Age-Related Changes in Retinal Functional Topography

Abstract

purpose. To perform a detailed topographical analysis of functional age-related changes over the retina.

methods. Fifty-nine normal phakic subjects aged 10 to 69 years were divided into six groups, according to decade of age. mfERG traces were recorded from the central 60° of the retina, with a resolution of 61 and 103 scaled hexagons. Group medians of peak amplitude and latency of the first- and second-order (first slice) responses were used to generate 3-D topographical maps.

results. With age, there was a continuous loss of amplitude and delay of implicit time of the first- and the second-order response components, but the topography of the loss was not uniform across the retina. Trend analyses on ring group data showed a significant decrease in amplitude of first- and second-order responses although the age relationship of second-order responses was more complex. The loss of first-order kernel amplitude was generally accompanied by a rise in implicit time. Second-order kernel latencies showed no uniform alteration with age.

conclusions. Consistent with previous work, a steady loss of amplitude and increase of implicit time was observed with age. The topographical 3-D data, however, reveal age-related functional alterations in the retina beyond those found in ring averages, suggesting that these are masked by the standard analysis. Thus, the choice of physiologically coherent regions of interest may increase the sensitivity of detecting age-related change in multifocal analysis of retinal function.

Retinal function can be objectively measured by recording the electrical potentials generated at the cornea in response to controlled stimulation. Age-related alterations in retinal function have been examined in several studies ^{1 2 3 4} with the Ganzfeld-electroretinogram (ERG), believed to originate in outer retinal layers. The results show a significant decrease in the photopic b-wave with age, accompanied by an increase in its implicit time. A progressive decline of both macular and paramacular responses with age beyond 20 years has been described ⁵ ; however, it has also been reported that only the amplitudes of foveal responses decrease significantly with age. ⁶  

The multifocal electroretinographic (mfERG) technique of Sutter and Tran ⁷ allows a more detailed topographical evaluation of retinal activity than that used in the studies just cited. Again, the main effect of aging is a continual reduction of amplitude with, in some cases, delayed implicit times for the first-order kernel components or linear response. ^{8 9 10 11 12 13} Only a few studies have been conducted to evaluate the effect of age on higher-order mfERG response kernels or nonlinear responses, ^{10 12 14 15} thought to reflect adaptational mechanisms. ^{16 17} As for the first-order kernel, results show decreases in amplitude, but implicit times appear less affected. Most of these studies have found that both central and peripheral mfERG responses decrease with increasing age, but that the decrease is more prominent in the central retina. 

Although the multifocal technique allows a detailed topographical analysis of retinal sensitivity, most of these studies did not take advantage of that resolution and analyzed responses averaged for five or six rings concentric on the fovea. ¹⁸ Moreover, as the retinal topography of the response is not uniform, local differences such as those between the superior and inferior retina ¹⁹ may be masked, particularly since the areas that are averaged increase in the periphery. In a numerical analysis of local mfERG responses of the first-order kernel, a linear decline in amplitude with age has been found. ¹⁹  

In this study, we examined in detail the retinal topography of mfERG amplitude and latency alterations with age in healthy phakic subjects with no clinically apparent ocular abnormalities. Each individual mfERG response element was analyzed, and profiles of the retinal response were then constructed for both the amplitude and latency of the P1 component for first- and second-order kernel analyses. We also compared the results of the first-order kernel obtained with a ring analysis to those found in a topographic analysis, to examine whether a ring analysis is adequate for monitoring age-related alterations. 

Methods

Subjects

Fifty-nine normal subjects (29 female, 30 male) aged 10 to 69 years participated in the study. They were divided into six age groups according to decade: 10 to 19 years (9 subjects; median 15 years), 20 to 29 years (11 subjects; median, 26), 30 to 39 years (11 subjects; median, 34), 40 to 49 years (10 subjects; median, 44), 50 to 59 years (10 subjects; median, 55), and 60 to 69 years (8 subjects; median, 64). All fulfilled the following inclusion criteria: Far and near best corrected visual acuity of 20/20 or 20/15, refractive error < ±3.0 spherical diopters and < ±1.0 cylindrical diopters, intraocular pressure ≤20 mm Hg, disparity of stereopsis ≥600 minutes, and clear ocular media ascertained using slit lamp examination in artificial pupil dilatation by the same certified ophthalmologist. The fundus was completely normal, blood pressure was ≤140/90 mm Hg, and there was no systemic disease and/or relevant medication. 

Refractive errors were corrected, and the pupils fully dilated (≥7 mm) with 0.5% tropicamide. ERG responses were recorded from the cornea with DTL fiber electrodes (UniMed Electrode Supplies, Farnham, UK). The reference and ground skin electrodes (gold cup electrodes) were attached to the ipsilateral temples and forehead, respectively. The research adhered to the tenets of the Declaration of Helsinki. 

Functional Testing

The stimulus was generated on a 20-inch monitor (Sony, Japan) with a frame rate of 75 Hz positioned 28 cm from the subject’s eye. Two mfERG recordings were performed after the guidelines of the International Society for Clinical Electrophysiology of Vision (ISCEV): one with a stimulus of 61 hexagons and one with a stimulus of 103 hexagons. For the 103 hexagons stimulus, ring 1 ranged from ∼0–1.5°; ring 2, 1.4–6°; ring 3, 4–11°; ring 4, 8–18°; ring 5, 13–26°; and ring 6, 19–35°. For the 61-hexagon stimulus, average eccentricities were ring 1, 0–2°; ring 2, 1.8–7°; ring 3, 5.0–13°; ring 4, 11–22°; and ring 5, 17–30°. 

The stimulus hexagons were modulated between white (100 cd/m²) and black (<1.0 cd/m²) according a standard m-sequence of Sutter and Tran ⁷ with a length of 2¹³ – 1. The luminance of the surrounding screen area was set to 51 cd/m², and a red central cross was presented for fixation. A new stimulus picture was presented every 13.33 ms. Recordings at 16 samples per frame yielded a temporal resolution of 0.833 ms. The recordings were amplified (×200,000) and band-pass filtered (10–100 Hz; model 12 amplifier; Grass Telefactor, Quincy, MA). The stimulation was divided into 16 segments, and those contaminated by artifacts were discarded and repeated. 

Data Analysis

The data from the left eye of each subject was used for further analysis. After the extraction of the first-order and the first slice of second-order responses (VERIS Science software for Macintosh Quadra; EDI, San Francisco, CA), all 61 and 103 traces were exported and processed further on an IBM-compatible personal computer. Each of the 61 and 103 responses of all subjects was then examined individually; three cursors were placed automatically at signal positions N1 (first negative deflection), P1 (positive deflection), and N2 (second negative deflection). Implicit times were calculated as the time difference between peak P1 and stimulus onset. Trough-to-peak amplitude was measured from P1 to N1. The age group medians for all 103 or 61 locations were used to extrapolate implicit time and amplitude topographies of first- and second-order kernel data, as previously described. ²⁰ In contrast to a ring analysis, the amplitudes in this type of representation are—similar to those in a trace array—not normalized to the area of stimulated retina, to enable visualization of local differences (unit nV instead of nV/deg²). 

Statistical Analysis

The ring data of each age group was analyzed for a potential age effect, to strengthen the evidence from the plot data. As the distribution of the data was skewed and presumably not normal, the statistical analysis was performed with the nonparametric Jonckheere-Terpstra test for trend. ²¹ The resultant levels of significance obtained for each ring group were then subjected to a Bonferroni correction (n = 6, see Table 1 ; n = 5, see Table 2 ). The direction of a potential trend is indicated in the tables by an upward arrow (increase across the age groups) or a downward arrow (decrease of the parameter with age). When the corrected level of significance was P < 0.05, filled arrows were used; otherwise, open arrows were used. 

The calculated percent change per decade, on the other hand, appeared to be normally distributed, allowing significance to be calculated with a Student’s t-test. Analyses were performed with commercial software (JMP 5.0; SAS Inc., Cary, NC). 

Results

In the top panel of Figure 1 , we show an example of a first-order kernel trace array (left) and normalized averages for retinal rings concentric to the fovea (right) that was recorded from a left eye stimulated with 103 hexagons. The P1 amplitude response (P1–N1 amplitude) ranged from 51.8 nV/deg² for the central retina (3° diameter) to 10.7 nV/deg² for the outer retinal ring (19–35°). P1 implicit time was ∼30 ms. An example of the second-order (first slice) analysis of a left eye for a stimulus with 61 elements is shown in the bottom panel of Figure 1 . The trough-to-peak amplitude in this case was much lower: 5.2 nV/deg² in the central area (4° diameter) and 2.8 nV/deg² in the outermost ring (17–30°). P1 implicit time was ∼31 ms. 

The topography of the peak P1 amplitude response and the topography of the P1 implicit time response were calculated for each of the 103 or 61 elements for both stimuli and for each kernel, by using the median calculated for each age group. In the following section, we focus on the 103-element data for the first-order analysis (due to the higher spatial resolution) and on the 61-element data for the second-order analysis (due to higher signal-to-noise ratio). 

First-Order Kernel Analysis

The visual field view of the extrapolated topography of the median first-order kernel P1 amplitude response is shown in Figure 2 , depicted as left eyes for each age group. The amplitudes are color-coded, going from red for the highest deviations (436–450 nV) through yellow, green, and blue to gray, indicating the lowest amplitudes (198–212 nV). In the youngest age group of 10- to 19-year-olds, there was a continuous decrease in amplitude from the foveal maximum in the center of the plot toward the periphery, apart from in the temporal retina (right) where there was a second area of maximum response. A minimum can be seen in the area of the blind spot. The amplitudes tended to be higher in the lower visual field (superior retina) than in the upper (inferior retina) and higher in the temporal retina than in the nasal. 

With age, there was a continuous loss of the average response amplitude in all retinal areas, but amplitudes in the upper-temporal retina (lower right) remained higher than those in the other quadrants. The youngest age group had the highest amplitudes (foveal maximum ∼450 nV, minimum in the nasal retina ∼310 nV), and in the oldest age group, the amplitudes were the smallest over the entire retinal area measured (central area, ∼290 nV; minimum in the nasal retina, ∼198 nV). 

Amplitudes generally show large intersubject variation, especially in the fovea. To analyze for age-related alterations, we tested for a potential trend of amplitudes and implicit times as a function of age when grouped into six concentric rings. In the top half of Table 1 , we show the median amplitudes calculated for each ring and age group, along with their ranges. We found a marked decrease of amplitude with age, and this trend was statistically significant in all the ring groups. 

The topography of the implicit time responses of all age groups is shown in Figure 3 , color coded for time: red for 32.3 to 32.55 ms to gray for 28.05 to 28.3 ms. Longer implicit times were generally found at the macula, around the optic nerve head (left), at the papillomacular bundle between the macula and optic nerve head and at the borders of the stimulated region. Shorter latencies were present in the area encircling the macula and most prominently in the temporal retina (right). Prolonged implicit times were found in older age groups (maximum in the fovea and the area of the blind spot, ∼32.5 ms; minimum in the temporal retina, ∼29.9 ms). The shortest implicit times were found in the groups with younger subjects (maximum in the fovea and the area of the blind spot, ∼30.6 ms; minimum in the temporal retina, ∼28.05 ms). 

The median implicit times and ranges, calculated for a ring analysis, are shown in the lower half of Table 1 . Again, the results were tested for a potential trend with age. We found a substantial increase in implicit time that was significant (3/6) or at the border of significance (i.e., close to but above P = 0.05, 2/6) in all but one ring group (1/6). There was little interindividual variability of implicit times. 

We additionally calculated for each hexagon the rate of change per decade, using linear regression analysis on the median data from each age group. In Figure 4 , left, change in first-order kernel amplitudes is depicted, color-coded for percentage of change in amplitude/decade, from yellow for 6.7% to 7%/decade down to gray for 0% to 0.5%/decade. The percentage of change in amplitude/decade is largest for hexagons at the center of the visual field (>8%/decade), at the upper visual field border (>7%/decade) and in the upper nasal visual field (6.8%–7.8%/decade), and lowest around the blind spot and in the lower temporal visual field 4.0% to 5.0%/decade. In the top half of the visual field (inferior retina) the amplitudes decrease faster than in the lower field half (6.19%/decade vs. 5.51%/decade; P < 0.001), and in the nasal half faster than in the temporal half (6.04 vs. 5.65%/decade; P < 0.03). 

On the right in Figure 4 , we show the average percentage of change in latency per decade for the first-order kernel data, based on the linear regression on median data from each age group for each hexagon. The average percentage of change in latency per decade was ∼1 ms (minimum, 0.1 lower nasal visual field; maximum, 1.8 upper nasal visual field). In this case, the maximum decrease was not in the center of the visual field (∼ 0.9%/decade), but in hexagons in the top nasal visual field (∼1.25%/decade). The percentage of increase in implicit time over six decades is significantly greater in the upper visual field than in the lower (1.13% vs. 0.94%/decade; t-test: P = 0.005). Nasotemporal differences are not significant. 

Second-Order Kernel Analysis

We show in Figure 5the topography of median amplitudes generated by a 61-element stimulus for the second-order kernel analysis in all age groups. In the youngest age group, amplitudes were generally largest in the nasal retina (left) with the maximum in the area of the papillomacular bundle. The minimum value was found in the temporal parafovea (right). Of interest, the highest group amplitudes were present, not in the youngest age group, but in the group of 20- to 29-year-olds (maximum in the area of the papillomacular bundle, 185 nV; minimum in the temporal retina, ∼100 nV). With increasing age, there was a steady general loss of group amplitudes in all regions. In the oldest age group, the amplitudes were the smallest (∼115 nV in the nasal retina and ∼68 nV in the temporal retina). 

In the top half of Table 2we list the median amplitudes calculated for retinal rings for each age group, along with their ranges. As before, we tested for a potential trend of amplitudes and implicit times as a function of age when grouped into concentric rings. Since the overall trend for an amplitude loss with age had more weight than the initial complexity described in the previous paragraph, the decreases in second-order kernel amplitude were, as in the case of first-order kernel data, significant in all ring groups. Like the first-order kernel, amplitudes showed a large intersubject variability, especially in the visual field center. 

We did not calculate the percentage of change in amplitude per decade for the second-order data, as they did not behave uniformly across age. 

The P1 implicit time topography of the second-order analysis is shown in Figure 6 . Longer latencies were found in the foveal region and on the borders of the stimulated region. Shorter values encircled the fovea, and temporal signals tended to be delayed compared with the nasal. For the oldest subjects, responses are especially delayed in the fovea. 

In the lower half of Table 2we list the median latency of P1 for the second-order kernel ring analysis. There were no uniform alterations with age, and only the outermost ring showed a significant tendency for continuous implicit time increase. 

Discussion

First-Order Responses

The first-order response kernel has been shown in monkeys to be generated in the outer retina from potential alterations in the photoreceptors, bipolar cells, and Müller cells, ^{17 22} although a ganglion cell component has also been identified. ^{23 24}  

Amplitudes.

The topography of first-order kernel P1 amplitudes that we found (Fig. 2)was generally in line with that reported in other studies ^{7 12 19 25} : The largest P1 amplitudes were in the fovea and temporal retina in all age groups, and P1 was larger in the upper retina than in the lower. We also found that amplitudes were greater in the temporal retina than in the nasal. The topographical pattern of the nasal–temporal and inferior–superior responses resembled the position of the large retinal vessel arcades (Fig. 2) . It is beyond the scope of this study to explore this further, but one may speculate that vascular (e.g., hydrostatic) differences between the upper and lower retina may exist and correlate with the functional differences. Alternatively, the distribution of photoreceptors, which is known to show superior–inferior gradients in many mammalian species including humans, ²⁶ may contribute to the topographical differences. 

The results show a progressive loss with age of the P1 amplitude of the mfERG for first-order responses (Fig. 2) . Although the loss occurred in all retinal regions, it was more prominent in the fovea and lower nasal retina. The percentage of decrease in amplitude per decade (Fig. 4 , left) showed a wide variation over the retina, ranging from 6.8% to 7.8%/decade around the macula and in the lower nasal retina to 4.0% to 5.0%/decade around the blind spot and in the upper temporal retina. For comparison, other estimates in the literature are an average of 10.5%/decade over all hexagons, ¹³ 7.5%/decade around the fovea, and 3.3%/decade in the periphery. ¹⁹ We found that the amplitudes decreased significantly faster in the lower retina than in the upper, although there were topographical differences within each hemifield. Tzekov et al. ¹⁹ report a faster decline in the upper retina than in the lower using Burian-Allen contact lens electrodes. It remains open whether the difference in electrodes ²⁷ —for example, the known tendency of contact electrodes to tilt downward slightly may be involved in this discrepancy. 

The standard ring analysis of our data (Table 1)shows significant evidence of an amplitude decrease together with a prolongation of implicit times in the central retina, confirming other studies showing a prominent decline of central vision with age. ^{8 10 13 19} The results of our study show, however, that a more sensitive data analysis based on anatomic and functional grounds may be superior to the standard geometrical ring shapes, in particular with respect to nasal–temporal and inferior–superior differences. 

There are many possible causes of the decrease in amplitude with age. Undetected preretinal media changes that reduce retinal illuminance and increase intraocular scattering, or reductions in photopigment density have been shown in other studies to fail to explain such a large amplitude decrease. ^{4 11 12 15 28 29} It has been estimated that less than 20% of the changes between younger and older age groups may be caused by preretinal media alterations. Seiple et al. ¹³ find that the mfERG amplitude reduction with increasing age is almost completely due to neural factors, whereas Tam et al. ¹⁸ reported no significant age difference in a comparison between young phakes and old pseudophakes, but did find a difference in a comparison between two groups of old pseudophakes, which raises the question of whether neural or metabolic processes are responsible for the early age-related decrease of mfERG response amplitudes. The influence of other factors that affect ERG amplitudes, such as refractive errors and intraocular pressure, were minimized by suitable subject selection. 

Damage at, or before, the bipolar cells will substantially decrease the amplitude of the mfERG. ²² The decrease in P1 amplitude that we find may thus reflect, at least in part, age-related deterioration in outer retinal function. The loss of sensitivity may be due to structural and morphologic changes in the photoreceptors or retinal pigment epithelium during aging but bipolar or Müller cell death could also account for much of the amplitude decline. ^{30 31 32 33 34 35 36 37 38 39 40 41} Curcio and Drucker ⁴² observed no significant age-related changes in amacrine cells within the central 5 mm of the retina. It has been proposed that the accumulation of lipofuscin in the older eye may be involved in the amplitude loss. ¹⁹  

Latencies.

The topography of first-order kernel implicit times (Fig. 3)is also similar to that described in previous mfERG examinations of retinal activity. ^{19 20 25} In the youngest age group the longest implicit times were found in the macula and around the blind spot, but latency did not change very much over the visual field, as noted by Parks et al. ⁴³ In addition, P1 was delayed at the upper and lower borders of the field. Again, the topography could be influenced by the large vessel arcades. In addition, there is a temporal-nasal gradient reflecting the impact of the second order signal (Fig. 3)on the first-order waveforms as has been shown in glaucomatous monkeys lacking second-order contributions (Seeliger MW, et al., IOVS 2001;42:ARVO Abstract 788). 

The overall percentage of change per decade in latency (Fig. 4 , right) ranged from 1.8 to 0.12 ms, which is within the range reported elsewhere. ^{10 11 12 13 18 19 20} There was a steady increase in implicit times for first-order kernel responses with age in the whole area, a trend found to be significant for many rings in our study (Table 1) . However, the upper visual field altered faster with age than did the lower, and our topographical analysis revealed many details hidden in the standard ring group data, supporting the need for a more sensitive measure of age-related alterations. 

Prolonged mfERG latencies in older patients may be caused by a slowing of the regeneration of photopigments in the older eye, as photoreceptors are affected by a gradual metabolic dysfunction of the retinal pigment epithelium. ^{12 44} A slowing of temporal adaptation in the older aged retina may also play a role. ¹² Instability of fixation in older patients can also cause an increase in implicit time, although in this study, recordings were monitored and the results—for example, Figure 3 —show that there were no such problems significantly contaminating the data. 

The interindividual visual variability is less for implicit times than amplitudes as previously reported. ^{19 20}  

Second-Order Kernel (First Slice)

Amplitudes.

The second-order kernel responses of the mfERG are generated proximal to the photoreceptor outer segment in the outer plexiform layer and reflect mechanisms involved in the short-term adaptation process. ^{16 17} Our results showed an amplitude topography similar to those in previous reports of second-order responses. ^{9 10 12 15} In contrast to the results of the first-order kernel analysis, amplitudes were highest in the nasal retina in the area of the papillomacular bundle, and there was no clear difference between upper and lower retinal responses (Fig. 5) . 

The P1 amplitude of the mfERG showed overall a progressive decline with age of second-order responses. The loss occurred in all retinal regions, and an analysis for trend showed significant changes for all ring groups (Table 2 , top panel). The problem of ring formation becomes particularly apparent: The topography of the second order data (Fig. 5) , featuring a peak and trough NOT centered on the fovea, is little represented in the ring data, as both peak and trough fall into the same ring and thus mostly cancel out each other. 

Older subjects are in a more light-adapted state than younger ones, ¹² and the reduction in second-order kernel amplitude with age indicates that an abnormality in the circuits involved in adaptation may develop in the older eye. It is possible that the impaired adaptation is caused by factors affecting the neural integration and blood supply of the retina. ¹²  

It is surprising that the highest amplitudes were not found in the youngest age group, indicating that there is possibly a slower maturation of inner retinal function. Several processes resembling relatively slow maturation of brain tissues have been described (see e.g., Ref. ⁴⁵ ). Also, the vitamin A/retinoic acid metabolism does show age-related topographical differences that may have an impact on postnatal retinal function. ^{46 47} In particular, Sakai et al. ⁴⁶ show that the observed alterations are located preferentially in the inner retina. At this stage, we can of course only speculate about possible mechanisms. 

Latencies.

The implicit time of the second-order kernel peak shows, as for the first-order kernel, little topographical alteration (Fig. 6) . There are also no significant alterations with age in a ring analysis, except for a delay in the most peripheral group. This may correspond to the significant delay that has been described for subjects above 50 years. ⁹  

In conclusion, our data show a progressive loss of mfERG P1 amplitudes and an increase in their implicit times with age in most retinal regions for both first- and second-kernel analysis. The degree of change is dependent on retinal location, and first-order kernel alterations may be influenced directly or indirectly by the large vessel arcades. Information about changes in discrete retinal areas with age should therefore be considered when analyzing mfERG traces for age-related alterations. 

 Supported by a German Academic Exchange Service (DAAD) grant, a grant from the University Tübingen (Nachlass Kurt Max von Siegl Edler von Geregelfels) (HL), and DFG (Deutsche Forschungsgemeinschaft) Grants Se837/1-2 and Se837/6-1 (MWS).

 Submitted for publication October 10, 2007; revised November 30, 2007, and April 19, 2008; accepted August 27, 2008.

 Disclosure: H. Langrová, None; E. Zrenner, None; A. Kurtenbach, None; M.W. Seeliger, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 7.

Figure 8.

Figure 1.

 

Top left: averaged first-order kernel trace arrays for 103 stimulus elements recorded form the left eye of a subject from the second age group (20–29 years). Top right: normalized group averages for six concentric rings. Ring 1, ∼0–1.5°; ring 2, 1.4–6°; ring 3, 4–11°; ring 4, 8–18°; ring 5, 13–26°; and ring 6, 19–35°. Bottom: Example of second-order kernel trace arrays for 61 stimulus elements (left) and normalized group averages for five concentric rings (right). Ring 1, ∼0–2°; ring 2, 1.8–7°; ring 3, 5.0–13°; ring 4, 11–22; and ring 5, 17–30°.

Figure 2.

 

First-order kernel: visual field view of peak P1 amplitude (in nanovolts) topography constructed from the median of the first-order kernel response component. The maps are depicted as left eyes for each of the six age groups.

Figure 3.

 

First-order kernel: visual field view of peak P1 implicit time topography (in milliseconds) of the left eye in normal subjects constructed from medians of each age group.

Figure 4.

 

Map of percentage change per decade in first-order kernel amplitudes (left) and latencies (right).

Figure 5.

 

Second-order kernel: visual field view of peak P1 amplitude topography (in nanovolts) of the left eye in normal subjects, constructed from medians of each age group.

Figure 6.

 

Second-order kernel: visual field view of peak P1 implicit time topography (in milliseconds) of the left eye in normal subjects, constructed from medians of each age group.