Insights into the modulation of age‐related macular degeneration: Molecular and systems‐level perspectives

ENGLISH SUMMARY

In developed countries, age‐related macular degeneration (AMD) is the leading cause of visual impairment in the aging population. Understanding of its pathogenesis is limited and the number of patients is predicted to continue rising. AMD is classified into two main forms: dry and neovascular AMD (dAMD and nAMD, respectively). Dry AMD is the most prevalent form (80–90%) of AMD cases and has no approved treatment in Europe. Neovascular AMD (10–20% of AMD cases) is treated with frequent intravitreal injections with medicines intervening with vascular growth, but unfortunately, treatment is not curative and only a portion of patients respond to the current treatments.

Increasing number of patients will create a challenge to the healthcare system and will underline the need to find either novel, more durable treatment options or ways to intervene with the disease course already in its early stages and prevent or delay disease escalation. Thus, understanding factors behind development of the disease and treatment response is important. Currently, the information on the factors influencing the progression and incidence of the disease is fragmented and does not provide sufficient tools for individualised patient care. In this study, our aim was to create a deeper understanding of factors modulating development and progression of AMD.

By analysing data from nAMD patients treated in Kuopio University Hospital, we were able to identify that heavy smoking and high BMI increase the probability for earlier emergence of nAMD, while the presence of intraretinal fluid (IRF) resulted in a fast‐progressing disease. We also observed a relationship between IRF and the appearance of subretinal drusenoid deposits, HRF and retinal atrophy. Thus, we observed that IRF plays a role in the escalation of the disease towards non‐reversible changes in retinal anatomy.

By analysing serum RNA profile of nAMD patients, we observed that the RNA profile of nAMD patients corresponds with anatomical changes observed in the retinal fluid compartments as well as atrophic changes in retinal layers. Especially, thioredoxin‐related transmembrane protein 4 (TMX4) was identified as a protein whose function during AMD development should be evaluated more in detail in the future.

Moreover, long‐term follow‐up of women living in the North Savo region revealed the role of systemic components during AMD progression and showed that people with long‐term corticosteroid and immunosuppressant use have an increased risk for developing AMD. Our results highlight that further research on how systemic components behind corticosteroid and immunosuppressant use modulate AMD progression is needed.

Thus, results of this study build more evidence from factors modulating AMD development and progression. Giving guidance to future analyses in a situation where comprehensive understanding from factors influencing AMD progression and development are lacking.

1. INTRODUCTION

Age related macular degeneration (AMD) is a progressive disease of the eye that affects the central part of the retina, macula, important for sharp and colour vision (Figure 1a). During AMD development degenerative changes are observed in multiple layers including neural, structural and vascular beds. Development of AMD may result in a blurred, distorted or no vision in patient's central vision (Figure 1b). Changes may complicate reading and recognising faces, creating challenges for a patient to manage everyday tasks. Destructive effects of AMD are most prominent in the macula, which is responsible of high‐resolution and colour vision due to high cone density (Vemala et al., 2017; Kaarniranta et al., 2020; Fleckenstein et al., 2024). AMD is the one of the most common causes of irreversible vision loss and blindness (i.e. best corrected visual acuity (BCVA) < 0.3 in better eye) among the elderly population in developed countries (Flaxman et al., 2017; Tuuminen et al., 2017). Early and intermediate stages of AMD do not affect the vision, but during the development of late‐stage AMD photoreceptor degeneration leads to loss of central vision. Photoreceptors are specialised neurons that convert light into electrical signals, which are further processed in the visual cortex to form images (Fleckenstein et al., 2024).

FIGURE 1.

Anatomy of the human eye (Kivinen, 2017) (A) and visual representation of normal and AMD vision (B, Pictures from Hossa, Finland).

AMD is classified as early, intermediate and late‐stage AMD. Drusen size and pigmentary changes separate AMD patients with early or intermediate AMD from healthy individuals. Drusen are extracellular deposits that accumulate in the outer retina, comprising materials such as lipids, minerals and proteins. Small drusen (<63 μm) are normal changes during aging and can be present also in eyes of healthy people (Ferris et al., 2013; Hyttinen et al., 2024). Early AMD is usually asymptomatic, even though accumulation of intracellular lipofuscin in the retinal pigment epithelium (RPE) and small extracellular drusen between RPE cells and Bruch's membrane (BrM) can be clinically detected (Figure 1) (Flaxel et al., 2020). Patients with early AMD have medium sized drusen (≥63–<125 μm) and have no pigmentary abnormalities. Instead, people with intermediate AMD have medium drusen accompanied with large drusen or pigmentary abnormalities (Ferris et al., 2013; Hyttinen et al., 2024). Late‐stage AMD is classified into two forms based on clinical features; dry AMD (dAMD) and neovascular AMD (nAMD). DryAMD is the most prevalent form, accounting 80–90% of all AMD cases (Flaxel et al., 2020). It is characterised by outer retinal atrophy, also known as geographic atrophy (GA), whereas in nAMD neovascularisation associates with subretinal and/or intraretinal fluid accumulation and is also known as exudative AMD. The prevalence of early‐stage AMD among Europeans is 3.5% among people aged 55–59 years and 17.6% among those aged ≥85 years. Instead, for late AMD prevalence was 0.1% for people aged 55–59 years and 9.8% for those aged ≥85 years. For all people aged ≥70 years, the prevalence was 13.2% for early AMD and 3.0% for late AMD. In same meta‐analysis prevalence of early and late AMD seemed to be similar between genders, with exception that women ≥85 seemed to have higher prevalence for late AMD (Colijn et al., 2017). Moreover, in recent Finnish publication incidence and prevalence of nAMD were higher in women (Purola et al., 2025).

Neovascular AMD is characterised by newly formed leaky vasculatures. This neovascularisation is thought to be induced by increased expression of growth factors such as vascular endothelial growth factor A (VEGF‐A). VEGF‐A can be released in response to stimuli including oxidative stress, hypoxia and complement activation. VEGF‐A promotes angiogenesis primarily by binding to its receptor VEGFR2 and activating downstream pathways that destabilise existing vasculature and lead to angiogenesis initiation (Helotera & Kaarniranta, 2022). Immediate visual changes in nAMD are caused by edema created by newly formed leaky blood vessels. This may further cause subretinal and intraretinal fluid accumulation, haemorrhages and fibrosis. In addition to term nAMD, this condition is also known as exudative AMD. Neovascular lesions can be divided into three groups: type 1–3 macular neovascularisation (MNV). Type 1 MNV is characterised by neovascularisation beneath the RPE, whereas in type 2 vasculature has broken through the RPE layer into the neuroretina. Both type 1 and type 2 MNV originate from choroidal vasculature. Instead, in type 3 MNV newly formed vessels originate from the layers of neuroretina (Mathis et al., 2023). It is known that treatment outcomes vary between lesion types. Type 1 lesions are associated with better long‐term outcomes but often require more intense anti‐vascular endothelial growth factor dosing. Compared to other lesion types, type 1 MNV is also 6.7 times less likely to progress to GA (Xu et al., 2015; Mathis et al., 2023). Instead, when type 1 and type 2 MNV are compared, type 2 typically responds quickly to treatment but develops more often fibrotic scars, which increases the risk for poor visual outcomes (Daniel et al., 2018; Mathis et al., 2023). It has been suggested that type 1 MNV would precede type 2 MNV as for 6–12 months follow up 23% of initial type 1 lesions were transformed into type 2 lesions (Schneider et al., 2005). Type 3 MNV are prone to be highly sensitive to anti‐VEGF treatment, which may be due to shorter distance between lesion and vitreous, and thus better exposure to treatment. However, type 3 lesions are associated with a high incidence of GA during long‐term follow up (Grunwald et al., 2015; Mathis et al., 2023).

During the development of dAMD, one of the manifestations of GA is the development of areas of atrophy involving photoreceptors and RPE. Recently, classifications and nomenclatures for different levels of atrophy were provided: incomplete or complete outer retina atrophy (iORA or cORA) and incomplete or complete RPE and outer retinal atrophy (iRORA or cRORA) (Spaide et al., 2020). From these, iORA represents an early phase in retinal atrophy with only architectural thinning in the outer nuclear layer (Spaide et al., 2020). In iORA, discontinuous loss of ellipsoid zone (inner segments of the photoreceptors) has occurred, and the interdigitation zone (layer located between the outer segments of photoreceptors and RPE) is not visible. Instead, in cRORA, the affected region is at least 250 μm wide, and in that zone, absence of the RPE is associated with choroidal hypertransmission, and the findings of cORA can be seen (Spaide et al., 2020). Thus, iORA and cRORA represent different ends in atrophy development.

Diagnosis of AMD includes tests for visual acuity, clinical examination and multiple imaging tools. Visual acuity assessment is often performed by charts showing letters arranged in progressively smaller lines. The clinical diagnosis typically involves biomicroscopy of the ocular fundus (surface at the back of the eye), which may reveal drusen, pigmentary changes, and RPE atrophy, as well as signs of exudation, haemorrhages, and fibrosis (Figure 2). Imaging tools such as colour fundus photography, optical coherence tomography (OCT), and fluorescein angiography facilitate diagnosis and management of AMD. OCT provides 3‐dimensional visualisation of ocular structures and angiography detects vascular networks and potential neovascularisation. Angiography can be done either with intravenously administered fluorescein dye or by OCT‐A, which uses laser light reflectance of the moving red blood cells to visualise vessels (Fleckenstein et al., 2024).

FIGURE 2.

Clinical findings during AMD development. Fundus photographs (upper panel) and optical coherent tomographs (lower panel) from healthy individual and different stages of AMD. Pigment mottling (black arrows), drusen (white arrows), GA (blue arrows), choroidal neovascularisation and haemorrhage (red arrows), intra‐ (yellow arrow) and subretinal (green arrow) fluids. In, GA neural retina (black column) and the RPE layer (asterisk) are atrophic or missing. BrM (red line) is not optically recognised right when pathology increases and thus it starts to wave. AMD, age‐related macular degeneration; BrM, Bruch's membrane; Ch, choriocapillaris; d, dry; GA, geographic atrophy; n, neovascular and RPE, retinal pigment epithelium. (Figure originally published in Helotera and Kaarniranta (2022) under CC BY licence).

Regardless of the higher prevalence of dAMD, the majority of visual impairments are due to nAMD as the disease is more aggressive and faster progressing (Flaxel et al., 2020). The management of nAMD has dramatically improved during the last two decades as therapies targeting VEGF (vascular endothelial growth factor) were introduced to the market (Flaxel et al., 2020). These treatments limit neovascularisation, normalise existing vasculature and reduce vascular leakage, which results in a significant delay of vision loss. Current treatment options include ranibizumab, brolucizumab, aflibercept and faricimab. However, in addition to approved medications, bevacizumab is also frequently used. In addition to VEGF inhibition, faricimab and aflibercept also target other growth factors. Faricimab also binds to angiopoietin 2 (ANG2), whereas aflibercept also binds to placental growth factor (PlGF) and VEGFB (Helotera & Kaarniranta, 2022).

In general, safety profile of intravitreal injections is considered favourable. The most common side effects with intravitreal anti‐VEGF inhibitors include, for example, conjunctival haemorrhage, eye pain and temporal intraocular pressure increase. However, rarer inflammatory related events are the ones which cause more concern, such as endophthalmitis, vasculitis, uveitis and vascular occlusions. Vison threatening Intraocular inflammation (IOI) and occlusion events are currently under discussion as abnormally high levels of these complications have been seen with recently approved brolucimab (Dugel et al., 2020; Helotera & Kaarniranta, 2022; Witkin et al., 2020, 2021). Interestingly, patients with brolucizumab induced IOI have been reported to be more often women. In addition, patients with antibrolucizumab antibodies have been reported to have higher rates of IOI events (6%) than patients without these antibodies (2%) (Witkin et al., 2020). Pre‐existing antibrolucizumab antibodies are observed in approx. 50% and treatment emerged antibodies in approx. 75% of the patients (Kim & Woo, 2024). Recent study proposed that nAMD associated blood–retina barrier (BRB) breakdown may enable transport of antidrug antibodies (ADA) from blood circulation to ocular tissues. They further demonstrated that ADA‐ brolucizumab complex may activate endothelial cells, trigger platelet aggregation, induce cytokine release and exacerbate immune response (Kearns et al., 2023). It is also known that even though retina is generally considered to have immune privilege, immune cells including antibody producing cells can enter this tissue under certain conditions (Bharadwaj et al., 2017). From other commonly used anti‐VEGF agents ADAs are relatively common for ranibizumab and faricimab, even though they are not at all as frequently observed as with brolucizumab (Kim & Woo, 2024). For example, in registration trials of faricimab treatment‐emergent antifaricimab antibodies were detected with 13.8% of nAMD patients and IOIs were observed to occur in 12.2% of ADA positive patients and 1.4% ADA negative patients (SPC, 2025). Rare cases of faricimab related sterile IOI has been reported also with real‐life patients. For example, bilateral IOI after faricimab injection was reported in five patients. After discontinuation of faricimab and resolution of inflammation, 4 eyes from 4 patients failed to recover and mean vision loss was 11.5 ± 8.6 early treatment of diabetic retinopathy study (ETDRS) letters. Authors concluded that real‐life risk for vision threatening IOIs may be higher than reported in trials. However, also real‐life reports with IOI incidence comparable to clinical trials have been published (Dhillon & Macleod, 2025). As development of ADAs towards therapeutic antibodies is recognised, commercially available kits already exist for ADA detection in routine use for some of the most common therapeutic antibodies (Alez‐Martin et al., 2025). However, these are not yet available and routinely utilised in the monitoring of AMD patients.

Neovascular AMD can be a bilateral disease. Treatment will start for the majority of patients initially only for one eye, and the second eye may follow later. It has been reported that during the first 2–5 years, 16–38% of patients will receive bilateral treatment (Barbazetto et al., 2010; Zarranz‐Ventura et al., 2014; Joachim et al., 2017; Lovestam Adrian et al., 2022). Furthermore, the severity of nAMD in the first eye predicts the severity of nAMD in the fellow eye (Gangnon et al., 2015). Considering that current therapies for nAMD may fail initially for subgroups of patients and eventually also the responders develop resistance and the disease can escape, there remains a need for novel modes of action to be introduced for the control of this complex disease.

Unfortunately, advances made in the treatment of dAMD are more limited as no treatments have been approved in the European Union (Flaxel et al., 2020). However, two treatments have been recently approved by the FDA. These treatments act by inhibiting complement function and reduce GA lesion growth rate. However, the clinical significance of this is questionable, as no clinically meaningful effect on the improvement or preservation of vision has been noted. From these, pegcetacoplan is an anti‐complement component 3 (C3) peptide and avacincaptad pegol is a complement factor C5 targeting ribonucleic acid (RNA) aptamer (Csaky et al., 2024).

Due to aging population, number of patients with AMD is predicted to increase. Estimations have projected that 288 million individuals globally will suffer from AMD in 2040 (Wong et al., 2014; Colijn et al., 2017; Li et al., 2020). Dramatic increase in nAMD patients in ophthalmology clinics has been already reported also in Finland (Helotera, Viita, & Laine, 2024; Karesvuo et al., 2020; Purola et al., 2025). Increase in the global burden of AMD will pose a burden on the healthcare system, but also for patients suffering from these disabling diseases. Encouragingly, the proportion of visual impairment has been observed to be reduced after 2006, which coincides with the emergence of anti‐VEGF treatment options for nAMD (Bloch et al., 2012; Claessen et al., 2012; Colijn et al., 2017).

Currently, information on the factors influencing the progression and incidence of AMD is fragmented and does not provide sufficient tools for individualised patient care. The general aim of this study was to build more evidence from factors modulating incidence and progression of AMD.

2. REVIEW OF THE LITERATURE

2.1. Aging and development of AMD

Aging could be described as a decline in the ability to respond and adapt to the different exposures. Twelve interdependent hallmarks of aging have been proposed: telomere attrition, epigenetic alterations, genomic instability, loss of protein homeostasis, disabled macroautophagy, deregulated nutrient‐sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, altered intercellular communication, chronic inflammation and dysbiosis. As these hallmarks are interconnected, the activation of one can also induce the appearance of others. This highlights the fact that aging‐related disorders should be considered as complex processes, which should be understood as a whole (Lopez‐Otin et al., 2023).

As in many age‐related diseases, AMD is also linked with mitochondrial dysfunction, altered proteostasis, disabled autophagy, altered intercellular communication, and low‐level chronic inflammation (Helotera & Kaarniranta, 2022; Kaarniranta et al., 2020; Kauppinen et al., 2016). In addition to aging, AMD is a multifactorial disease where normal homeostasis of the retina is disrupted by genetic factors, smoking and unhealthy diet (Anderson et al., 2002; Kauppinen et al., 2016; Mitter et al., 2014; Moon et al., 2012). During the development of AMD, retinal and vascular layers of the eye undergo degenerative changes leading to cell death in these compartments. Pathological changes of AMD can be observed in RPE, photoreceptors, choriocapillaris and BrM. Vision loss in AMD occurs due to the death of photoreceptors (Kaarniranta et al., 2020).

RPE cells, BrM and choriocapillaris regulate exchange of nutrient and oxygen between outer retina and choroid as well as removal of waste products from the outer retina (Borrelli et al., 2018; Hammadi et al., 2023). Passage of substances and immune cells from systemic circulation to retinal compartments is especially regulated by BRB. BRB can be divided into outer BRB (oBRB) and inner BRB (iBRB), which control fluid, nutrient and cellular entry in outer and inner retina, respectively. Understanding pathological processes in oBRB during AMD development is important as AMD originates at the site of oBRB (Tisi et al., 2021). Outer BRB consists of RPE cells, BrM and adjacent choroidal capillaries containing fenestrated endothelium. However, some sources do not count choroidal capillaries as a part of oBRB as they do not provide strong barrier by themself due to fenestrations (Dujardin et al., 2024; O'Leary & Campbell, 2021; Tisi et al., 2021).

Correct functioning of RPE cells, BrM and choriocapillaris maintain health of the photoreceptors, and failure in these functions leads to oBRB breakdown and following photoreceptor degradation in AMD (Kaarniranta et al., 2020; Tisi et al., 2021).

2.2. RPE

The RPE is a highly specialised monolayer of polarised epithelial cells, which appear dark brown due to their melanin content. Melanin provides protection towards ultraviolet light caused damage. The apical side of RPE cells is in contact with photoreceptor outer segments (POS), and its basal side adheres to BrM. Adjacent RPE cells are connected by tight junctions, which form major components in BRB. However, transport of nutrients, electrolytes and water between choroid and retina are needed. Thus, membrane pumps, endocytosis, passive diffusion and other mechanisms are used to serve this function and move substances through RPE layer (George et al., 2021; Yang et al., 2021). RPE cells are considered regeneration‐incompetent in mammals, whereas in non‐mammalian vertebrates they can regenerate after RPE injury. However, some level of RPE repair has been observed with animal models, for example with mice, rats and pigs (George et al., 2021). Interestingly, it has been shown that equatorial RPE cell density will degrease during aging, whereas foveal RPE cell density appears to be more stable over the same period (Gao & Hollyfield, 1992).

To maintain visual function, RPE cells play a role in several processes that maintain the health of photoreceptors and choriocapillaris. These processes include for example phagocytosis of POS, light absorption to optimise vision, vitamin A metabolism and secretion of growth factors. (Harrison, 2019; Yang et al., 2021).

Microvilli at the apical side of RPE cells surround POS, and RPE cells maintain renewal of photoreceptors by phagocytosis and degradation of dispatched POS. It is considered that in mammals each RPE cell will take care of 30 photoreceptors, and POS are renewed once every 10 days (Sajovic et al., 2022; Yang et al., 2021). Phagocytosed POS are degraded in lysosomes in the process called heterophagy. Reactive oxygen species (ROS) promote dysfunctions of these lysosomes, which leads to reduced capacity of lysosomes to process POS. Incomplete degradation of POS will further lead to accumulation of autofluorescent lipofuscin, which is considered a biomarker in AMD (Kaarniranta et al., 2020). It is shown that intralysosomal accumulation of toxic lipofuscin can further inhibit lysosomal degradative functions. It can also sensitise lysosomes to visible light, oxidative stress, jeopardise lysosomal stability, and lead to cellular apoptosis (Moreno‐Garcia et al., 2018).

Vitamin A plays an essential role in mammalian vision and lack of vitamin A will lead to visual dysfunction, especially to reduced night vision. The human body cannot synthesise vitamin A, thus it needs to be ingested. However, its daily intake is not needed as a fat‐soluble vitamin is stored in the body. The majority of vitamin A is stored in the liver, but it can be found also in adipose tissue, lungs and RPE. Thus, dietary vitamin A is absorbed in the intestine, stored and transported by blood circulation to RPE cells. RPE cells uptake vitamin A and after processing transport end products to photoreceptors to form rhodopsin, a light‐sensitive pigment containing receptor protein that triggers visual phototransduction in rods. Rhodopsin is formed when the 11‐cis‐retinal binds to opsin. During the phototransduction process light is converted to electrical signals. The first step in this process is the absorption of photons by the visual pigment rhodopsin, which causes isomerisation of 11‐cis‐retinal to all‐trans‐retinal. During the canonical visual cycle, occurring in rods, all‐trans‐retinal is continuously recycled back to the 11‐cis‐retinal in RPE cells. All‐trans‐retinal is released from opsin, on the luminal side of the photoreceptor membrane of rod discs to be transferred to RPE cells. Cone visual cycle instead, occurs in the Müller cells (Harrison, 2019; Sajovic et al., 2022). In AMD thickening of BrM leads to reduced entry of vitamin A to RPE, creating a situation resembling a hypovitaminosis A. Interestingly, it has been shown that daily oral supplementation of a combination of vitamin A precursor beta‐carotene, vitamin E, vitamin C and zinc can reduce the development of late AMD. However, the role of beta‐carotene supplementation alone could not be answered with this study (Age‐Related Eye Disease Study Research, 2001; Sajovic et al., 2022).

Another important function of the RPE cells is secretion of growth factors that support development and survival of the adjacent cells and sustain health of the choriocapillaris as well as retinal layers (Blaauwgeers et al., 1999; Dutt et al., 2010). For example, it has been shown in a cell culture model that conditioned media from RPE cells induced neuronal differentiation of photoreceptor or bipolar cells (Dutt et al., 2010). In addition, VEGF secretion from RPE cells is polarised, and high levels of VEGF are secreted from the basal side facing choriocapillaris. Indeed, VEGF receptors VEGFR‐2 and VEGFR‐3 can be found in choriocapillaris at the side facing the RPE cell layer. It is considered that VEGF secreted by RPE cells is important for the maintenance of the choriocapillaris and their fenestrations (Blaauwgeers et al., 1999). However, secretion of VEGF in physiological levels should be separated from the pathological overexpression of VEGF in nAMD, where excessive VEGF secretion leads to neovascularisation and fluid accumulation due to leaky vessels (Helotera & Kaarniranta, 2022).

The precise mechanism behind RPE degeneration, order during events in the onset and progression of AMD are not fully understood. However, malfunction of various processes, such as mitochondrial dysfunction, protein aggregation, increased ROS production, impaired autophagy and chronic inflammation, has been described (Helotera & Kaarniranta, 2022; Kaarniranta et al., 2020). It has been shown that cultured RPE cells extracted from healthy and AMD donors are functionally distinct. Extracted RPE cells from AMD patients accumulated lipid droplets, glycogen granules and showed disintegrated mitochondria's as well as increased autophagosomes. Moreover, RPE cells from AMD donors were more sensitive to oxidative stress, had decreased mitochondrial activity and impaired autophagy (Golestaneh et al., 2017). It is known that alterations in mitochondria of RPE cells emerge due to aging, but in AMD patients these alterations appear 10–15 years earlier than in control specimens (Feher et al., 2006). Aging is also associated with increased damage in mitochondrial deoxyribonucleic acid (DNA), but significantly more damage is observed when AMD patients are compared with age matched controls. It is considered that excess of intracellular ROS production plays an important role during mitochondrial damage and decreasing autophagy (Kaarniranta et al., 2020; Somasundaran et al., 2020). Even though ROS are byproducts of the normal cellular metabolism, excessive ROS can lead to oxidative stress and induce cellular damage. RPE cells have high metabolic demand and are enriched with mitochondria, which are one source for high levels of ROS. Also, POS phagocytosis and processing exacerbate ROS production in RPE cells (Murali et al., 2020). In general, oxidative stress is considered to induce autophagic flux in RPE cells, but chronic oxidative stress seems to lead to decreased lysosomal activity and autophagy as well as increased lysosomal lipofuscin. Under physiological conditions, ROS levels are controlled by the antioxidant system. Thus, oxidative stress is caused If ROS levels exceed cells antioxidant capacity of the cell. Exact reasons behind ROS build up are still unclear, but theories include for example exposure to visible light, lipofuscin accumulation and cigarette smoking (Kaarniranta et al., 2020; Somasundaran et al., 2020).

Autophagy is a self‐clearance process that removes damaged or dysfunctional proteins and cellular components through lysosome dependent mechanism and thus maintains cellular homeostasis. In RPE cells autophagy plays an essential housekeeping role as RPE cells are long‐lived, have undergone terminal differentiation and thus do not divide or divide only rarely. It is considered that dysfunctional lysosomal clearance leading to accumulation of waste materials plays a key role during AMD development. Also autophagy in late AMD occurs at a lower rate than in the early AMD. It is considered that in late AMD, an increased amount of damaged organelles overload autophagy mechanism and lead to dysfunctionality. Impairment of autophagy in RPE cells can thus lead to detrimental accumulation of damaged organelles and abnormal or toxic proteins. Disturbed proteostasis is characterised by the accumulation of toxic lysosomal lipofuscin in the RPE and extracellular drusen deposits between the RPE and choriocapillaris. (Hyttinen et al., 2025; Kaarniranta et al., 2020; Somasundaran et al., 2020). In addition to AMD, accumulation of fluorescent lipofuscin is associated also with many other human diseases including for example Alzheimer's and Parkinson's diseases. Lipofuscin is formed by oxidised proteins, lipids and sugars. Composition varies between tissues but is observed to include oxidised proteins (30–70%), lipids (20–50%), metals cations (Fe3+, Fe2+, Cu2+, Zn2+, Al3+, Mn²⁺, Ca²⁺) (2%) and sugar residues. Interestingly, lysosomes cannot degrade lipofuscin once it is formed and it cannot be cleared by exocytosis (Moreno‐Garcia et al., 2018). Phagocytosis and degradation of POS is one of the key functions of the RPE, thus the proper function of the lysosomes is essential (Kaarniranta et al., 2020). It has been shown that lipid peroxidation in POS leads to reduced lysosomal degradation of POS, lysosomal dysfunction, increased lipofuscinogenesis and degreased autophagy in RPE cells (Krohne et al., 2010).

Thus, mitochondrial damage together with oxidative stress will lead to decreased intracellular recycling and degradation of proteins in aging RPE. This will result in protein damage and aggregation that further induces inflammation. For example, oxidative stress can lead to oxidatively damaged proteins, lipids and DNA with novel oxidation‐specific epitopes. These novel structures can induce a pro‐inflammatory response, and it is known that these modifications are found in RPE cells. As a short‐term cascade, acute inflammation is a beneficial process that leads to the clearance of pathogens and repair of damaged tissues. However, prolonged chronic inflammation can lead to tissue injury. Indeed, complement and inflammasome activation has been linked especially to the development of dAMD (Datta et al., 2017; Kaarniranta et al., 2020). Interestingly, it has been shown that in cultured RPE cells mitochondrial damage can lead to nuclear gene expression changes including increase in expression of genes involved in outer segment phagocytosis, components of drusen, components of the extracellular matrix, genes associated with lipid transport and genes associated with a response to inflammation or oxidative stress. Intriguingly, one of the upregulated genes was VEGF, which is linked to both inflammation and neovascularisation (Miceli & Jazwinski, 2005).

2.3. BRM and drusens

Transport of nutrients, oxygen and visual cycle by‐products between RPE and choriocapillaris is regulated by BrM. 2–4 μm thick BrM is a collagen and fibrin rich connective tissue that provides structural support for RPE cells and creates a selective barrier between RPE and choroid. Structurally BrM has three layers: inner and outer layers are collagenous, whereas the layer in between is elastic. Also, the choroidal endothelial basement membrane and RPE basement membrane are sometimes counted as part of BrM. Composition of BrM membrane is essential to its function and changes may lead to dysfunction (Bergen et al., 2019; Borrelli et al., 2018; Clark et al., 2025; Hammadi et al., 2023). Reduced elasticity, increased calcification and BrM thickening due to lipid and protein accumulation resulting in reduced permeability is associated with aging. Created diffusion barrier will further accelerate accumulation of biomaterials (Bergen et al., 2019; Huang et al., 2008; Moore & Clover, 2001). Composition of BrM is normally maintained by extracellular matrix turnover, which is altered during aging and may contribute to observed age‐related thickness (Clark et al., 2025; Hussain et al., 2010; Kumar et al., 2010; Murali et al., 2020). It has been also reported that BrM‐RPE complex is thicker in early AMD than in control patients (Brandl et al., 2019).

In addition to BrM thickening, drusen that accumulate between the RPE and BrM affect the nutrient flow between choroid and RPE. Drusen are yellowish lipid and protein rich deposits of extracellular debris, whose accumulation will eventually lead to functional loss of the retinal photoreceptors. In clinic they can be observed as yellow spots in colour fundus images or as dome shaped accumulations under the RPE in OCT. Drusen are defined by their location, size, autofluorescence and colour. Small round drusen with well‐defined edges with size <63 μm are called hard drusen. Whereas the size of intermediate and large drusen is 63–125 μm and >125 μm, respectively. Soft drusen represent ill‐defined borders (Bergen et al., 2019; Flaxel et al., 2020; Spaide & Curcio, 2010). Drusen should be separated from the subretinal drusenoid deposits (SDD) as they are located between the retina and RPE. SDD were originally termed ‘reticular pseudo‐drusen’ and regardless of the location are aggregations sharing similarities with druse‐associated molecules (Hyttinen et al., 2025; Spaide & Curcio, 2010).

The largest single component of drusen is lipids and especially esterified cholesterol, unesterified cholesterol, and phosphatidylcholine. Nearly 150 proteins have been identified from drusen, including components of the complement system, vitronectin, apolipoproteins E and B. In addition to lipids and proteins, drusen have been shown to contain trace elements, including zinc, iron and calcium. The composition of drusen is not homogeneous, suggesting a heterogenous origin of these deposits (Bergen et al., 2019; Spaide & Curcio, 2010). Even though drusen can also be seen in normal eyes, their deposition is considered the hallmark of AMD. Moreover, the accumulation of large drusen increases an individual's risk of developing advanced AMD. However, drusen accumulation is not unique to AMD but is also observed in other disorders such as Alzheimer's disease and renal diseases. The development of drusen is poorly understood, but hypotheses about the origins of the components include POS, RPE, choroidal and blood. Based on drusenomics studies, it looks like a large portion of the proteins originate from blood circulation. Next comes RPE, and only a minor portion would originate from POS or choroid (Bergen et al., 2019). Interestingly, it has been suggested that drusen accumulation may result partly due to endothelial dysfunction. Moreover, histological and pathological similarities have been observed in atherosclerotic plaques and drusen. For example, both accumulations are associated with changes in proteoglycan turnover, the presence of low‐level inflammation, and a noticeable high portion of drusen‐associated proteins can also be found in atherosclerotic plaques (Bergen et al., 2019). Due to the complex nature of both diseases, it is difficult to say whether there is a common mechanism behind AMD and atherosclerotic cardiovascular diseases, but increasing evidence suggests a relationship between these diseases (Mauschitz & Finger, 2022). However, when genetic overlap was analyzed for 34 AMD‐associated genetic loci and cardiovascular diseases, increased AMD risk was associated with reduced cardiovascular disease risk (Grassmann et al., 2017).

It is not currently fully understood how molecules will be recruited to and retained in drusen. By current understanding, average sized proteins, e.g. 53 kDa molecules, can easily diffuse through BrM. Whereas migration of macromolecules is slower or limited (Bergen et al., 2019). It has been shown that BrM permeability for proteins decreases 10‐fold from the first to ninth decade of life. In same study, size‐exclusion limit for elderly membranes was 100–200 kDa (Moore & Clover, 2001). Whereas size‐exclusion limit for young BrM has been suggested to be 180–500 kDa. However, many proteins found in drusen are average size and should be able to diffuse through BrM, if only size would be considered. Thus, it is likely that proteins become entrapped to drusen by forming aggregates (Bergen et al., 2019).

The mechanism of drusen formation is not completely understood. It has been speculated that, for example, chronic inflammation, oxidative stress, stressed autophagy and changes in choriocapillaris may play roles in drusen formation (Anderson et al., 2002; Gurubaran et al., 2025; Hyttinen et al., 2025; Mullins et al., 2011; Rabin et al., 2013; Wang et al., 2009). For example, increased drusen density was shown to be associated with reduced vascular density in choriocapillaris. However, this study did not show causality of the effect. Choriocapillaris are maintained by VEGF secretion from RPE, so drusen may interrupt this signalling and destabilise choroidal capillaries. Alternatively, vascular loss may impair removal of the debris and waste products from the sub‐RPE space and thus lead to drusen formation. The same study also showed increased leukocyte accumulation in early AMD samples when compared to unaffected controls (Mullins et al., 2011). Indeed, subretinal accumulation of immune cells, especially around drusen, GA lesions and choroidal neovascularisation, have been observed in AMD (Fletcher, 2020; Guillonneau et al., 2017). In addition, a recent study done with a mouse model demonstrated that recruitment of memory T cells in the choroid and outer retina occurs in early phases of AMD development and precedes degenerative changes (Sturzbecher et al., 2025). As subretinal space physiologically lacks immune cells, these cells are infiltrated from other compartments. Mounting evidence supports blood origin for these infiltrating immune cells (Guillonneau et al., 2017).

Once formed, drusen can probably escalate disease progression as they increase distance between choroidal vasculature and RPE, which will result as reduced nutrient and waste product enhance. This may lead, for example, to hypoxia, which will further escalate disease progression, for example by stimulating VEGF production. Also, drusen material has been proven to be pro‐inflammatory. Drusen can enhance inflammation, for example through amyloid structures, and inflammasome pathways (Anderson et al., 2002; Khan et al., 2016; Lipecz et al., 2019; Liu et al., 2013). In addition, C3a, C5a, CFB and membrane attack complex (MAC) complexes found in drusen can increase expression of VEGF and the formation of choroidal neovascularisation (CNV, Figure 3) (Gold et al., 2006; Kunchithapautham & Rohrer, 2011; Nozaki et al., 2006).

FIGURE 3.

Vascular wall in homeostasis and during angiogenesis. VEGF, vascular endothelial growth factor; ANG, angiopoietin. (Figure originally published in Helotera and Kaarniranta (2022) under CC BY licence).

2.4. Choroid

Choroid is a vascularised layer that provides blood supply for the outer retina and consists of three vascular layers including: the choriocapillaris, the Sattler's vascular layer with medium vessels and the Haller's vascular layer with large vessels. Choriocapillaris are part of oBRB and are a highly anastomosed meshwork of fenestrated capillaries located in the innermost choroid, just under the BrM and RPE (Hammadi et al., 2023). Fenestrated nature of the choroidal vessels makes them much leakier than non‐fenestrated vessels (Bergen et al., 2019).

Fenestrations in choriocapillaris are maintained by the VEGF, which is produced from the basolateral side of the RPE cells. Choroidal capillaries supply nutrients and remove waste products from outer retinal layers (Marneros et al., 2005; O'Leary & Campbell, 2021). Interestingly, the choriocapillaris are further polarised as fenestrations locate on the retinal side of the vessels and pericytes are found on the outer wall facing the sclera. This composition facilitates efficient movement of macromolecules between the retina and choroid (Kim et al., 2020).

Under physiological conditions, 85–90% of oxygen demand of outer retina is sourced by choriocapillaris, while the remaining 10–15% is acquired from the deep retinal capillary plexus. Retina has high metabolomic activity and in the outer retina, most of the oxygen is consumed in the metabolically active photoreceptor inner segments. Choroid is perfused at a high rate to respond to the retina's high oxygen consumption. Thus, the proper functioning of choriocapillaris is critical to support RPE and the rest of the retina (McHugh et al., 2019).

Choroidal endothelial cells form lumen of the capillaries and are lined with a thin layer of basal lamina (Figure 3). Pericytes that partially surround the choroidal capillaries are also invested in basal lamina. It has been shown that during aging pericyte damage precedes endothelial changes (Nag et al., 2021). Thickness of the choroid, choriocapillaris density and diameter as well as choroidal blood flow are shown to be reduced during aging (Emeterio Nateras et al., 2014; Ramrattan et al., 1994). Moreover, reduction in choriocapillaris density, diameter and blood flow is even more prominent when AMD patients are compared to age matched controls (Grunwald et al., 1998; Ramrattan et al., 1994). It has been suggested that reduced blood flow would impair delivery of oxygen and nutrients as well as removal of waste products and thus contribute to disease progression (Emeterio Nateras et al., 2014). Interestingly, low choroidal perfusion was identified as a risk factor for CNV development in a study which followed CNV development in the fellow eye of patients with unilateral nAMD (Boltz et al., 2010).

It has been shown that during AMD development choroidal atrophy precedes RPE degeneration (Biesemeier et al., 2014). Moreover, a linear correlation between choroidal dropout and loss of RPE has been shown in GA, and in the same study choroidal atrophy was evident in nAMD samples without RPE atrophy. It was suggested that during nAMD RPE cells become hypoxic due to choroidal dropout, leading to upregulation of VEGF production from RPE cells and further stimulating choroidal neovascularisation (Figure 3) (McLeod et al., 2009).

It is considered that tight junctions between choroidal endothelial cells loosen during aging, which can result in increased permeability and extravasation of macromolecules from choroidal vasculature (Nakanishi et al., 2016). Increased permeability can further lead to extravasation of fluid and immune cells to and lead to low level inflammation in surrounding tissue microenvironment (Figure 4) (Helotera & Kaarniranta, 2022). Indeed, it was recently reported that aged choroidal samples show higher expression of genes encoding pro‐inflammatory transcription factors and proteins facilitating leukocyte transendothelial migration than their comparators (Voigt et al., 2020). In addition, abnormal complement activation has been associated with AMD and polymorphisms in complement factor H and other complement factors have been found to be a risk factor in AMD. Complement factor H is an inhibitor of complement, whose levels have been shown to be reduced in the choroid of AMD patients. Complement activation can lead to the formation of a pro‐lytic pore (MAC), which has been detected on the choriocapillaris of AMD patients (Mulfaul et al., 2022). Also, other studies have linked low‐level inflammation to the degeneration of choroids in AMD, for example the number and activation of mast cells have been identified to be increased in the choroid during all phases of AMD. It is suggested that release of proteolytic enzymes during mast cell activation could play a role in choroidal thinning and fibrosis during AMD development (Bhutto et al., 2016). Also, fibrosis observed in wet AMD can be a consequence of low‐level inflammation, vascular leakage and neovascularisation (Helotera & Kaarniranta, 2022).

FIGURE 4.

Development of age‐related macular degeneration. (a) Old age, light exposure, smoking, high‐fat diet and unknown factors elevate processes that will lead to development of AMD. (b) Angiogenic switch at nAMD (c) Fibrosis in nAMD can be a product of defective and excessive wound healing response. Pro‐inflammatory cytokines may promote differentiation and activation of matrix‐producing myofibroblasts (e.g., EndMT and EMT). RPE, retinal pigment epithelium; EMT, epithelial to mesenchymal transition; C3a, complement component 3a; C5a, complement component 5a; CFB, complement factor B; EMT, epithelial–mesenchymal transition; EndMT, endothelial–mesenchymal transition; MAC, membrane attack complex; MMT, macrophage‐mesenchyman‐transition; VEGF‐A, vascular endothelial growth factor A. (Figure originally published in Helotera and Kaarniranta (2022) under CC BY licence).

Interestingly, it has been reported that decreased choriocapillaris density correlates with increased drusen formation. Thus, it could be considered that proper functioning of choriocapillaris is important to prevent drusen formation. In advanced nAMD, findings from choroidal thickness are controversial, which can be affected by underlying neovascularisation. Interestingly, regions with choroidal neovascularisation are often surrounded by nonperfused choriocapillaris regions. Thus, it can be speculated that hypoxia can at least partly regulate the location of neovascular lesion development (Lipecz et al., 2019; McHugh et al., 2019).

2.5. Prognostic, predictive and risk factors in AMD

Rate of disease progression in AMD can vary between individuals. Thus, understanding prognostic and predictive factors can be helpful in development of personalised treatment approaches and in finding a most efficacious treatment path for each patient. Prognostic factors can help to estimate outcomes of the disease, and predictive factors can help to determine which patients respond to specific treatments. Instead, understanding risk factors can facilitate the development of preventive strategies.

High age is the strongest risk factor for AMD development, but other risk factors have been identified as well including environmental factors such as smoking and diet. Also, genetic factors have been shown to be involved. Studies have also explored various factors related to nAMD development or progression including both local retinal and systemic molecular regulators (Hagag et al., 2024; Heesterbeek, Lores‐Motta, et al., 2020). Interestingly regular exercise and higher education seem to have protective effects against nAMD. Whereas obesity has been linked to disease progression. Also, intake of vitamin A, vitamin B6, β‐carotene, lutein, zeaxanthin, magnesium, copper, docosahexaenoic acid and omega‐3 fatty acid has been shown to have protective role (Agron et al., 2021; Hagag et al., 2024; Heesterbeek, Lores‐Motta, et al., 2020; Ristau et al., 2014).

OCT is currently a key tool for predicting and evaluating treatment responses. Changes in OCT can indicate onset or progression of AMD even before patients can detect concrete changes in vision. Interestingly, OCT biomarkers that have been associated with poor visual outcomes include intraretinal fluid (IRF), width and height of subretinal hyperreflective material and fibrovascular pigment epithelial detachment (Nawash et al., 2023). Moreover, high baseline drusen volume, presence of reticular pseudodrusen, SDDs, baseline photoreceptor layer thinning, irregular choroidal vessels and hyperreflective foci (HRF) are associated with disease progression (Hagag et al., 2024; Heesterbeek, Lores‐Motta, et al., 2020; Nawash et al., 2023). Moreover, fluctuations in central subfield thickness (CST), thickness of the retina measured in the central 1 mm area around the fovea, seem to affect outcomes. Higher fluctuations indicate worse visual outcomes and may have associations with fibrosis and GA (Evans et al., 2020; Sheth et al., 2022).

HRF also serves as a predictive marker during anti‐VEGF treatment as their number is reduced in patients responsive to treatment. Markers related to higher demand of frequent anti‐VEGF injections include presence of IRF, vitreomacular adhesion and vitreomacular tractions (Nawash et al., 2023). There is also evidence that treatment outcomes vary between lesion types. Type 1 lesions show better long‐term outcomes than other MNV types but require more intense anti‐vascular endothelial growth factor dosing. It is also less likely that type 1 MNV progress to GA (Xu et al., 2015; Mathis et al., 2023). In recent analysis, eyes with type 2 and 3 MNV were associated with resolution of IRF and SRF after three faricimab loading injections (Han et al., 2025). However, generally when type 1 and type 2 MNV are compared, type 2 typically responds quickly to treatment but develops more often fibrotic scars, which increases risk for poor visual outcomes (Daniel et al., 2018; Mathis et al., 2023).

Many genetic markers have been associated with AMD progression, but more research is needed to clarify current data and understand which of the factors are most relevant. Currently, mutations rs10922109 and rs570618 in CFH, rs116503776 in C2/CFB/SKIV2L, rs3750846 in ARMS2/HTRA1 and rs2230199 in C3 variants are more consistently reported (Heesterbeek, Lores‐Motta, et al., 2020). Finding genetic prognostic factors that would enable early intervention would be highly beneficial to delay nAMD onset. Also, being able to identify patients who are at risk for fast disease progression, need more intensive treatment approaches or benefit from specific treatment would facilitate the development of personalised treatment approaches.

2.6. Role of systemic factors in AMD

AMD is traditionally considered a local disease. However, early changes during AMD progression can be seen in choroidal capillaries, which are exposed both to local and systemic components. Studies about inflammatory markers have revealed associations with serum proteins and AMD. For example, studies with components of the complement system have shown that both systemic and local sources are involved (Armento et al., 2021; Heesterbeek, Lechanteur, et al., 2020; Mulfaul et al., 2022). It has been shown that indicators of complement activation including FHR‐4, C3a, C3d, Ba, C5a and SC5b‐9 are significantly elevated in plasma from AMD patients compared to controls (Cipriani et al., 2020; Scholl et al., 2008). Also increased serum levels of IL‐6 are shown to be associated with late AMD and particularly with GA and low circulating levels of TNF‐α have been shown to be associated with increased visual acuity after anti‐VEGF therapy (Nahavandipour et al., 2020; Rozing et al., 2020). Moreover, increased systemic C‐reactive protein (CRP) levels are linked to late AMD, highlighting the role of systemic low‐level inflammation in AMD progression (Feng et al., 2020). In addition to findings on circulating inflammatory proteins, also results from drusenomics studies support the involvement of systemic factors in AMD development, as they have shown that large portion of the drusen proteins originate from the blood circulation (Bergen et al., 2019).

In addition to serum proteins, there is an increasing evidence to suggest that systemic risk factors such as smoking, hypertension, cardiovascular disease, lupus erythematosus, Sjogren's syndrome, Crohn's disease, giant cell arteritis and obesity are linked with AMD (Chen et al., 2014; Helotera, Siintamo, et al., 2024; Loukovaara et al., 2022; Moir et al., 2023; Velilla et al., 2013; Vujosevic et al., 2023; Xu et al., 2020; Zhang et al., 2016). Interestingly, patients with AMD are at increased risk of conditions outside of the eye, such as dementia, Alzheimer's and kidney diseases, indicating some common mechanism driving these diseases (Jung et al., 2023; Tsai et al., 2023).

As accumulating evidence suggests that circulating factors may modulate AMD progression, a question should be asked: Can orally ingested compounds affect AMD progression? Interestingly, there is evidence that orally taken medicines and dietary supplements may modulate AMD progression, for example regular use of dietary supplements containing non‐provitamin A carotenoids lutein and zeaxanthin have shown to reduce risk of late AMD (Age‐Related Eye Disease Study 2 Research et al., 2014; Sajovic et al., 2022).

Accumulating evidence also suggests that use of systemic treatments such as metformin and non‐steroidal anti‐inflammatory drugs may reduce risk of AMD (Aggarwal et al., 2024). However, results from studies done with aspirin are controversial (Yan et al., 2022). Interestingly, it was recently proposed that levodopa can reduce new onset nAMD and may reduce the frequency of needed anti‐VEGF injections (Hyman et al., 2023). However, observed results can be mediated due to the local action of the systemically administered medication as levodopa can cross the blood‐retina barrier (Motz et al., 2020). Also, use of the anti‐coagulant dabigatran has been linked to a reduced risk for new nAMD (Akter et al., 2022). Indeed, we recently showed that patients using anti‐coagulation medication receive nAMD diagnosis later than patients without medication (Helotera, Siintamo, et al., 2024). Therefore, the increasing evidence suggests that management of systemic conditions or systemically administered treatments may modulate the progression of AMD.

3. AIMS OF THE STUDY

Currently, information on the factors influencing the progression and incidence of AMD is fragmented and does not provide sufficient tools for individualised patient care. The general aim of this study was to build more evidence from factors modulating the incidence and progression of AMD.

The specific aims of this study are:

1. To explore factors affecting the progression of nAMD and identify predictive factors that can estimate the duration of intravitreal treatments. (Publication I)

2. To determine if serum RNA profile can distinguish control patients from nAMD patients and evaluate if the RNA profile of nAMD patients reflects anatomical changes observed during AMD progression. (Publication II)

3. To analyse whether long‐term use of systemic medications, including corticosteroids, immunosuppressants, diabetes and antithrombotic medications are associated with the development of AMD. (Publication III)

4. SUBJECTS AND METHODS

4.1. Ethics

All studies were performed according to the ethical standards of the Declaration of Helsinki and informed consent was obtained from all subjects involved in the studies. Patient information was collected and stored in accordance with the European Union's Regulation 2016/679.

The Ethics Committee of the Kuopio University Hospital has approved the studies with approval numbers 42/2014 (Publication I), 42/2014 (Publication II), 237/13.00/2024 (initial approval 28.10.1986, Publication III) and 221/13.02.00/2015 (initial approval 7/2014, 31.1.2014, Publication III). Considering the Publication III, permissions to register linkages were granted by the Finnish Social and Health Data Permit Authority Findata (Dnro THL/5436/14.06.00/2024, 12.09.2024).

4.2. Study subjects and blood samples

All three studies were done utilising patient data and Publication II included also patient and cadaver samples. When considering nAMD patients, the study subjects of Publication II were a subgroup of the patients of Publication I. Control group in Publication II included patients who were undergoing cataract surgery and who showed no signs of retinal degeneration. In Publication I and II, all study subjects were treated, and samples were collected at the Department of Ophthalmology at the Kuopio University Hospital (KUH). Study subjects in Publication III were independent groups of individuals. However, due to the same geographical region, it is likely that there exists some overlap between study populations in all considered publications (I‐III).

In Publication I and II, all nAMD patients (n = 421 in Publication I and 60 in Publication II) were subjected to examination in the Department of Ophthalmology at the KUH, including biomicroscopy examination, fundus photographs (Canon CX‐1 Retinal Camera, Canon, Tokyo, Japan) and/or optical coherence tomography (OCT) (SPECTRALIS OCT2, Heidelberg Engineering, Heidelberg, Germany) and, when needed, fluorescein angiography (FAG) (Canon CX‐1). Inclusion criterion for nAMD patients in these studies was intravitreal treatment for nAMD at KUH during the years 2007–2021. Only patients with enough information for each corresponding analysis were included. As these studies included analysis of retrospective real‐world data, there was variance in the recording of the individual variables. Thus, not all data was available from all patients. One eye per patient was selected for the study. In case only one eye had nAMD, it was picked as the study eye. If both eyes had nAMD, the eye with a longer treatment history was picked unless it was found to exhibit signs of fibrotic late‐stage nAMD. FAG and OCT analysis were carried out by retina experts.

In Publication II, blood samples from 60 nAMD patients and 64 controls were obtained at the KUH. Serum samples were collected between 2015 and 2019. For serum extraction, 10 mL of blood were collected to BD Vacutainer® Clot Activator Tubes (Becton Dickinson, Franklin Lakes, NJ, USA), gently inverted 10 times and allowed to sit for 30–60 min at room temperature. The serum was separated after centrifugation (3200  g at 20°C for 15 min) and transferred into two smaller Mekamini tubes (Mekalasi, Helsinki, Finland). The samples were stored at −70°C until sequencing. Data on retinal properties were gathered for nAMD patients only.

In Publication III, the study population consisted of a pooled cohort of two follow‐up studies, including the Kuopio Osteoporosis Risk Factor and Prevention (OSTPRE) Study and Kuopio Fall Prevention Study (KFPS), totaling 16 518 women living in the North Savo region, eastern Finland who were followed 1993 onwards. The cohort included Caucasian women born between 1932 and 1946 with a mean baseline age of 54.9 (SD 3.8) years. The OSTPRE and the KFPS studies have been described in detail previously (Rikkonen et al., 2023; Sund et al., 2014). The analysis utilised self‐reported health questionnaires for baseline height, weight, smoking and menopause age. The beginning of the year 1993 formed the baseline for the analysis. Depending on the outcome of interest, follow‐up was terminated on the date of death, AMD diagnosis, or at the end of the registry period on 31 December 2021. Information on time of death was obtained from the National Causes of Death Register. Women with dAMD were identified with a diagnosis code H35.30 and nAMD patients with a code H35.31 (https://icd‐codes.info/fi/VII/H30‐H36/H35/). AMD cases were identified from national and local health care registers using diagnosis codes (ICD‐10: H35.30 and H35.31, ICD‐9: 3625, ICD‐8: 37710). All AMD diagnoses were recorded from the hospital discharge register (HILMO).

In Publication III, women were classified as long‐term medication users after receiving annual reimbursements for prescription costs over a period of five consecutive years. The medications were identified with Anatomical Therapeutic Chemical (ATC) codes from the Finnish Social Insurance Institution's national drug reimbursement registry (KELA). The long‐term medications of interest were categorised by using the respective ATC codes for: diabetes medication (A10) and diabetes medications containing metformin (A10BA02, A10BA01, A10BA03, A10BD02, A10BD03, A10BD05, A10BD07, A10BD08, A10BD10, A10BD1, A10BD13–18, A10BD20, A10BD22–23, A10BD25–27), antithrombotic agents (B01, B01AA, B01AB, B01AC, B01AE, B01AF and B01AX), immunosuppressants (L04) and systemic corticosteroids (H02).

4.3. Anti‐VEGF treatments

Anti‐VEGF injections in Publication I and II were carried out with the modified pro re nata protocol according to the Finnish nAMD guidelines (Tuuminen et al., 2017). Treatment was started with three monthly loading intravitreal bevacizumab injections and bevacizumab treatments were continued as long as satisfactory treatment responses were observed. Ranibizumab or pegaptanib were primarily used during 2007–2008 and the use of aflibercept started in 2012. Patient eyes were imaged using OCT on their monthly visits at the early stage of nAMD. Follow‐up visit intervals were gradually extended by 2 to 4 weeks once their macula exhibited dry state signals.

4.4. RNA sequencing

For Publication II, serum samples were processed and sequenced at the Institute for Molecular Medicine Finland (FIMM) Genomics NGS Sequencing unit at the University of Helsinki, supported by the Helsinki Institute of Life Science and Biocenter Finland. RNA extraction was conducted at the HiPREP Core at the FIMM and sequenced at the FIMM Genomics NGS Sequencing unit at the University of Helsinki, backed by the Helsinki Institute of Life Science (HiLIFE) and Biocenter Finland. Total RNA was extracted (serum volume 1 mL) using the Maxwell® RSC 48 Instrument and Maxwell RSC miRNA Plasma and Serum Kit (Promega, Madison, WI, USA) following the manufacturer's instructions. The protocol included a DNase treatment. RNA was eluted in 50 μL of RNase‐free water (35–40 μL recovered), and sample quality and quantity were determined using Agilent Bioanalyzer 2100 with the RNA 6000 Pico Kit (Agilent Technologies, Santa Clara, CA, USA). The obtained RNA quality and yield were deemed good for a serum extraction.

Library preparation from 800 ng of total RNA was performed according to the Illumina Stranded Total RNA with Ribo‐Zero Plus Reference Guide (Illumina, San Diego, CA, USA). Library quality checks were performed using the LabChip GX Touch HT High Sensitivity assay (PerkinElmer, Shelton, CT, USA). The libraries were quantified for sequencing using the KAPA Library Quantification Kit (KAPA Biosystems, Wilmington, MA, USA). Sequencing was carried out with Illumina NovaSeq6000 system using S4 flow cell with lane divider (Illumina, San Diego, CA, USA). The data yield from sequencing was satisfactory within sample type constraints and the read length for the paired‐end run was 2 × 151 bp. The RNA‐seq datasets were quality‐checked and pre‐analysed according to FIMM‐RNAseq 2.0.7 workflow (19). The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE273435 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE273435) (20). Data validation to determine if RNA‐seq data correlates with RT‐PCR analysis is performed and described by Liukkonen et al. (2024).

4.5. Staining of cadaver samples

For Publication II, paraffin‐fixed human cadaver samples were stained with CHMP6 (1:100, ab235050, Abcam) and TUBGCP3 (1:100, Sigma Aldrich, HPA043913). Before incubation with the primary antibody, sections were deparaffinised with xylene (3 × 10 min) and rehydrated through an ethanol series (Abs EtOH 2 × 5 min, 94% EtOH 2 × 5 min, 70% EtOH 5 min, distilled H₂O 5 min). TUBGCP3 sections were pretreated with TRIS buffer (5 min, 90°C) and washed with TBS (pH 7.4, 2 × 2 min). Endogenous Peroxidase and Alkaline Phosphatase were blocked with BLOXALL (Vector laboratories, SP‐6000‐100, 10 min). Sections were next washed with TBS (5 min) and blocked with 2.5% horse serum (20 min). Primary antibodies were incubated overnight at +4°C. On the next day, sections were further washed for 5 min with TBS and incubated with the secondary antibody (ImmPRESS polymer kit AP Horse anti‐rabbit, MP‐5401, Vector Laboratories, 30 min). Next, sections were washed 2 × 5 min with TBS. Signal development was performed using the Vector® Red Substrate Kit, Alkaline Phosphatase (SK‐5100), followed by wash with TBS (5 min), 2× dips indistilled H₂O, a dip in haematoxylin and 3× wash with distilled H₂O. Sections were placed in TBS before mounting, which was carried out with Aquamount H‐5501.

Imaging of the samples was performed at the Biomedicum Imaging Unit, University of Helsinki, with the support of the Helsinki Institute of Life Science (HiLIFE) and Biocenter Finland. Images were acquired using a Zeiss Axio Imager microscope with 40× magnification. The microscope contains a Zeiss AxioCam 105 colour camera for transmitted light imaging and uses Zeiss Zen 2 software.

4.6. Statistical analysis

For Publications I‐III, Statistical analyses were explorative and descriptive; the studies did not aim to confirm or reject predefined hypotheses. Numerical values were compared using two‐tailed t‐test, Mann–Whitney U test, Pearson Chi‐square test or analysis of covariance (ANCOVA) analyses. A significance was considered if p < 0.05. Correlations were checked using multiple regression analysis. Calculations were performed using SPSS software (versions 27.0.1.0 and 29.0.2.0) or Excel 2016.

In Publication I, continuous variables are described by absolute values and as changes from baseline per analysis time point. All data reported were only for patients with assessments at each of the indicated time points. Furthermore, only patients receiving treatment each year have participated in calculations on the median number of injections.

In Publication II, the acquired count data was filtered to separate protein‐coding and non‐coding RNAs using data sourced from Ensembl BiomaRt, which were then filtered in R Statistical Software 4.1.3 with DESeq2 1.34.0 to include a minimum of 10 reads in at least 7 patients (21–23). Differentially expressed genes were selected by identifying genes with at least a 2‐fold change (downregulated if foldchange in median ≤0.5 and upregulated if foldchange in median ≥2) and at least a numerical difference of 1 in median values and statistical significance. RNAs were eligible for disease progression analysis only if they were differentially expressed between nAMD and control samples. Library size normalised counts per million values were used and were acquired with edgeR. A false discovery rate (FDR) adjustment was made using the Benjamini–Hochberg method. Integrative analysis of the lncRNAs, sncRNAs and mRNAs was performed with Ingenuity Pathway Analysis (Qiagen).

Given the high variance variability between the samples in Publication II, data was further checked by the bootstrap‐boosted ANCOVA adjusted for selected confounders, including gender, body mass index (BMI), smoking status, and for the use of blood pressure medication, anti‐cholesterol medication, anti‐coagulation medication and anti‐aggregation medication. Additionally, univariate significance tests were performed. One sample was excluded from the ANCOVA analyses due to a missing BMI value.

In Publication III, the time‐dependent Cox proportional hazards models were employed in survival analysis with their respective 95% CI to account for the impact of long‐term medication on AMD incidence over time.

5. RESULTS

5.1. Association of baseline characteristics to disease course and treatment duration in nAMD

In Publication I, aim was to explore factors affecting the progression of nAMD and identify predictive factors that can estimate the duration of intravitreal treatments. Study included 421 nAMD patients treated at the Kuopio University Hospital (KUH) during years 2007–2021. Patients were treated according to the Finnish pro re nata regimen and 32.5% of patients were males. Median age at treatment initiation was 79 years. Median baseline Visual acuity (VA) was 0.25 (mean 0.3) and 12.7% of patients (n = 47) had VA under the detection limit at baseline. 60.7% of patients remained on treatment during the 2nd treatment year, 44.8% during 3rd treatment year and 4.9% during 10th treatment year. During the 1st treatment year median number of injections was 5 and mean VA gain was 0.5 rows (snellen equivalent, median 0 rows).

To analyse whether the baseline factors were linked with disease progression we analysed the association of baseline medication, smoking status and BMI to patient age at the time of nAMD diagnosis (age at first anti‐VEGF injection, Table 1). In this patient population, use of anti‐coagulation (78.1 vs. 81 years, no med. vs. med. resp., p = 0.001), anti‐aggregation (78 vs. 79.7 years, no med. vs. med. resp., p = 0.024), and blood pressure medication (76.9 vs. 79.2 years, no med. vs. med. resp., p = 0.032) appeared to be associated with later onset of nAMD. Whereas patients with smoking history (76 vs. 80 years, active smokers vs. non‐smokers, p < 0.001) and high BMI (76.5 years vs. 79.4 years, BMI ≥30 vs. BMI <25 resp., p = 0.006) appeared to be diagnosed earlier for nAMD than their comparators. The relationship of diagnosis age of nAMD with smoking status, anti‐coagulation medication, BMI, anti‐aggregation medication, and blood pressure medication was further analysed using multiple regression analysis. 310 patients had data from all these variables and were thus included in the analysis. In multiple regression analysis, 15.7% of the variation in the age of nAMD diagnosis could be explained by these factors. All the chosen factors remained highly statistically significant except the data on blood pressure medication.

TABLE 1.

Association of baseline medication, body mass index and smoking to treatment initiation.

		Patients, n	Age at anti‐VEGF initiation		p‐value
			Mean	STDEV
Anti‐coagulation medication	No medication	262	78.1	7.5	0.001
	Medication in use	62	81.0	5.5
Anti‐aggregation medication	No medication	184	78.0	7.9	0.024
	Medication in use	139	79.7	6.3
Anti‐cholesterol medication	No medication	164	78.3	8.0	0.322
	Medication in use	157	79.1	6.4
Blood pressure medication	No medication	70	76.9	8.0	0.032
	Medication in use	257	79.2	7.0
BMI	<25	136	79.4	7.3	<25 vs. ≥30; p = 0.006
	25 to <30	80	79.3	7.2
	≥30	66	76.5	7.1
Smoking	Non‐smoker	253	80.0	7.0	Non vs. Random p = 0.016*; Non vs. heavy p < 0.001
	Random smoking	56	77.3	7.5
	Heavy smoking	58	76.0	7.0

Note: Significant values are bolded.

Presence of IRF is more detrimental to the vision than is the SRF. Thus, we further analysed relationships of retinal fluid status and treatment duration in our data set. We observed that the presence of IRF (Figure 5) at baseline was associated with shorter treatment duration. For example, during the third year of treatment, 30.4% of IRF‐only patients and 75% of SRF‐only patients remained on treatment (p = 0.003). Moreover, we observed that patients with only IRF at baseline were older than patients with only SRF at baseline (median 87 vs. 73.5 resp. p = 0.001). Also, baseline VA differed significantly between patients with IRF and only SRF (p = 0.0001, 0.26 vs. 0.5 for IRF vs. SRF‐only). However, patients with IRF at baseline showed better VA development during the first treatment year (baseline IRF +1 row vs. SRF‐only 0 rows, p = 0.015). Our observations are consistent with the previous observations that when type 1–3 MNVs are compared, type 2 and 3 typically respond quickly to treatment but are at increased risk for poor visual outcomes (Daniel et al., 2018; Foss et al., 2025; Mathis et al., 2023). SRF is shown to be the most common fluid in type 1, whereas IRF is frequently observed in type 2 MNV (Foss et al., 2025).

FIGURE 5.

Relationship between intraretinal fluid and disease severity. Percentage of patients remaining on treatment per treatment year and fluid status. IRF, intraretinal fluid; SRF, subretinal fluid.

As a clear effect of IRF on treatment duration was observed, we wanted to look more deeply at how the presence of IRF affects disease severity on an anatomical level. Thus, we compared the presence of sub‐RPEF, HRF, drusen, SDDs and the level of atrophy at baseline (before treatment initiation) between the SRF only and IRF only or IRF all patients. Interestingly, more HRFs and more severe atrophy (cRORA 20.7% vs. 5%, IRF all vs. SRF only, p = 0.020) were observed when IRF was present (Table 2). Also, more SDDs were observed in patients with IRF only than in patients with only SRF (IRF only vs. SRF only: 43.5% vs. 15%, p = 0.039; IRF all vs. SRF only: 31% vs. 15%, p = 0.104).

TABLE 2.

Presence of OCT biomarkers in IRF (SRF+/−) and SRF only patients.

Number of patients, n		Baseline		p‐value
		IRF (+/−SRF)	SRF (no IRF)
		87	20
SRF % (n)		73.6 (64)	100 (20)
Sub‐RPEF % (n)		82.8 (72)	100 (20)	5.8E‐05
Retinal thickness, median μm (STDEV)		412 (162)	380 (137)	0.31
Peak retinal thickness, median μm (STDEV)		506 (166)	435 (163)	0.15
Cuticular drusen, % of patient		6.9	5	0.74
Subretinal drusenoid deposit, % of patients		31.0	15	0.10
HRFs. % of patients	0–3	6.9	5	5.5E‐04
	4–24	66.7	95
	25–49	18.4	0
	50 or more	8.0	0
Drusens. % of patients	Small	12.6	5	0.18
	Medium	0	5
	Large	0	10
	Small and medium	12.6	3.4
	Small and large	16.1	35
	Medium and large	2.3	0
	Small, medium and large	56.3	30
Atrophy, % of patients	No atrophy	1.1	5	0.46
	iORA	26.4	50	0.07
	cORA	21.8	20	0.88
	iRORA	24.1	15	0.1
	cRORA	20.7	5	0.02
	RPE tear	4.6	10	0.57

Note: Significant values are bolded.

Abbreviations: cORA, complete outer retinal atrophy; cRORA, complete RPE and outer retinal atrophy; HRF, hyper reflective foci; iORA, incomplete outer retinal atrophy; IRF, intraretinal fluid; iRORA, incomplete RPE and outer retinal atrophy; RPE, retinal pigment epithelium; SDDs, Subretinal drusenoid deposits; SRF, subretinal fluid; sub‐RPEF, subretinal pigment epithelial fluid.

5.2. Serum RNA profile reflects fluid status and atrophic retinal changes in nAMD

In Publication II, our aim was to gain understanding of the circulating RNA profile of nAMD patients. Serum RNA samples from 60 nAMD patients and 64 control patients were analysed. Mean age at sampling was 76.0 years (67.2% females) for controls and 79.2 years for nAMD patients (72% females). To identify differentially expressed RNAs, we followed two approaches to control false leads. In one approach, we used the Benjamini–Hochberg method to adjust differentially expressed RNAs for the false discovery rate. In another approach, we created an initial list of differentially expressed RNAs and combined it with information about disease progression markers. To create an initial list, differentially expressed RNAs were defined as those having at least a 2‐fold change and a minimum numerical difference of one in median values alongside statistical significance when RNAs from nAMD patients and controls were compared. RNAs were eligible for disease progression analysis only if they were differentially expressed between nAMD and control samples. Initial list included 563 mRNAs, 32 lncRNAs, and two sncRNAs that were downregulated in nAMD samples. Only five mRNAs and no lncRNAs or sncRNAs were upregulated in nAMD samples based on our criteria. All the previously mentioned RNAs, later referred to as the initial list, were taken for disease progression analysis.

To evaluate the crosstalk of lncRNAs, sncRNAs and mRNAs on the pathway level, we performed an integrative analysis of the molecules with ingenuity pathway analysis (Qiagen, Hilden, Germany, Figure 6a,b). Based on the expression patterns of the molecules, the analysis inferred significant downregulation of gene expression and translation of related terms, enriching pathways such as translation initiation and elongation, thereby explaining the trend of general downregulation of the molecules. Also, pathways important in the regulation of the vasculature, such as the RHO GTPase cycle, mTOR pathway, ROBO receptors, and protein kinase A, were observed to be altered. Overall, the molecular interactions resulted in increased necrosis and apoptosis, and decreased cell viability, movement, and cellular organisation, reflecting degenerative changes observed during AMD.

FIGURE 6.

mRNAs altered in nAMD vs. control serum. (a) Most significantly altered biological functions between nAMD and control samples based on serum RNA analysis. Blue indicates downregulation in nAMD and the stronger the colour, the stronger the association. (b) Most significantly altered biological functions in pathway analysis. Blue indicates downregulation and orange colour upregulation in nAMD. The colour intensity indicates association strength.

After false discovery rate adjustment using the Benjamini–Hochberg method, two downregulated lncRNAs and 27 downregulated mRNAs from the initial list remained significant, providing a list of RNAs with a lower likelihood of false leads about differentially expressed RNAs between nAMD and control patients (Table 3).

TABLE 3.

RNAs significant after FDR adjustment. * One value is missing in the nAMD group.

RNA type	Gene name	nAMD (n = 60)		Control (n = 64)		FDR
		Average	STDEV	Average	STDEV
lncRNA	SNHG1	427.2	1105.2	1887.7	2276.2	0.014
	FAM211A‐AS1	20390.3	14738.9	31776.7	15030.0	0.021
mRNA	ZNF292	100.4	84.8	175.1	84.0	0.026
	CYB561A3	55.0	64.1	112.3	75.8	0.026
	AFF4	59.4	62.3	111.9	68.9	0.026
	PDCD6	17.0	26.6	47.1	46.0	0.026
	EXOC6B	57.9	60.0	106.6	64.8	0.036
	NKTR	32.9	47.2	90.6	102.8	0.043
	GGNBP2	26.9	43.2	61.3	51.8	0.043
	TMX4	44.0	53.0	88.4	69.2	0.043
	FEM1B	27.6	40.4	59.4	48.2	0.043
	CHMP7	51.6	60.4	101.5	76.8	0.043
	SCML4	19.8	35.2	56.8	60.1	0.043
	ADK2	53.6	56.2	104.9	81.8	0.043
	CSTF2T	19.4	34.9	46.2	39.9	0.043
mRNA	S1PR1	55.0	61.5	107.4	83.6	0.043
	CD96	14.2	29.2	46.8	53.9	0.043
	ANK3	10.4	23.3	36.5	42.3	0.043
	SP140L	10.3	21.2	31.9	35.0	0.043
	BZW2	10.7	24.2	34.0	38.3	0.043
	MDN1	44.9	57.1	98.6	92.7	0.050
	SH2D1A	11.3	25.3	36.8	45.3	0.051
	KCTD9	7.2	16.9	26.4	35.2	0.052
	GGA2	94.7	87.3	163.9	113.3	0.053
	AAK1	81.2	74.3	165.6	156.9	0.053
	TXLNA	41.8	49.6	77.3	53.7	0.053
	SETDB2	43.4	49.9	80.6	57.9	0.053
	LANCL1	16.7	29.1	43.1	46.3	0.053
	SIKE1	12.2	26.3	35.2	39.6	0.053

Abbreviations: nAMD, neovascular age‐related macular degeneration; BMI, body mass index; FDR, false discovery rate.

To analyse how the serum RNA profile changes during nAMD progression, the differentially expressed mRNAs in the initial list were further examined based on clinical markers and anti‐VEGF treatment (Table 4). We compared the following conditions: 1. Sampling before anti‐VEGF treatment vs. sampling during anti‐VEGF treatment; 2. Patients with only intraretinal fluid (IRF) vs. patients with only subretinal fluid (SRF); 3. Patients with iORA vs. patients with cRORA.

TABLE 4.

RNA profile differences in nAMD vs. control samples in disease progression analysis. (a) Altered RNAs detected based on anti‐VEGF analysis. (b) Altered RNAs detected based on SRF vs. IRF analysis. (c) Altered RNAs detected based on atrophy analysis. Adjusted values from the ANCOVA analysis are shown. This table contains only values for control vs. nAMD. Values for comparison of modifying agents can be found from original publication.

RNA type	Gene name	Ensembl ID	Adjusted means		Adjusted p‐value
			Control, N = 64	nAMD, N = 59
					Control vs. nAMD
a
lncRNA	SNHG29	ENSG00000175061	34896.0	20597.8	4.0E‐06
mRNA	SKAP1	ENSG00000141293	52.3	24.9	0.029
	PAG1	ENSG00000076641	81.0	41.8	0.003
	GRSF1	ENSG00000132463	81.4	49.2	0.010
	CDK5RAP2	ENSG00000136861	63.5	44.4	0.020
b
lncRNA	FTX	ENSG00000230590	4468.7	2260.0	0.016
	NINJ2‐AS1	ENSG00000177406	2479.0	1590.7	0.071
mRNA	CRCP	ENSG00000241258	34.2	30.6	0.159
	TUBGCP3	ENSG00000126216	60.4	29.8	0.012
	CENPT	ENSG00000102901	76.5	63.7	0.011
mRNA	TGM2	ENSG00000198959	63.0	46.4	0.032
	ANGPT1	ENSG00000154188	35.0	30.2	0.047
	TMX4	ENSG00000125827	98.8	48.9	<0.001
	CEP162	ENSG00000135315	44.5	19.0	0.008
	CHMP6	ENSG00000176108	79.4	54.6	0.142
	GPR65	ENSG00000140030	39.7	19.1	0.014
	R3HDM1	ENSG00000048991	58.4	39.9	0.005
	STARD3	ENSG00000131748	50.1	26.1	0.005
	TMEM64	ENSG00000180694	69.5	56.1	0.124
	PREPL	ENSG00000138078	37.6	23.7	0.013
	TAB3	ENSG00000157625	61.8	54.0	0.016
	CDC14B	ENSG00000081377	296.7	259.8	0.023
	CDC42BPA	ENSG00000143776	59.0	40.9	0.029
	ARMCX3	ENSG00000102401	57.8	35.7	0.013
	CAMKK2	ENSG00000110931	128.8	86.1	0.011
	ZBTB37	ENSG00000185278	48.5	20.8	0.006
	SPATA13	ENSG00000182957	99.0	73.9	0.020
c
mRNA	CMTR1	ENSG00000137200	44.0	17.6	0.013
	CD28	ENSG00000178562	38.2	15.7	0.094
	EIF6	ENSG00000242372	41.7	21.3	0.025
	CEBPG	ENSG00000153879	31.0	25.0	0.061
mRNA	FLNB	ENSG00000136068	43.5	41.8	0.061
	PIH1D1	ENSG00000104872	38.5	18.4	0.020
	DUSP16	ENSG00000111266	45.7	32.9	0.007
	MIOS	ENSG00000164654	25.5	11.5	0.007
	RBM28	ENSG00000106344	39.4	22.3	0.011
	PPIL2	ENSG00000100023	56.7	22.5	0.003
	EIF1AX	ENSG00000173674	81.6	48.0	0.069
	ARHGEF7	ENSG00000102606	73.0	46.4	0.022

Note: Significant values are bolded.

Abbreviations: ANGPT1, angiopoietin 1; ARHGEF7, Rho guanine nucleotide exchange factor 7; ARMCX3, armadillo repeat containing X‐linked 3; CAMKK2, calcium/calmodulin‐dependent protein kinase kinase 2; CD28, CD28 molecule; CDC14B, cell division cycle 14B; CDC42BPA, CDC42 binding protein kinase alpha; CDK5RAP2, CDK5 regulatory subunit associated protein 2; CEBPG, CCAAT enhancer binding protein gamma; CENPT, centromere protein T; CEP162, centrosomal protein 162; CHMP6, charged multivesicular body protein 6; CMTR1, cap methyltransferase 1; cORA, complete outer retinal atrophy; CRCP, CGRP receptor component; cRORA, complete RPE and outer retinal atrophy; DUSP16, dual specificity phosphatase 16; EIF1AX, eukaryotic translation initiation factor 1A X‐linked; EIF6, eukaryotic translation initiation factor 6; FDR, false discovery rate; FLNB, filamin B; FTX, FTX transcript XIST regulator; GPR65, G protein‐coupled receptor 65; GRSF1, G‐rich RNA sequence binding factor 1; iORA, incomplete outer retinal atrophy; IRF, intraretinal fluid; iRORA, incomplete RPE and outer retinal atrophy; lncRNA, long non‐coding ribonucleic acid; MIOS, meiosis regulator for oocyte development; mRNA, messenger ribonucleic acid; nAMD, neovascular age‐related macular degeneration; NINJ2‐AS1, NINJ2 antisense RNA 1; PAG1, phosphoprotein membrane anchor with glycosphingolipid microdomains 1; PIH1D1, PIH1 do‐main containing 1; PPIL2, peptidylprolyl isomerase‐like 2; PREPL, prolyl endopep‐tidase‐like; R3HDM1, R3H domain containing 1; RBM28, RNA binding motif protein 28; RPE, retinal pigment epithelium; SKAP1, Src kinase‐associated phosphoprotein 1; SNHG29, small nucleolar RNA host gene 29; SPATA13, spermatogenesis‐associated 13; SRF, subretinal fluid; STARD3, StAR‐related lipid transfer domain containing 3; TAB3, TGF‐beta activated kinase 1 (MAP3K7) binding protein 3; TGM2, transglutaminase 2; TMEM64, transmembrane protein 64; TMX4, thioredoxin‐related transmembrane protein 4; TUBGCP3, tubulin gamma complex component 3; ZBTB37, zinc finger and BTB domain containing 37.

By comparing samples taken before the initiation of anti‐VEGF treatment (n = 23) to samples collected during anti‐VEGF treatment (n = 34), we discovered that four mRNAs and one lncRNA were downregulated in the samples taken during anti‐VEGF treatment (Table 4a). They were also downregulated when nAMD and control samples were compared. None of the sncRNAs were differentially expressed in baseline vs. anti‐VEGF analysis.

Interestingly, in the IRF only group, 19 downregulated mRNAs, one upregulated mRNA, and two downregulated lncRNAs were identified when patients with only IRF (n = 18) were compared to patients with only SRF (n = 21) at the time of sampling (Table 4b). None of the short ncRNAs were differentially expressed in the IRF only vs. SRF only analysis.

When patients with iORA (n = 11) were compared to patients with cRORA (n = 17), 10 mRNAs were observed to be downregulated, and two were upregulated in the cRORA group (Table 4c). All atrophy levels, including cRORA and iRORA, were included in the ANCOVA analysis. In the ANCOVA analysis, all identified mRNAs were downregulated in nAMD groups compared to control groups. None of the lncRNAs or short ncRNAs were differentially expressed either in the iORA vs. cRORA analysis.

As high variability was observed between the samples, we explored whether gender, body mass index (BMI), smoking status, or the use of blood pressure, anti‐cholesterol, anti‐coagulation or anti‐aggregation medications could modify observed RNA levels as confounding variables. One sample was excluded from the ANCOVA analyses due to the lack of a BMI value. All the tested variables occurred to be confounding factors for some specific RNAs (Table 5). However, none of the tested variables applied for all RNAs and we could not identify confounding variables for all tested RNAs. Thus, based on our analysis each tested RNA had their own set of confounding variables, if any.

TABLE 5.

Summary of confounding factors for RNAs influenced by anti‐VEGF treatment (a), RNAs affected by atrophy level (b), and for RNAs identified when samples with IRF and SRF were compared (c).

a
Confounding variable	Number of RNAs affected
	Control vs. nAMD	Baseline vs. anti‐VEGF
Gender	0/5	1/5
BMI	2/5	1/5
Smoking	0/5	1/5
Blood pressure	1/5	1/5
Anti‐cholesterol	0/5	3/5
Anti‐coagulants	2/5	3/5
Anti‐aggregation	0/5	2/5

b
Confounding variable	Number of RNAs affected
	Control vs. nAMD	Atrophy status
Gender	1/22	3/22
BMI	2/22	3/22
Smoking	1/22	2/22
Blood pressure	1/22	1/22
Anti‐cholesterol	1/22	1/22
Anti‐coagulants	1/22	1/22
Anti‐aggregation	0/22	0/22

c
Confounding variable	Number of RNAs affected
	Control vs. nAMD	SRF vs. IRF
Gender	1/12	3/12
BMI	1/12	2/12
Smoking	0/12	4/12
Blood pressure	1/12	2/12
Anti‐cholesterol	3/12	0/12
Anti‐coagulants	0/12	1/12
Anti‐aggregation	2/12	1/12

Abbreviations: nAMD, neovascular age‐related macular degeneration; IRF, intraretinal fluid; SRF, subretinal fluid.

Taken together, to obtain a deeper understanding of serum RNA changes, we followed two approaches to control false leads. In one approach, we used the Benjamini–Hochberg method to adjust differentially expressed RNAs for the false discovery rate. In another approach, we combined our initial list of differentially expressed RNAs with information about disease progression markers. The aim of the latter step was also to evaluate the link between disease progression and the serum RNA profile. When results from both analyses were combined, interestingly, only the RNA that was differentially expressed by both approaches encoded the protein thioredoxin‐related transmembrane protein 4 (TMX4).

5.3. The impact of systemic long‐term medications for the development of age‐related macular degeneration

In Publication III, our aim was to determine whether long‐term use of systemic medications, including diabetes and antithrombotic medications, corticosteroids, and immunosuppressants, is associated with the development of AMD. The study included 16 518 women living in the North Savo region, eastern Finland. Follow‐up was for the period between 199 and 2021, and the mean follow‐up time was 27.9 (SD 7.0) years. The mean age at baseline was 54.8 (SD 3.8, range 46–61). The all‐cause mortality during the follow‐up was 36.8% (n = 6076). During the follow‐up, 7.8% of women (n = 1308) were diagnosed with AMD. The mean age at diagnosis of dAMD was 76.9 years (SD 6.1, n = 847, 64.7% of all AMD diagnoses) and 78.1 years (SD 5.7, n = 461, 35.2% of all AMD diagnoses) at the diagnosis of nAMD.

For further analyses, patients were divided into the long‐term medication users (medication in use >5 years) and their respective controls (no use or medication in use <5 years, Table 6). Table 6a shows how patients in control groups are divided into non‐users and those who had medications in use under 5 years. Patients with diabetes medication in use (n = 1988) had a mean of 12.8 (SD 8.5) years of medication history during the follow‐up. Altogether, 1501 patients with diabetes had over 5 years of medication and were considered long‐term users in the analysis. All patients with anti‐thrombotic medication (n = 8027) had a mean of 8.4 (SD 7.1) years of medication history during the follow‐up, including 5596 patients with over 5 years of medication, considered long‐term users. Instead, patients with any use of systemic corticosteroids (n = 5751) had a mean of 6.6 (SD 8.2) years of medication history during the follow‐up, including 2382 long‐term users. During the follow‐up, patients with immunosuppressants (n = 796) had a mean of 7.1 (SD 6.8) years of medication history. Out of these, 407 are long‐term users. Some overlap between categories of long‐term medication usage was present in the cohort. The proportion of women with simultaneous long‐term use in two, three or four categories of medications was n = 1970 (11.8%), n = 300 (1.8%), and n = 14 (0.1%), respectively.

TABLE 6.

Age at diagnosis of AMD. (a) Age at diagnosis of non‐users, short term users and long‐term users. (b) Prevalence of AMD, age‐adjusted hazard ratio and mean age at diagnosis among long‐term medication users (use >5 years) and their respective controls (no use or use <5 years).

a
	Category of use	n, AMD/total (AMD prevalence, %)	Mean age at AMD, (SD)	p‐value* (AMD age)
Diabetes medication	Non‐user	1160/14 681 (7.9)	77.4	0.754
	Short‐term in use, <5 years	14/336 (4.1)	76.5
	Long‐term use, ≥5 years	135/1501 (9.0)	77.1
Antithrombotic medication	Non‐user	489/8491 (5.8)	76.4	<0.001
	Short‐term in use, <5 years	193/2431 (7.9)	77.8
	Long‐term use, ≥5 years	627/5596 (11.2)	77.9
Corticosteroids	Non‐user	887/12 474 (7.1)	77.4	0.599
	Short‐term in use, <5 years	147/1662 (8.8)	76.9
	Long‐term use, ≥ 5 years	275/2382 (11.5)	77.4
Immuno‐suppressants	Non‐user	1227/16 002 (7.7)	77.3	0.543
	Short‐term in use, <5 years	26/317 (8.2)	78.6
	Long‐term use, ≥ 5 years	56/407 (13.8)	77.5

b
		AMD prevalence, % (95% CI) during follow‐up	HR (p, vs. control)	Mean age at AMD (years, SD)	p (age, vs. control)
Study population		7.8 (7.4–8.2)		77.4 (6.0)
Diabetes medication	Medication (Any)	9.0 (7.7–10.3)	1.075 (0.488)	77.1 (6.0)	0.754
	Metformin	9.2 (7.7–10.3)	0.879 (0.401)	78.4 (6.1)	0.293
	Other that metformin	8.7 (6.9–10.5)	1.107 (0.518)	75.3 (5.4)	0.026
	Control	7.8 (7.4–8.2)		77.3 (6.0)
Antithrombotic medication	Medication	11.2 (10.5–11.9)	1.145 (0.042)	77.9 (5.7)	<0.001
	Control	6.2 (5.7–6.7)		76.1 (6.4)
Corticosteroids	Medication	11.5 (10.7–12.3)	1.348 (<0.001)	77.4 (5.9)	0.994
	Control	7.3 (6.8–7.8)		77.4 (6.0)
Immuno‐suppressants	Medication	13.8 (11.3–16.1)	1.623 (0.004)	77.5 (5.4)	0.875
	Control	7.7 (7.3–8.1)		77.3 (6.0)

Note: ANOVA Tukey's test between categories in a, and Pearson Chi‐square and t‐tests were used for statistical analysis in b. Significant values are bolded.

Abbreviations: AMD, age‐related macular degeneration; HR, hazard ratio; SD, standard deviation.

Interestingly, the mean age at diagnosis of AMD was later among antithrombotic medication users compared to control group (77.9 vs. 76.1, respectively, p < 0.001) and earlier among patients using other diabetes medications than metformin (75.3 vs. 77.1, p = 0.026, Table 7). Overall, diabetes medications did not affect the mean age at the time of AMD diagnosis or cumulative prevalence of AMD. However, significant difference in AMD diagnosis age was observed when long‐term non‐metformin users were compared to diabetes control group.

TABLE 7.

AMD incidence in age‐matched cohorts with 5‐year increments. The control group includes women with no long‐term medication at the age of interest.

	Age matching group, years	AMD incidence during next 5‐years: Control group vs. long‐term, %	p‐value
Diabetes medication	65	0.7 vs. 0.7	0.926
	70	1.6 vs. 2.4	0.076
	75	3.0 vs. 3.8	0.359
	80	3.7 vs. 3.3	0.687
	85	2.9 vs. 3.0	0.772
Antithrombotic medication	65	0.6 vs. 1.5	0.024
	70	1.7 vs. 1.6	0.976
	75	2.9 vs. 3.8	0.140
	80	3.6 vs. 3.8	0.240
	85	3.0 vs. 2.7	0.793
Corticosteroids	65	0.6 vs. 1.4	0.009
	70	1.2 vs. 2.2	0.076
	75	3.0 vs. 3.9	0.155
	80	3.5 vs. 4.6	0.010
	85	2.8 vs. 3.5	0.201
Immuno‐suppressants	65	0.6 vs. 4.8	<0.001
	70	1.6 vs. 2.5	0.398
	75	3.0 vs. 5.2	0.107
	80	3.6 vs. 6.4	0.004
	85	2.9 vs. 2.4	0.889

Note: Significant values are bolded.

Abbreviation: AMD, age‐related macular degeneration.

Prevalence for AMD was increased among long‐term users of systemic corticosteroids (HR 1.348 p < 0.001), immunosuppressants (HR 1.623 p = 0.004), and antithrombotic (1.145 p = 0.042) medication (Table 7). However, when the incidence of AMD during the 5 years period in age matched groups was examined, the results with antithrombotic medications were not as consistent as with corticosteroids and immunosuppressants. Age matched groups were created for ages 65, 70, 75, 80 and 85.

6. DISCUSSION

6.1. Summary

AMD is a progressive disease affecting the central vision and is a leading cause of severe visual impairment among elderly people in developed countries. Due to aging population, number of patients is expected to continue rising. During the last two decades there have been significant advances in the management of nAMD, allowing patients to maintain vision for a longer period (Fleckenstein et al., 2024; Wong et al., 2014). However, this has created increasing resource demand for the ophthalmology departments as patients who are on treatment are long term customers for these clinics (Helotera, Viita, & Laine, 2024). Nonetheless, current treatment options are not curative and only delay visual impairment (Fabre et al., 2022). Thus, detailed understanding from the mechanisms driving early disease development should be created to be able to find curative treatments or ways to do early intervention to delay disease onset. Moreover, when disease has been diagnosed and patient is on treatment, it would be beneficial if treatment paths could be individualised based on patient characteristics. Possibility to find most efficacious treatment for each patient and treatment schedules based on patients' needs could guide resources where they are truly needed. In this study, our aim was to increase understanding of factors modulating AMD incidence and progression.

In Publication I, the aim of the study was to explore factors affecting the progression of neovascular age‐related macular degeneration (nAMD) and identify predictive factors that can estimate the duration of intravitreal treatments. Explored factors included patient demographics, baseline medications and anatomical variables. Even though many factors have already been described as linked with the development and progression of nAMD, more research is needed to clarify current data and understand which of the factors are most relevant. Novel unexplored markers could provide tools to assemble this puzzle together. Finding prognostic factors that would enable early intervention to delay nAMD onset would be highly beneficial. In a similar vein, being able to identify patients who are at risk for fast disease progression and need more intensive treatment approaches is much needed.

In Publication I, we were able to show that heavy smoking and high BMI are accelerating factors for earlier emergence of nAMD, while the use of anticoagulation and anti‐aggregation medication was associated with a later onset of nAMD. When the relation of analysed baseline factors to treatment duration was analysed, patients with SRF or anticoagulation medication at baseline stayed longer on treatment than other analysed patient groups. Interestingly, the presence of IRF at baseline and low number of injections during the first treatment year resulted in a fast‐progressing disease. More intriguingly, we observed a link between IRF and the appearance of subretinal drusenoid deposits, HRF, and increased retinal atrophy. Thus, our results show that IRF plays an important role in the escalation of the disease towards non‐reversible changes in retinal anatomy.

Role of some of these factors e.g. smoking and obesity is already recognised as accelerating factors during AMD progression (Saigal et al., 2025). However, also novelty can be found in Publication I, which underlines research questions that should be carefully examined in the future. For example, Akter et al. have reported that the thrombin inhibitor dabigatran reduces the risk of new nAMD and in a mouse model accelerates CNV repair. They showed that thrombin reduces transepithelial resistance of RPE cells, generates unique complement C3 and C5 cleavage products, and led to VEGF secretion (Akter et al., 2022). As we also showed that anti‐coagulation medication may play a role in modifying AMD progression, we consider that the relationship between anti‐coagulation medication and AMD should be addressed more in detail in the future. Interestingly, high BMI and smoking had opposite effects on nAMD diagnosis age than anti‐coagulation and anti‐aggregation medication. This observation is intriguing as altered vascular health is associated with both smoking and high BMI (Faber et al., 2009). Additionally, obesity and smoking have been shown to be linked with hypercoagulability (Barua & Ambrose, 2013; Faber et al., 2009; Kornblith et al., 2015). Hypercoagulability associated with obesity may be due to chronic low‐grade inflammation, altered hepatic synthesis of coagulation factors, or adipose tissue derived coagulation mediators (Faber et al., 2009; Kornblith et al., 2015). Also, smoking has been shown to alter the balance of coagulation factors (Barua & Ambrose, 2013).

Moreover, our data from Publication I indicates that the presence of IRF already at baseline is linked with faster disease progression and worse disease outcome. Retinal fluid status can fluctuate during the nAMD treatment; thus, our results highlight that when analysing relationships of retinal fluids and disease outcomes, longitudinal analysis is essential to avoid misinterpretation of the results. As our data revealed, the previously undescribed link of IRF to HRF, SDDs and cRORA, future studies should explore whether there is a common mechanism driving all these phenomena. Pathological fluid accumulation occurs when the leakage exceeds the local capability to remove the fluid. Macular fluids can develop in various ways; they can, for example, result from the breakdown of the blood retinal barrier, osmotic pressure of accumulated proteins or via dysfunctional fluid transport of the RPE layer. Newly formed incomplete blood vessels can also leak fluid and immune cells to the surrounding tissue microenvironment (Haydinger et al., 2023; Helotera & Kaarniranta, 2022). Accumulating fluid can further damage the tissue microenvironment and may lead to increased inflammatory cytokine release, inflammatory cell infiltration or hypoxia, causing further progression of the disease (Haydinger et al., 2023; Helotera & Kaarniranta, 2022). Development of HRF is not completely understood, but theories of their origin include that HRF may represent anteriorly migrating RPE cells or that HRF may also include lipid‐filled immune cells in nAMD (Echols et al., 2020). Observed RPE migration may be driven by hypoxia (Echols et al., 2020). Hypotheses that have been created on SDD development include them stemming from choroidal vascular dysfunction, RPE dysfunction or that SDDs are para‐inflammatory phenomena (Monge et al., 2022). Furthermore, the presence of SDDs is linked with AMD severity and photoreceptor disruption, photoreceptor loss, localised Müller cell gliosis and RPE alterations (Monge et al., 2022). Based on our results, the presence of IRF at any point during patient treatment history may be an indicator of a more advanced stage of the disease.

Interestingly, all our findings in Publication I can be linked to the health of vasculature and thus our results underline the potential role of disturbed vasculature during the whole disease course of nAMD. Overall results from Publication I may help ophthalmologists assess an individual's risk for disease progression and may draw attention to lifestyle intervention with patients who are at risk for fast disease progression.

In Publication II, our goal was to understand how serum RNA profile differs between the control and nAMD patients. Our aim was also to understand if serum RNA profile reflects anatomical changes during the nAMD progression and whether it can be altered by anti‐VEGF medication. Unsatisfactory treatment outcomes in nAMD remain one of the greatest challenges in ophthalmology, imposing a significant social and economic burden (Kaarniranta et al., 2023). To recognise the factors behind the insufficient treatment responses of individual patients, a precise understanding of the processes and biomarkers associated with the disease is needed. Currently, developments in clinical AMD diagnostics are more focused on imaging, artificial intelligence‐based image analysis, role of imaging markers in prediction of treatment outcomes and self‐monitoring devices (Abd El‐Khalek et al., 2024; Mathis & Kodjikian, 2022; Sarkar & Dyawanapelly, 2021). Serum biomarkers are not used utilised in diagnosis of nAMD. Serum as a source for biomarkers is a desirable target since its acquisition requires a far less invasive procedure than samples obtained from the target tissue. More diagnostic biomarkers for nAMD are needed for both research use and to identify patients at high risk for rapid disease progression, and for more personalised treatment approaches. Emerging evidence points that the serum or blood proteins, lipids, and microRNAs are associable with macular degeneration (Blasiak et al., 2019; ElShelmani et al., 2020, 2021; Hyttinen et al., 2021; Lad et al., 2023; Rozing et al., 2020; Yen et al., 2024). Although circulating non‐coding RNAs (ncRNAs), such as miRNAs, are routinely detected in the blood, far less is known about the significance of circulating mRNAs (Campillo et al., 2024; Cheng et al., 2021; ElShelmani et al., 2020, 2021). We recently showed changes in the serum mRNA profile in nAMD patients and initial evidence that serum mRNA may reflect changes in retinal compartments was established (Liukkonen et al., 2024). In Publication II, we analysed serum RNA profiles more in detail and showed an association of altered signatures of circulating mRNA, ncRNA and lncRNA with nAMD progression at an anatomical level. Data search criteria differed between this study and our previously published analysis. In this study, we analysed mRNA, long ncRNA (lncRNA) and small ncRNA (sncRNA) profiles in the serum of nAMD patients and associated them with anatomical changes observed at the retinal level as well as with anti‐VEGF treatment status.

RNA‐seq analysis was used to identify serum RNA as it provides the opportunity to identify novel RNA and RNA variants compared to traditional microarray analysis, which analyzes the transcriptome using known RNA probes. Additionally, RNA‐seq offers a higher percentage of differentially expressed genes, especially genes with low expression (Wang et al., 2014). This is particularly why RNA‐seq was chosen for this analysis as serum normally contains only low amounts of RNA. Furthermore, unlike many publications involving RNA‐seq analysis, we utilised DNase treatment to eliminate the risk of DNA contamination, which could affect RNA‐seq readings (Li et al., 2022).

In publication II, we were able to demonstrate that serum RNA profile differs between neovascular AMD (nAMD) patients and controls but also reflects anatomical changes in the retinal fluid compartments as well as atrophic changes of the retina. Majority of the observed differences indicated downregulation of the molecules in nAMD. Based on general expression patterns, the analysis showed significant downregulation of gene expression and translation‐related pathways. Enriching pathways included, for example translation initiation, RNA polymerase II and III transcription and microRNA biogenesis, thereby explaining the trend for general downregulation of the molecules. Also, pathways important in the regulation of the vasculature, such as RHO GTPase cycle, mTOR pathway, ROBO receptors and protein kinase A, were observed to be altered. Overall, the molecular interactions enriched for increased necrosis and apoptosis and decreased cell viability, movement, and cellular organisation, reflecting degenerative changes observed during AMD. Interestingly, when the RNA profile was evaluated for anatomical changes, many of the altered RNAs had known functions in the regulation of the cell cycle, senescence, vasculature and retinal development of the retina. Thus, these changes were in line with the general trend from the comparison of the control and nAMD samples.

In our analysis, we followed two independent ways to control false‐positive leads, and when these approaches were combined, TMX4 was observed to be differentially expressed by both approaches. Interestingly, the biological functions of TMX4 have not been well established and its role during the development of AMD is still unclear. TMX4 has been previously linked to oxidative damage in cultured RPE cells and has been shown to interact with calnexin, a protein whose impairment leads to retinal degeneration (Cano et al., 2014; Rosenbaum et al., 2006; Sugiura et al., 2010). Interestingly, Zhao et al. recently reported the TMX4 expression in vascular endothelial cells and platelets. Inhibition or lack of TMX4 in a mouse model resulted in decreased platelet aggregation and thrombosis (Zhao et al., 2025). Thus, the limited amount of what is known from TMX4 highlights that it functions during AMD development should be examined in the future.

Serum as a source of RNA contains only low amounts of RNA and due to the circulating nature of fluid it is exposed to a variety of factors that may influence the serum transcriptome. As expected, high variance between the samples was observed during the analysis in Publication II. These results highlight that it is important to identify confounding factors that can explain at least a part of the variance seen in the individual RNA expression levels between samples. We observed that gender, BMI and smoking, as well as the use of blood pressure, anti‐cholesterol, anti‐coagulation or anti‐aggregation medications, could be confounding factors that should be considered in the analysis of serum RNA data from nAMD patients. However, it should be noted that in our analysis each identified RNA had its unique set of confounding factors.

Although circulating non‐coding RNAs (ncRNAs), such as miRNAs, have been analysed more in detail and it has been demonstrated that serum miRNA profiles differ between AMD and control patients, far less is known about the significance of circulating mRNAs (Blasiak et al., 2019; ElShelmani et al., 2020, 2021; Yen et al., 2024). As we could observe some level of correlation in serum mRNA profile and progression of AMD, our finding opens a number of novel questions. For example, the origin of altered RNAs is still unclear. We were unable to answer the question about the origin of identified RNAs in publication II, but this study provided clues that our analysis enriches mRNAs with protein products observed in the posterior segment of the eye. To evaluate if protein products of identified mRNAs can be detected from the human eye, TUBGCP3 and CHMP6 were stained from cadaver samples. Intriguingly, both were identified from multiple compartments of the eye including photoreceptor outer segments, RPE and vasculature of the choroid. By our understanding, this is the first report to show staining of these molecules in the human eye.

Overall, serum as a source for research‐based biomarkers is a desirable target since its acquisition requires a far less invasive procedure than samples obtained from the target tissue. Thus, our results from Publication II also encourage finding alternative approaches for studying AMD, which are limited due to restricted access to live human target material and the limited value of animal models of human AMD.

In publication III, our goal was to demonstrate how long‐term use of systemic medications, including diabetes and antithrombotic medications, as well as corticosteroids and immunosuppressants, affects the development of AMD. Increasing evidence suggests that both local changes in the retina microenvironment and systemic components are involved in the development of AMD. There exists a substantial amount of evidence that low‐level inflammation is involved in AMD pathogenesis (Helotera & Kaarniranta, 2022; Khan et al., 2022; Nahavandipour et al., 2020). The role of non‐steroidal anti‐inflammatory drugs has been more under inspection; thus, we decided to focus on systemic corticosteroids and immunosuppressants (Xu et al., 2021; Yang et al., 2024). Instead, metformin targets multiple pathways of aging; thus, the use of metformin as a treatment for age‐related diseases such as AMD is under inspection (Barzilai et al., 2016). Intriguingly, Aggarwal et al. reported recently that the use of metformin in non‐diabetic patients can reduce the odds of AMD development (Aggarwal et al., 2024). Instead, inconsistent results from the protective association have been received from patients with diabetes (Gokhale et al., 2023; Lee et al., 2019; Romdhoniyyah et al., 2021). In addition to immunosuppressants, corticosteroids, and metformin, we decided to include antiaggregation medications in our analysis, as it has been shown previously that the anticoagulant dabigatran may reduce the risk for new nAMD (Akter et al., 2022). In line with Akter et al.'s findings, we recently showed in Publication I from the smaller patient population that patients on antiaggregation (ASA or clopidogrel) and anticoagulation (warfarin or direct oral anticoagulants, DOACs) medication receive nAMD diagnosis later than their comparators. The difference was especially noticeable among anticoagulation medication users, and patients with anticoagulation medication also stayed longer on anti‐VEGF treatment than patients without anti‐coagulation medication (Helotera, Siintamo, et al., 2024). Thus, recent observations with metformin, anti‐coagulants and factors related to inflammation have highlighted the role of the systemically administered medications for AMD development, and the underlying diseases behind the medications should be examined further and with additional patient populations (Aggarwal et al., 2024; Akter et al., 2022; Helotera, Siintamo, et al., 2024; Jiang et al., 2022; Xu et al., 2021; Yan et al., 2022; Yang et al., 2024). To get a clear understanding, prospective randomised trials with large cohorts would be needed. However, findings from retrospective analyses can provide clues from the most relevant associations.

With Publication III, we were able to demonstrate that people with long‐term corticosteroid and immunosuppressant use are at greater risk for developing AMD. Initial results with antithrombotic medications suggested that people with antithrombotic medications could also have increased odds for AMD, but when 5‐year incidence for AMD was analysed in age‐matched groups results for antithrombotic medication were not as consistent as with corticosteroids and immunosuppressants. Interestingly, the mean age at diagnosis of AMD was later among antithrombotic medication users compared to the control group and earlier among patients using other diabetes medications than metformin. Based on results from Publication I, we were hoping to clarify the role of anticoagulants (warfarin and DOACs) during AMD progression with Publication III. However, due to the limited number of anticoagulant users, especially DOAC users, we could not separate these medications from the antiaggregation medications. Based on our results, the role of the systemic components behind corticosteroid and immunosuppressant during AMD progression should be examined in more detail in the future.

Overall, results from Publications I‐III and increasing evidence outside of these publications highlight that both local changes in the retina microenvironment and systemic components are involved in the development of AMD. Interestingly, our findings also suggest that altered vascular health and low‐level inflammation may modulate the disease course of AMD.

6.2. Limitations of the study

Greatest limitations in Publication I rise from the nature of the study which involves real world patients. Thus, data is more likely descriptive and will raise research questions than give concrete finalised answers. For example, patients in different arms on comparisons may have different background factors that may affect outcomes. Moreover, data is not collected for study, and real‐world recording practices may have caused limitations for data availability. One limitation of the study was that available information from baseline medications was limited; for example, we had no data from the durations of baseline medications before the AMD diagnosis. Moreover, with this study we cannot separate the role of diseases behind medications from the role of medication.

In publication II, we had samples from 60 nAMD and 64 control patients. One limitation of the study is the lack of duplicate from an independent patient population. We also had different patients in different comparisons, for example nAMD patients who had not received anti‐VEGF were different than patients who were on treatment. Thus, we did not have the opportunity to see if changes will occur also within one patient if individuals are observed for a longer period. For example, can we see the same changes when an individual starts anti‐VEGF, when SRF changes to IRF, or when atrophy increases? We also observed a number of confounding factors and showed that patient characteristics can modify serum RNA profiles. In an optimal setting, patient backgrounds would be balanced between comparator arms, and a longitudinal analysis for serum RNA would be performed. As variation in RNA levels between samples was high, a more standardised approach could also facilitate controlling that issue.

In publication III, one limiting factor is that the cohort was still relatively young for an AMD study, and it is expected that more AMD cases will arise in the future. Thus, we consider that analyses for 5‐year incidences, comparison of different anti‐thrombotic medications, as well as diabetes medications should be revisited once the study population ages. One significant limitation in publication III was also that medications were only withdrawn from the KELA registry for reimbursed medication purchases. Thus, medications administered at hospitals or purchased without any reimbursement are not included in the analysis. Therefore, future analysis would benefit from including also medication provided in hospitals. In addition, with this study we could not differentiate whether observed differences occur due to medications used or underlying baseline disease. Thus, based on our results, the role of systemic changes behind the most common baseline diseases of systemic corticosteroid and immunosuppressant use, and AMD development should be examined in more detail in the future.

6.3. Future directions

In Europe, there are currently approved medications only for the treatment of nAMD and patients with dAMD do not have treatment options available. Vision is one of the factors that strongly affects the quality of life and functionality. Thus, lack of effective treatments in dAMD is an obvious unmet medical need that should be addressed in the future. However, regardless of available efficacious treatments for nAMD, the number of open questions remains also in nAMD research. Current treatment options are not curative, and they are able to only slow down disease progression. To find curative treatment opportunities, early stages of disease development should be understood more in detail. Currently, there is a vast amount of information available, but the information is fragmented, partly contradictory, and a clear overall picture is lacking. Raising the question of whether the key to binding all findings together is still waiting to be found.

As the number of patients with AMD is high, causing an economic burden for society and resource demand in hospitals, it would be beneficial also to be able to identify people with high risk for AMD and ways to interfere with the development. Early intervention, to delay in AMD emergence and evident vision loss would be also desirable for the patient, whose ability to function can be dramatically affected due to visual impairment. To identify high risk patients and propose preventive actions, more work should be done to understand factors behind AMD development. Even now, there is a lot of information available, but we need a way to consolidate all this information into a unified algorithm. Perhaps then we could better understand where the gaps still are, and which topics need more detailed examination. Currently, patients are identified when the disease is already in advanced stage and when damage has already developed. Moreover, nowadays clinicians cannot provide individualised prognoses of disease progression and possibilities for individualised treatment paths are limited. In the future, personalised treatment paths could allow cost savings in big picture when resources are directed based on patients' needs but also considering cost benefit ratio. There are a number of factors that could be considered when providing personalised treatment paths. To provide a couple of examples, things that could be considered in the future during treatment selection could include: 1. Patient selection based on molecular profiling. Can early intervention with correct medication provide better prognosis for the patient and delay costs for society from visual impairment? 2. Detection of patients with high risk for treatment provoked serious side effects. For example, in case of vision threatening side effects. Could we avoid cost from the side effect management for the ophthalmology department and individuals' loss of quality of life due to side‐effect triggered visual impairment?

AMD is a complex disease, with alterations in multiple signalling pathways. To create plans for early intervention and discovery of curative treatments, fragmented information should be merged to find the most important effectors. The information generated in this thesis will hopefully add pieces to this puzzle and benefit patients with AMD in the future.

7. CONCLUSIONS

1. Heavy smoking and high BMI are accelerating factors for earlier emergence of nAMD, while the presence of IRF results in a fast‐progressing disease. Moreover, there is a link between IRF and the appearance of subretinal drusenoid deposits, HRF, and increased retinal atrophy. Thus, IRF plays a role in the escalation of the disease towards non‐reversible changes in retinal anatomy.

2. RNA profile of nAMD patients corresponds with anatomical changes in the retinal fluid compartments as well as atrophic changes of the retina. Especially, the role of thioredoxin‐related transmembrane protein 4 (TMX4) during AMD development should be evaluated in the future.

3. Our results highlight the role of systemic components during AMD progression and show that people with long‐term corticosteroid and immunosuppressant use are at greater risk for developing AMD. Further research on systemic components behind corticosteroid and immunosuppressant use and modulating AMD progression is warranted

ACADEMIC ADVISORS

Series Editors:

Professor Jari Halonen, M.D., Ph.D., M.A. (education), Institute of Clinical Medicine, Surgery, Faculty of Health Sciences, University of Eastern Finland, Kuopio, Finland.

Professor Ville Männistö, M.D., Ph.D., Institute of Clinical Medicine, Internal Medicine, Faculty of Health Sciences, Kuopio, Finland.

Research Director Toni Rikkonen, Ph.D., Institute of Clinical Medicine, Faculty of Health Sciences, Kuopio, Finland.

Professor Tarja Malm, Ph.D., A.I. Virtanen Institute for Molecular Sciences, Faculty of Health Sciences, Kuopio, Finland.

Lecturer Veli‐Pekka Ranta, Ph.D., School of Pharmacy, Faculty of Health Sciences, Kuopio, Finland.

University Lecturer, Docent Päivi Kankkunen, Department of Nursing Science, Faculty of Health Sciences, Kuopio, Finland.

Supervisors:

Professor Kai Kaarniranta, M.D., Ph.D., Ophthalmology, Institute of Clinical Medicine, School of Medicine, Faculty of Health Sciences, University of Eastern Finland, Kuopio, Finland.

Niko Kivinen, M.D., Ph.D., Ophthalmology, Kuopio University Hospital, Kuopio, Finland.

Reviewers:

Professor Renāte Ranka, Ph.D., Faculty of Pharmacy, Riga Stradins University, Riga, Latvia.

Professor Maciej Gawecki, M.D., Ph.D., Dobry Wzrok Ophthalmological Center, Gdansk, Poland.

Opponent:

Professor Panu Jaakkola, M.D., Ph.D., Department of Oncology and Radiotherapy, FICAN West Cancer Centre, University of Turku, Turku University Hospital, Turku, Finland.

LIST OF ORIGINAL PUBLICATIONS

This dissertation is based on the following original publications:

1. Heloterä H, Siintamo L, Kivinen N, Abrahamsson N, Aaltonen V, Kaarniranta K. Analysis of prognostic and predictive factors in neovascular age‐related macular degeneration Kuopio cohort. Acta Ophthalmol. 2024, Sep;102(6):703–713.

2. Heloterä H, Kostanek J, Liukkonen M, Siintamo L, Linna‐Kuosmanen S, Watala C, Blasiak J, Kaarniranta K. Serum RNA Profile Reflects Fluid Status and Atrophic Retinal Changes in Neovascular Age‐Related Macular Degeneration. Int J Mol Sci. 2025, May 19;26(10):4852.

3. Heloterä H, Sund R, Rikkonen T, Kaarniranta K. The impact of systemic long‐term medications for the development of age‐related macular degeneration. Acta Ophthalmol. 2026, Mar 6; 00, 1–8.

The publications were adapted with the permission of the copyright owners.

Heloterä, H. (2026) Insights into the modulation of age‐related macular degeneration: Molecular and systems‐level perspectives. Acta Ophthalmologica, 104(Suppl. 293), S4–S35. Available from: 10.1111/aos.70239