Introduction

Neovascular age-related macular degeneration (nAMD) can be treated with intravitreal injection of drugs that inhibit the effects of vascular endothelial growth factor (VEGF) on choroidal neovascularization [1]. The therapeutic effects of VEGF antagonists have primarily been studied in the first eye developing the disease which prevents bias due to interdependence of observations from the two eyes [2,3,4,5]. However, knowledge of the pattern of development of nAMD in the fellow eye is relevant since symptoms on the first eye may remain unnoticed and therefore only are appreciated when the fellow eye is affected. Additionally, the planning of follow-up examinations to detect disease in the fellow eye depends on knowledge of risk factors and the time lag between disease development in the two eyes.

Previous studies of bilateral nAMD have focused on the point prevalence [6], the influence of baseline visual acuity [7], morphological symmetry of lesions in the two eyes [8], and the effect of delayed onset of treatment in the fellow eye [9]. However, these studies were based on patient materials selected for randomized clinical studies, clinical risk factors other than visual acuity were not considered, and changes in incidence of fellow-eye involvement over time were not studied.

Therefore, the present study reports epidemiological and clinical risk factors for the development of neovascular AMD in the fellow eye among 2516 consecutive patients from a defined population already diagnosed with the disease in one eye during a period of more than 10 years.

Methods

Design

Prospective observational cohort study of the development of nAMD in the fellow eye from the only treatment center for a population of 0.9 million citizens of which approximately 38% is above the age of 50 years. All clinical data were continuously entered into a structured patient record from which the data were transferred to a database for surveillance of activity and quality of treatment. The study was approved by The Central Denmark Region Committees for Health Research Ethics.

Population

The population has been described in detail previously [10] and included all patients presenting with visual symptoms suspected to represent nAMD in the Eastern part of the Central Denmark Region. The patients had free access to a practitioning ophthalmologist within a few days of onset of visual symptoms. After a primary examination including inspection of the fundus and optical coherence tomography (OCT) scanning, patients were referred for treatment at the Department of Ophthalmology, Aarhus University Hospital, which is the only site administering anti-VEGF treatment to patients in the area. The basis of the present study is all 2516 patients in whom intravitreal treatment with at least three injections with anti-VEGF compound for nAMD was initiated in one or both eyes during a period of 10 years and 1 month until April 1, 2017.

Examination

At referral, the patients were interviewed about the nature and duration of visual symptoms, followed by measurement of visual acuity (VA) using Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity charts (Polaphor, Block Optic LTD, Dortmund, Germany). Mydriasis was induced by tropicamide 1% (Alcon, Copenhagen, Denmark) and phenylephrine 10% (Region Midtjylland Hospital Pharmacy, Aarhus, Denmark) eye drops, followed by optical coherences tomography (OCT) scanning using Topcon equipment (DRI OCT or 3D OCT 2000, TRITON, Topcon, Tokyo, Japan), to detect retinal pathology suggesting subretinal neovascularization. The average thickness of the circular retinal area with a diameter of approximately 1 mm centered on the point of fixation [11] was noted as central retinal thickness (CRT). When a choroidal neovascularization (CNV) was suspected, fluorescein angiography (FA) was performed with the filling phase on the most severely affected eye, using a Canon CF-1 angiography system (Canon, Amstelveen, Netherlands) from 2007 and a Topcon camera (DRI OCT Triton, Topcon, Tokyo, Japan) from 2016. Overt or suspected CNVs were assessed as being subfoveal, parafoveal, peripapillary, or extrafoveal, and the type of lesion was defined to be either classic, predominantly classic, minimally classic, or occult. The largest diameter of the area that showed increasing hyperfluorescence through the angiographic phases was assessed with the horizontal diameter of the optic disk set to a reference of 1000 μm.

Treatment was planned as three monthly intravitreal injections with anti-VEGF compound followed by monthly controls with measurement of visual acuity and OCT scanning, and treatment pro re nata (PRN) until the central retinal edema had disappeared or remained stable for 6 months.

Patient selection

The exclusion criteria were the following: residence outside the defined geographical area, one or more treatments performed in other regions (n = 21), age below 50 years (n = 6), previous treatment with retinal photocoagulation (n = 15) or photodynamic therapy (n = 20), cases where a CNV might be due to other causes than nAMD (n = 117), and patients in whom intercurrent events had resulted in the administration of less than three intravitreal injections of anti-VEGF compound (n = 14).

The inclusion followed Danish national guidelines which implied the presence of choroidal neovascularization documented by FA, retinal edema, a history of visual impairment during the preceding 6 months, and visual acuity better than 20 ETDRS letters.

This resulted in the initiation of treatment in 3057 eyes from 2516 patients who received at least three anti-VEGF injections for nAMD in the department during the period. The national death registry automatically updated the electronic patient record from which it appeared that 616 of the patients had died after (mean ± SD, range) 4.57 ± 2.13 [0.24; 9.56] years, whereas the remaining 1900 patients had been followed (mean ± SD, range) 4.17 ± 2.65 [0.16; 10.04] years. When retinal thickness had been stable for at least 6 months, follow-up was continued with the private practitioning ophthalmologist. The patient was informed to check vision regularly using Amsler grid and had free access to the ophthalmologist within a few days if vision declined. The ophthalmologists referred the patient to the eye department for re-assessment if central retinal thickness increased or a need for further treatment in either eye was suspected. Therefore, the material can be considered to contain all patients qualifying for treatment in either eye during the follow-up period.

Treatment was performed using ranibizumab 10 mg/ml (Lucentis®, Novartis, Basel, Switzerland) from the onset of treatment in the clinic, with the option to change to aflibercept 4 mg/ml (Eylea®, Bayer, Leverkusen, Germany) in non-responders from 2013. Aflibercept was the first-choice treatment in all cases since 2015, with the possibility to change to ranibizumab in non-responders. Treatment had been switched in 547 eyes from 509 patients.

All clinical data were continuously entered into a structured patient record from which the data were transferred to a database for surveillance of activity and quality of treatment.

Data analysis

The anatomical characteristics observed on fluorescein angiograms were converted to factor variables as follows: membrane position—subfoveal = 1, parafoveal = 2, peripapillary = 3, extrafoveal = 4. Membrane type—classic = 1, predominantly classic = 2, minimally classic = 3, occult = 4. The number of injections, VA, and CRT at the first examination after 12 months was calculated for 409 (75%) of the patients who were still followed at this time.

Statistical analysis

All analyses were performed in STATA (version 14.0, StataCorp, Texas, USA).

Unpaired t tests (continuous variables) or chi-square tests (contingency values) were used to test for differences at baseline between patients who had bilateral treatment requiring nAMD at the primary referral and those in whom the diagnosis on the fellow eye was established later, i.e., more than 1 month after the first eye.

Paired t tests (continuous variables) or chi-square tests (categorical variables) were used to test for differences between the right and the left eye in patients who had bilateral disease at referral and between the first and the fellow eye in patients in whom only one eye was affected at referral.

The risk for developing nAMD in the fellow eye was assessed using survival analysis with death as competing risk. The influence of individual risk factors was assessed at 6 months and at each year between 1 and 8 years after the disease was diagnosed in the first eye using multiple regression based on the pseudo-values method. The continuous variables were modeled in a restricted cubic spline with four nodes in the fractiles 5%, 35%, 65%, and 95% and were transformed to the best linear fit before entering into the model. According to this, visual acuity could be entered without transformation, leakage diameter after log transformation, and age after a quadratic transformation. CRT could not be transformed to a linear fit with simple functions and was therefore included in the analyses using splines.

Results

The patient selection is shown in Table 1. It appears that among the 2516 nAMD patients in whom anti-VEGF treatment had been initiated, 541 (21.5%) had developed treatment requiring nAMD on the fellow eye during the observation period. In 132 (24.4%) of these patients, the condition was bilateral at the time of diagnosis, whereas in the remaining patients nAMD had been diagnosed in the fellow eye at least 1 month later than in the first eye. The patients who had bilateral nAMD at the time of diagnosis were significantly older than the patients who initially presented with unilateral disease.

Table 1 The numbers of patients and the distribution of background parameters in patients diagnosed with bilateral disease and in whom the fellow eye was diagnosed later than the first eye. Age is indicated as mean ± SEM (number) with the median and range below
Full size table

Table 2 shows the annual number of patients diagnosed with nAMD in the first eye, who had died and who had been diagnosed in the second eye, and patients retained at risk for second eye involvement year-wise back from March 31, 2017.

Table 2 The patients retained at risk calculated as this number from the preceding year corrected for newly diagnosed patients and patients who had died and who had been diagnosed in the second eye year-wise back from March 31, 2017
Full size table

Table 3 shows that in the patients presenting with bilateral disease, there were no significant differences between the studied parameters among the two eyes. However, Table 4 shows that in the patients who initially presented with unilateral disease, the later presentation in the fellow eye was accompanied with a significantly higher VA, a lower CRT, a membrane position more distant from the fovea, and a smaller leakage area, but showed no significantly different membrane type. Additionally, after 12 months, the fellow eye had a significantly lower effect in VA and CRT than in the first eye. In spite of this, VA was still significantly higher in the fellow than in the first eye after 12 months (mean ± SD) 64.4 ± 16.6 ETDRS letters versus 58.2 ± 20.8 ETDRS letters (p < 0.0001), whereas there was no significant difference in CRT among the two eyes after that time (mean ± SD) 254 ± 79.4 μm versus 257 ± 87.1 μm (p = 0.61).

Table 3 Visual function and morphological characteristics of the two eyes of patients diagnosed with bilateral nAMD, at baseline and after 12 months. Data are presented as numbers or mean ± SEM (number) and for visual acuity, age and time at follow-up also with the median and range (below)
Full size table
Table 4 Visual function and morphological characteristics of the two eyes of patients diagnosed with nAMD on the second eye at least 1 month after diagnosis on the first eye, at baseline and after 12 months. Data are presented as numbers or mean ± SEM (number) and for visual acuity age and time at follow-up also with the median and range (below)
Full size table

Figure 1 shows the cumulative incidence of developing nAMD in the fellow eye after correcting for the competing risk due to death and after censoring for differences in follow-up time. The incidence curve starts at 5.2% corresponding to the 132/2516 of the patients with bilateral disease at the time of diagnosis of nAMD. Table 5 shows the numbers of patients underlying the curve.

Fig. 1
figure 1

The cumulative incidence of developing nAMD on the fellow eye. The dashed lines delimit the 95% CI intervals

Full size image
Table 5 The number of patients at risk, who had died, and who had developed fellow-eye involvement after each of the first 8 years after diagnosis of nAMD on the first eye
Full size table

Table 6 shows the contribution of the studied risk factors to fellow-eye involvement 12 months after the diagnosis of nAMD in the first eye based on 2333 observations where all parameters were present. Negative coefficients indicate that parameters increase the time to occurrence of nAMD in the fellow eye. It appears that male gender, increasing leakage area, and peripapillary location of the subretinal neovascular membrane in the first eye reduced the risk of developing disease in the fellow eye. The quadratic fit of age implied that the risk was bimodal with an increasing risk with age until a maximum, followed by a decreasing risk with age. This age of maximum risk decreased from 98.8 years for fellow-eye involvement within 6 months to 88.3 years for fellow-eye involvement within 1 year to reach a minimum of 78.9 years for fellow-eye involvement within 4 years and longer. However, smoking at the diagnosis of nAMD, eye laterality, membrane type, and VA had no influence on the risk of developing disease on the fellow eye. Increasing age was a significant risk factor after 6 months, and higher VA was a significant risk factor after 3 and 4 years. Otherwise, the risk factors were similar when the calculation was repeated with the follow-up time set to 0.5, 1 (shown in Table 4), 2, 3, 4, 5, 6, 7, and 8 years.

Table 6 The coefficients, 95% confidence intervals, and p values of the studied risk factors for developing nAMD on the fellow eye within 12 months
Full size table

Discussion

The overall finding of bilateral disease in slightly more than one out of five patients with nAMD is in accordance with a previous study of the prevalence of bilateral disease 2 years after the diagnosis of the disease in the first eye [6] and is lower than reports from more selected patient materials [12, 13]. However, the overall prevalence is only a crude measure of bilateral involvement, and therefore the objective of the present study was to extend previous studies by correcting data for censoring due to death and varying follow-up time, by differentiating between patients who present with bilateral disease and those in whom the fellow eye is affected later than the first eye, and by considering possible risk factors for fellow-eye involvement.

The baseline parameters in both eyes of patients referred with bilateral disease were similar to previous findings in the first eye [10]. However, the visual outcome after 12 months was lower than previous findings of an average 2.2 ETDRS letters improvement in from our clinic [10], which may indicate that these patients had reached a more severe disease condition by the bilaterality of the disease at the time of referral. However, it cannot be excluded that the results at follow-up are biased since this was based on the 75% of the patients who were still followed in the eye department after 12 months, who may have been the most severely affected patients. The patients referred with bilateral disease are represented by the start position around 5% of the cumulative incidence curve. The lack of difference in clinical and morphological parameters among the two eyes in these patients at the time of diagnosis and after 12 months indicates that the disease had developed in parallel and argues for a strong coupling between disease development in the two eyes in these patients [8].

The finding that patients who developed nAMD in the fellow eye later than the first eye in our real-life setting had a higher VA at the time of diagnosis is opposite to results from a study on patient materials selected for inclusion in randomized clinical trials [14]. In the present study, fellow eyes also had lower CRT, smaller leakage area, and new vessels located more distantly from the fovea than in the first eye, and had a significantly lower effect of anti-VEGF treatment on VA and OCT in the fellow eye than in the first eye after 12 months, which may reflect a lower potential for improvement in the eyes with less severe disease [7]. In spite of this, VA was still significantly better in the fellow than in the first eye after 12 months, which may be due to the fact that experience with nAMD on the first eye had incited the patients to be more aware of symptoms of fellow-eye involvement [15]. A positive effect of anti-VEGF treatment on visual function in fellow eyes cannot be reached by treating fellow eyes in patients diagnosed with nAMD in one eye [9], and therefore attention should be directed at detecting incipient symptoms of fellow-eye involvement. The value of the patient’s experience with the first eye might be interpreted as the potential gain from a monthly screening of an informed general population rather than the outcome when patients unexperienced with the disease notice and react to subjective symptoms of nAMD for the first time. This might be included in analyses of cost-effectiveness for public health programs aimed at managing nAMD.

The slope of the survival curve can be seen to gradually decrease representing a decreasing point incidence with time. It appears that approximately 6 years after the diagnosis of nAMD on the first eye, the slope has declined to less than 1% and after nine years the cumulative incidence is around 28%. It is possible that some of these patients had developed non-exudative neovascularizations earlier [16], which may have affected the time of overt fellow-eye involvement. In addition to this, the finding suggests that some patients are prone to second-eye involvement and others are not, which supports evidence of a genetic disposition for fellow-eye involvement [17, 18]. The fact that fellow-eye involvement seems to be negligible later than 6 years after the diagnosis of the disease in the first eye might be used in the planning of follow-up programs for the disease.

The similarities in the risk factors for fellow-eye involvement at different follow-up times argue for the validity of the survival analysis model. Thus, all analyses showed a reduced risk for fellow-eye involvement in men and no influence of smoking, membrane type, and eye laterality, although the number of peripapillar new vessels was too low for a conclusion about this membrane type to be made. This corresponds to previous findings of an increased risk of reaching a treatment-requiring condition in the first eye in women [19, 20], whereas the increased risk of reaching the disease in the first eye in smokers could not be confirmed for the fellow eye [21]. This may reflect a different risk profile for disease development later in the disease and corresponds to the lack of influence of smoking on the therapeutic response of anti-VEGF treatment [10]. The previously observed increasing risk with age for developing nAMD in the first eye [22] could be confirmed for the fellow eye, but with the modification that this risk reached a maximum which was approximately 79 years for patients with a risk of fellow-eye involvement after 4 years. The background for the reduced risk of fellow-eye involvement after a specific age requires further investigation. The reduced risk of developing nAMD in the fellow eye with increasing leakage area might be related to a better preserved choroidal vascularization and supports that choroidal ischemia may participate in the pathogenesis of nAMD [23]. This is also supported by findings that in first eyes with nAMD, drusen volume is larger [24], the choroid is thinner [25], and choroidal blood flow is lower [26] than in fellow eyes without nAMD. The protective effect of peripapillary localization of the CNV in the first eye might indicate that this location represents a particular type of CNVs with a higher probability of unilateral involvement. This might be used to weigh the risk for fellow-eye involvement. The positive effect of high visual acuity on the first eye for developing CNV on the fellow eye after 3 and 4 years, but not after shorter and longer time periods, remains obscure and probably represents more complex risk relations that cannot be determined from the present material. This complexity is supported by the fact that CRT could not be transformed to a linear variable that could be included in the analysis and confirms the lack of direct correlation between VA and CRT [27, 28].

Altogether, the findings suggest that the planning of follow-up examinations to detect fellow-eye involvement of patients should consider the time and morphology of nAMD diagnosed in the first eye. Fellow-eye involvement occurs with a higher probability within the first 6 years after the first eye, in women and in cases where the leakage area was small and the location of the new vessel is not peripapillary in the first eye.