Long-Term Decline of Central Cone Function in Retinitis Pigmentosa Evaluated by Focal Electroretinogram

Abstract

Purpose.: We evaluated long-term changes of central cone-mediated function in retinitis pigmentosa (RP) patients by recording focal electroretinograms (fERG).

Methods.: A cohort of 43 RP patients was followed from 4 to 16 years (average follow-up 9.3 years, average 10 examinations/patient) by recording the fERG response to a flickering uniform red field overlaying the central 18° of visual field (VF). Statistical censoring led to a reduced dataset of 32 patients (autosomal dominant 9, recessive 5, sporadic 5, x-linked 1, Usher II 12), from which long-term decay rates were estimated by global fitting of individual fERG amplitude time-curves.

Results.: Long-term follow-up of central cone fERG amplitude showed two main features: short-term variability and long-term decline. fERG short-term variability range was 0.14 to 0.2 log units. Mean yearly decay rate of central fERG was 5.6% (95% confidence interval [CI] 4%–7%). Yearly decline depended on inheritance pattern, being significantly greater in autosomal recessive and sporadic compared to autosomal dominant RP. The degree of central cone fERG decline was unrelated to the size of the residual VF.

Conclusions.: The decline of central cone function is significantly slower than global cone function decline in RP. Central cone fERG loss is independent of residual VF.

Introduction

Retinitis pigmentosa (RP) is a heterogeneous group of inherited retinal diseases causing primary degeneration of rod photoreceptors and secondary critical degeneration of cone photoreceptors. ¹ In early stages of disease, the function of either peripheral or central cones may be abnormal, and it tends to deteriorate as the disease progresses. ^2–5 At histopathologic levels, cone degeneration occurs in all genetic forms of RP, including autosomal dominant RP (ADRP) caused by rod-specific genes, such as rhodopsin. ¹ Cone photoreceptors mediate color and high acuity vision, providing daylight vision. Many of the RP genes are expressed only in rods, but cones still malfunction and die. The nonautonomous death of cones likely is due to a common problem(s), as it is seen in all species where there is a rod-specific gene defect. 

The Ganzfeld (i.e., whole-field) electroretinography (ERG) has been used as an objective and direct assay of retinal cone function in longitudinal ^2,3,6–8 and cross-sectional ^9,10 RP studies, leading to a well-defined time frame for overall disease progression. However, because the Ganzfeld ERG is a summed response from the entire retina, it does not provide a direct evaluation of how the function of specific regions of the retina degenerates. This is relevant particularly for the central retina, which mediates some of the most specialized functions of the cone system, including visual resolution and fine spatial discrimination. 

In the natural history of most RP subtypes, the central retina appears to be the last retinal region to lose functionality, ⁶ although functional alterations in this region are detectable since early in the course of the pathology. ^4,5 Recent findings in animal models, reviewed by Punzo et al., ¹¹ showing that the rate of cone degeneration is related inversely to receptor density, suggest that central cone function should be more robust than global field function in RP, because cones density peaks in the central retina in humans. In agreement with this expectation, central islands of almost intact cones are preserved in a significant fraction of RP patients, ¹² and visual acuity (VA) declines later and less sharply than does global field ERG. ^{2,7,8,13–16} Yet, this issue becomes unclear when objective ERG measures are considered. In the pioneer three-year study of RP natural history, Berson et al. estimated an annual decay rate of 17% for global ERG and a 5% decay for central cone ERG. ³ This trend for a slower central decay is documented by a multifocal ERG study, showing that central region recodings correlate with VA and more external regions decay as the Ganzfeld ERG. ¹⁶ However, additional studies, using longer follow-up and larger patient cohorts decreased the decay rate estimate for global ERG to 7% to 13% per year, ^1,7,13–15 and increased the estimate for central ERG decay to 10%. ⁸ Thus, at present contrasting estimates exists for global and central cone ERG decay rate. 

Biologic and technical factors have been identified at the basis of these discrepancies. These include genetic and epigenetic heterogeneity of the patient cohorts examined, better estimate accuracy provided by long-term studies compared to cross-sectional or short-term studies, and the use of different criteria for data censoring and analysis (for an in depth discussion, see the report of Berson et al. ¹³ ). In addition, when comparing central and global cone function, further difficulties derive from the existence of a much larger set of studies evaluating global ERG, ^{2,3,6–10,13–16} compared to those addressing central ERG decay. ^3,5,8,16  

In light of these considerations, we thought it useful to provide here a quantification of central cone functional decay based on the long-term follow-up of a patient cohort of known RP inheritance patterns, using censoring and analysis criteria derived from previous global ERG studies. ^6,13–15 Patients were monitored with a central cone focal ERG recording (fERG), whose first harmonic amplitude has been shown to correlate with perimetric sensitivity in the same retinal region (Iarossi et al. ¹⁷ ). The long-term decline in fERG amplitude was used to estimate central cone function decay rate, while the relationship between fERG amplitude and visual field (VF) extent was used to investigate whether central function decay is related temporally to global field decay. 

Methods

Patients

A group of 43 RP patients (21 males, 22 females) from central and southern Italy, were selected from the database of patients followed clinically at the Visual Electrophysiology Service of the Institute of Ophthalmology at Università Cattolica del S. Cuore, and included in the study, based on a minimum follow-up time of 4 years, and a minimum of 3 visits. Patients had sought consultation because of visual symptoms. Following the first visit, they were invited to adhere to the institutional schedule of one examination per year. The actual date of each visit was established by an independent administrative office. All patients gave informed consent to participate in the study, which adhered to the tenets of the Declaration of Helsinki and was approved by the Universita' Cattolica Review Board. 

All patients had progressive forms of RP based on history, clinical findings and ERG abnormalities. The data set derived from 437 ocular examinations performed from 1995 to 2012 under the same test conditions. Table 1 summarizes individual patient baseline features. 

Table 1. 

 

Patient Baseline Visual Characteristics

Patient	Sex	Inheritance Pattern	Age at Baseline	Follow-Up, y	No. of Visits	VA (RE)	VA (LE)	Baseline fERG (μV)
1	M	ADRP	6	14	8	0.9	0.9	1.86
2	F	Usher	14	4	4	1.0	1.0	1.73
3	M	Recessive	74	8	10	0.8	0.9	1.66
4	M	Usher	37	8	3	1.0	1.0	1.63
5	F	ADRP	33	6	4	0.8	0.8	1.38
6	M	Usher	17	11	17	1.0	1.0	1.12
7	M	Recessive	49	7	8	1.0	1.0	1.09
8	M	Sporadic	16	11	15	0.9	0.8	1.00
9	M	Usher	22	12	10	0.9	0.8	0.96
10	M	Sporadic	26	12	19	0.8	0.9	0.93
11	F	Sporadic	18	15	15	1	1	0.89
12	M	Usher	14	11	11	0.3	1.0	0.76
13	M	Usher	23	10	19	1.0	1.0	0.69
14	F	ADRP	43	12	7	1.0	1.0	0.62
15	F	Recessive	32	6	6	0.3	0.3	0.59
16	M	Usher	12	9	11	0.8	0.7	0.59
17	F	Recessive	57	12	12	1.0	1.0	0.58
18	F	ADRP	36	11	17	0.2	0.2	0.58
19	F	X-linked	76	11	5	0.2	0.2	0.54
20	M	Usher	17	12	15	1.0	1.0	0.50
21	F	ADRP	25	5	6	0.8	0.6	0.46
22	F	Usher	47	4	4	0.1	0.3	0.46
23	M	ADRP	38	10	7	0.4	0.5	0.46
24	F	Sporadic	6	16	10	0.7	0.4	0.44
25	F	Usher	55	7	8	0.1	0.3	0.44
26	F	Recessive	49	13	17	0.2	0.4	0.44
27	F	Sporadic	19	10	15	1.0	1.0	0.43
28	M	ADRP	18	14	18	0.8	0.8	0.41
29	F	Usher	51	9	13	0.5	0.5	0.40
30	M	Usher	50	10	11	0.7	0.7	0.39
31	F	ADRP	31	15	17	1.0	1.0	0.36
32	F	ADRP	32	10	20	0.03	0.1	0.32
33	M	Usher	44	5	3	0.3	0.3	0.27
34	M	X-linked	57	5	4	0.1	0.1	0.27
35	M	X-linked	9	9	6	0.5	0.5	0.27
36	F	Usher	27	4	4	0.1	0.1	0.26
37	F	ADRP	30	12	12	1.0	1.0	0.26
38	M	Recessive	16	6	6	0.4	0.6	0.25
39	F	X-linked	38	9	10	0.1	0.3	0.22
40	M	Sporadic	59	4	4	1.0	0.1	0.21
41	F	Usher	45	8	12	0.2	0.2	0.21
42	F	Sporadic	58	9	7	0.1	0.080	0.17
43	M	X-linked	52	6	8	0.6	0.5	0.13

 fERG amplitude refers to the first harmonic amplitude averaged between the two eyes. Patients are ordered by their fERG baseline values. M, male; F, female; RE, right eye; LE, left eye.

Inclusion Criteria

Patients met the following inclusion criteria: Typical RP was present with a rod-cone pattern of retinal dysfunction, as determined by standard Ganzfeld ERG, dark-adapted Tübinger perimetry, and classic fundus appearance, and an inheritance pattern determined unequivocally by a detailed family and medical history. VF by Goldmann V/4e was >10° at baseline. Patients had a known inheritance pattern and/or genotype under study. Patients underwent at least four years of fERG and clinical examination follow-up, with a minimum of 3 visits. There were no or minimal ocular media opacities, and no concomitant ocular (e.g., glaucoma, amblyopia) or systemic diseases. Patients with non-Usher syndromic subtypes of RP, Leber's congenital amaurosis or early onset RP with atypical functional patterns were not included. 

Measures of Ocular Function and ERG

A full general and ophthalmologic examination (including detailed family history, anterior segment biomicroscopy, corrected Early Treatment Diabetic Retinopathy Study [ETDRS] VA, direct and indirect ophthalmoscopy, intraocular pressure measurement) was performed on each patient at baseline and on several consecutive visits. 

Best corrected VAs were obtained with a projected Snellen chart. Kinetic VFs were measured to the V4e white test light of the Goldmann perimeter against the standard background of 31.5 apostilbs. Fields were digitized and converted to areas and equivalent diameters 

Cone fERGs were recorded from the central 18° region using a uniform red field superimposed on an equiluminant steady adapting background, used to minimize stray-light modulation. ^17,18 The stimulus was generated by a circular array of eight red LEDs (λ maximum 660 nm, mean luminance 93 cd/m2) presented on the rear of a Ganzfeld bowl (white-adapting background). A diffusing filter in front of the LED array made it appear as a circle of uniform red light. fERGs were recorded in response to the sinusoidal 95% luminance modulation of the central red field. Flickering frequency was 41 Hz. Patients fixated monocularly at a 0.25° central fixation mark, under the constant monitoring of an external observer. Pupils were pharmacologically (1% tropicamide and 2.5% phenylephrine hydrochloride) dilated to a diameter of ≥8 mm, and all subjects underwent a preadaptation period of 20 minutes to the stimulus mean illuminance. fERGs were recorded by an Ag-AgCl electrode taped on the skin over the lower eyelid. A similar electrode, placed over the eyelid of the contralateral patched eye, was used as reference (interocular recording). fERG signals were amplified (100,000-fold), bandpass filtered between 1 and 100 Hz (6 decibels/octave [dB/oct]), and averaged (12-bit resolution, 2-kHz sampling rate, 200–600 repetitions in 2–6 blocks). Off-line discrete Fourier analysis quantified the amplitude and phase lag of the response fundamental harmonic (first harmonic) at 41 Hz. For test–retest variability, patients were recalled within a 6-month interval of the time of a given examination to establish short-term variability for the measure of visual function. Six of the 32 patients in the reduced dataset did not perform the test. 

Data Analysis

Previous studies have shown that a simple exponential decay represents a good model for full field cone ERG decay in RP patients, and a good approximation for cone loss in animal models of RP. ¹³ For these reasons we used a single exponential decay model for central cone function changes. 

Because fERG amplitudes were skewed, individual fERG values were normalized to baseline and transformed into common (base 10) logarithms. Because an exponential decay for fERG means a linear decay for logfERG, we performed linear regression analysis of logfERG amplitude ratio as the dependent variable and time of follow-up as the independent variable to derive a slope (rate of change) for each patient during follow-up. 

Following previously described methods, ^13–15 the distribution of such slopes as a function of fERG at baseline was analyzed for “ceiling” and “floor” effects (see Results). Longitudinal data were censored to avoid these effects, and final slopes were recalculated from the reduced data set, including only patients with at least 3-year follow-up following censoring. Statistical outliers were identified from the longitudinal data using the generalized extreme Studentized deviate test (P < 0.05). Only two outliers were identified and they were discarded. Average decay then was derived as the mean of the individual slopes obtained from the censored data, ^13–15 as well as by global linear fitting. The latter let individual intercepts vary with patient and optimized the fit with a decay rate parameter (slope) common for all individual time-course curves. The advantage of this model is the simultaneous use of the entire dataset to determine a common average decay without external assumptions about weighting factors and modalities. 

We also used an alternative method to determine the average long-term decay. For each patient, we considered all the possible pairs of fERG measures and computed the ratio between the two measures of each pair (by dividing the more recent by the previous one). The rationale is that if the decay rate is constant, the ratio between any two measures from a patient provides an estimate of how much the fERG changed in the interval between such measures. Two factors contributed to this change: short-term variability and long-term decline. The former should vary from measure to measure, the latter should depend only on the interval between the two visits. By taking all possible measure pairs for each patient and plotting data from all patients as a function of the interval between the two measures of the pair, we expected to transform the contribution of short-term variability into a scatter of data points centered on the average long-term decay curve, which was extracted by averaging the distribution. Yearly averages obtained from at least 50 points were used to estimate the average decay rate. 

Statistical Analysis

Data were analyzed with Origin 8Pro (Microcal, Inc., Piscataway, NJ) and Graphpad Prism 5.04 (Graphpad Software, Inc., San Diego, CA). Unless otherwise specified, data are presented as mean ± SD. Normality test was performed using the D'Agostino-Pearson omnibus K2 test. 

In linear regression, residuals were analyzed by the D'Agostino-Pearson omnibus K2 normality test, and run tests. Fit results were considered only when the fit converged and P < 0.01. Following established convention, the quality of fit was recorded as adjusted coefficient of determination (AdjR ²) and P value. Values obtained by best fitted regression lines are reported together with their 95% confidence interval (CI). 

Results

The present analysis is based on the long-term follow-up of the 43 RP patients with diverse inheritance patterns (Table 1). Follow-up ranged from 4 to 16 years (mean 9.5 ± 3.2 years), with an average of 1 visit per year. The youngest patient was 6 years old at baseline, and the oldest was 87 years old at the final follow-up. The mean baseline age was 34.4 ± 18.5 years. 

Between-Eye Correlation in Central Cone fERG Amplitude

Most previous studies focus on visual measures averaged between the two eyes, because these tend to be correlated in the eyes of RP patients. To assess whether this is true for central cone fERG, we evaluated the correlation between the log amplitude of the fERG first harmonic component (from now on logfERG amplitude) of the left eye and that of the right eye at baseline (Fig. 1). The resulting Pearson correlation coefficient was 0.857 (95% CI 0.729–0.926, N = 47, two-tailed P < 0.0001). Therefore, in the following analysis, we averaged fERG amplitude between eyes for each visit. 

Figure 1. 

 

Between-eye correlation in central cone fERG amplitude. The fERG amplitude recorded from the left eye at baseline is plotted versus the corresponding right eye fERG for all patients in the dataset. Pearson correlation coefficient 0.857 (95% CI 0.729–0.926, N = 43, two-tailed P < 0.0001). The logarithms of fERG amplitude are plotted, because they follow a normal distribution (D'Agostino-Pearson normality omnibus K2 test, P = 0.28 left eye, P = 0.68 right eye, N = 43)

Overall Progression: Raw Data and Censoring

Examples of the time course of individual fERG amplitude are shown in Figure 2 for some RP patients with main inheritance patterns and an Usher patient. To allow comparison, each curve is normalized to baseline, and presented on a logarithmic scale. Two salient features are apparent: persistent short-term oscillations and a long-term decline. 

Figure 2. 

 

Time course of fERG amplitude in individual cases. Examples of fERG amplitude variation with time in some individual patients. Each point corresponds to a visit. To facilitate comparison, data are normalized to baseline and presented as logfERG.

To quantify such features, we first needed to censor data to avoid ceiling and floor effects deriving from ERG measures that do not reflect yet (or anymore) real disease progression. To this purpose we performed linear regression analysis of individual patient logfERG versus time, obtaining a slope for each patient (Fig. 3A). These slopes were transformed into individual rates of change in fERG amplitude, and plotted as a function of the patient baseline fERG (Fig. 3B). The plot showed no ceiling effect, because none of the patients with high baseline fERG (amplitude >1.5 μV) was more likely than other patients to demonstrate zero change (dashed horizontal line). On the contrary, a floor effect was evident at low fERG amplitudes (<0.27 μV; i.e., signal-to-noise ratio <5), where regressors were as common as progressors. Therefore, we censored all data from patients with baseline fERG amplitude <0.27 μV (dashed vertical line), as well as other patient follow-up data following the occurrence of a fERG amplitude <0.27 μV. The reduced dataset obtained following censoring consisted of 32 patients (patients 1–32 in Table 1), with a normal distribution of estimated annual change in fERG amplitude (D'Agostino-Pearson omnibus K2 normality test, P = 0.957). The average follow-up after censoring was 9.1 ± 3.35 years. 

Figure 3. 

 

fERG amplitude follow-up: raw data and censoring. (A) logfERG time curves (gray dots and lines) and individual best linear regression fit lines (black lines) are shown for all 43 patients. (B) The slope of each fit line shown in (A) is converted to annual percentage change in fERG amplitude and plotted as a function of the patient baseline fERG. The dashed horizontal line indicates zero change, the dashed vertical line the censoring fERG threshold of 0.27 uV.

The Short-Term Variability Range

Short-term variability in visual performances of RP patients has long been reported. ^2,3,19–21 To determine its range for central cone fERG amplitude, patients were invited to a visit within 6 months of a previous one. The two fERG amplitude measures thus obtained for patients in the censored dataset were highly correlated (Fig. 4, left; Pearson correlation coefficient 0.78; N = 26; two-tailed P < 10⁻⁵). Plotting these same data in the form of percentage difference in fERG measures versus their average (Fig. 4, right), we found that the distribution had a mean of 4.77%, its fifth percentile was −41.9%, and its 95th percentile was 81.2%. These values are in line with previous results. ^2,3,19–21  

Figure 4. 

 

Short-term variability range. Left: logfERG measures obtained in two consecutive visits no more than 6 months are highly correlated for patients in the censored dataset. The Pearson correlation coefficient was 0.78 (N = 26, two-tailed P < 10⁻⁶, one pair of measures for each patient). Average interval between measures was 104.9 ± 46.5 days. logfERG values followed a normal distribution (D'Agostino-Pearson normality omnibus K2 test, P = 0.55 [first rec], P = 0.59 [second rec], N = 26). Right: Bland-Altman plot of the difference versus the average of the fERG amplitudes recorded in the two visits. The mean percentage difference was 4.77%. The 5% percentile of the difference distribution was −41.9%, its 95% percentile was 81.2%.

Long-Term fERG Decay in RP

A first estimate of the average long-term decay rate of central cone fERG was obtained by averaging the individual rates of decline obtained by independent linear regression analysis of each patient follow-up within the censored dataset. The result was a 5.3% estimated annual decay of fERG amplitude (95% CI 4.0%–6.7%). This corresponds to a 50% loss of fERG amplitude (−0.3 log units) in 11.4 years (95% CI 9–15.4 years). This raw averaging does not consider that longer follow-up and more measures are likely to provide more accurate estimates. Thus, a second estimate of decay rate was obtained by global linear regression on the reduced patient dataset (Fig. 5A). In global fitting, the decay rate is shared by all curves and optimized to best fit all curves, while the intercept of each curve is optimized individually (see Methods). Global fitting estimated an average annual decay rate in fERG amplitude of 5.6% (95% CI 4.5–6.7%, P < 10⁻¹⁴, AdjR ² = 0.56). This estimate is very close to the previous one. 

Figure 5. 

 

Estimated average decay rate of central cone fERG in RP. (A) Time course of logfERG normalized to baseline for all patients in the censored dataset (black dots and lines). The thick gray line illustrates the shared best fit slope obtained by global fitting linear regression analysis of the censored dataset. The corresponding annual decay rate in fERG amplitude is 5.6% (95% CI 4.5–6.7%). (B) logfERG ratio values obtained from all pairs of fERG measures for each patient are plotted as a function of the time between the two measures (dots). White squares represent the yearly averages of this distribution for all years with more than 50 data points (years 0–9). The gray line is the best linear fit of the yearly averages. It corresponds to an annual decay rate of 5.68% (95% CI 4.63–6.72) in fERG amplitude.

Global fitting provided a relatively low value of AdjR ² (0.56). Considering the short-term variations of individual fERG curves (Figs. 3, 4), this was not surprising for a simple decay model. Indeed, a very similar estimate of decay rate and a much higher AdjR ² (0.87) are obtained when individual patient data are averaged on annual basis before global fitting (not shown). To confirm that short-term variability accounts for most deviation from a simple decay model, we used an alternative analysis. We reasoned that if decay rate is constant, the fERG change between any two measures from the same patient is determined by such constant decay and by short-term variability. The first contribution only depends on the time between the measures, the second should vary randomly with time. Therefore, taking the log ratio of any two fERG measures in each patient, plotting it as a function of the time between the two measures, and pulling all patient data together, we expected a distribution of points surrounding the mean long-term decay line (Fig. 5B). The latter was derived by averaging the data distribution on a yearly basis (open squares), and fitting the yearly averages with a linear regression model (gray line). This fit predicts an annual decay rate of 5.68% (95% CI 4.63–6.72, P < 10⁻⁵, AdjR ² = 0.89), in close agreement with our previous estimates. Finally, the estimated decay rate obtained by global fitting of the entire dataset was 5.6% (95% CI 4.3–6.9%, P < 10⁻¹⁴, AdjR ² = 0.38, N = 43 patients). This was very close to the estimates obtained from the censored dataset, but, as expected, it had a larger 95% CI and a lower AdjR ². 

Long-Term Decay of Central Cone fERG Depends on Inheritance Pattern

The average decay rates obtained by separate analysis of the recessive, sporadic, ADRP and Usher patient subsets in our reduced cohort are shown in Figure 6 and Table 2. The differences in decay rate were statically significant when the recessive or sporadic sets were compared with ADRP (P < 0.05, ANOVA, Holm-Sidak method Multiple Comparisons), and they correlate well with the different disease severity observed clinically. No estimate was obtained for X-linked forms, because most such patients had very low fERG amplitudes since baseline and failed censoring criteria. 

Figure 6. 

 

Estimated long-term decay of central cone fERG for different inheritance patterns. The estimated annual rate of central cone fERG amplitude decay (horizontal segments) and its 95% CI (gray area) are shown for the recessive, sporadic, ADRP, and Usher patient subsets. Values were determined by global fitting of the corresponding patient subsets (see also Table 2).

Table 2. 

 

Estimated Annual Decay Rate (and 95% CI) of Central Cone fERG for the Four Main Inheritance Patterns Present in the Reduced Data Set

	No. of Pts.	Mean Annual Decay Rate	95% CI LB	95% CI UB	AdjR 2	df	P Value
Recessive	5	7.3%	4.7%	9.8%	0.51	40	<10−13
Sporadic	5	8.3%	6%	10.6%	0.51	63	<10−13
ADRP	9	4.1%	2.1%	6.1%	0.51	84	<10−15
Usher 2	9	5.8%	4%	7.7%	0.31	94	<10−14

 Estimates are obtained by global fitting of the longitudinal follow-up of fERG amplitude versus time in the corresponding patient subsets. LB, lower bound; UB, upper bound; df, degrees of freedom.

No significant difference in average decay rate was observed when patients were subdivided in subsets according to sex, age at baseline, or initial fERG value (data not shown), indicating that these variables do not affect decay rate significantly, in accordance with previous studies. ^{2,6–8,13–16} Similarly, we did not observe any clear tendency in implicit time variation with time, in accordance with previous studies. ⁹  

VA and VF Follow-Up

Longitudinal changes in VA and VF area were evaluated following censoring according to published criteria. ¹³ The resulting annual rates of loss obtained by global fitting were 8.2% for VF area (95% CI 3%–13.3%, N = 20 patients), and 3.8% for VA (95% CI 2.7%–5.0%, N = 31). 

Central Cone Function Decay is Unrelated to the Extent of Residual VF

Finally, we tested whether the degree of central cone functional loss was related to the progression of RP as determined by VF restriction. Using baseline values from the present cohort and 43 additional RP patients from our clinic, we found no relationship between VF radius and fERG amplitude (Fig. 7), indicating that central cone function decay is independent of the extent of residual VF in RP. 

Figure 7. 

 

fERG amplitude was not correlated to residual VF size in RP. Baseline values of fERG are plotted as a function of VF radius for 86 RP eyes. Squares are left eye data, circles are right eye data. Pearson correlation coefficient 0.06 right eye and 0.09 left eye.

Discussion

We presented here the analysis of long-term follow-up of central cone fERG recordings in 43 RP patients followed between 4 and 16 years. Three different methods (average individual fit, global fit, and multipuncta analysis), led to very similar estimates of fERG amplitude decay rate, indicating an average 5.6% annual loss of central cone function in RP patients (95% CI 4–7%). When considering individual inheritance patterns, in agreement with clinical observations, slower progression rates were found in ADRP and Usher patients, while the worst predicted outcome was for recessive forms. Sporadic forms had a decay rate similar to recessive forms, in agreement with the former being found commonly to be recessive, once the specific case mutation is characterized. 

An annual decay rate between 4% and 7% is in good agreement with previous estimates of central cone function obtained with smaller stimuli, including a 5.2% ³ rate obtained with a flickering white 4° stimulus, a 10% decay rate (based on individual data) obtained with a flickering 9° red light, ⁸ and a 6% to 10% decay rate obtained for the central 6° ring of the multifocal ERG. ¹⁶ The usefulness of the present ERG assay is to extend beyond the few degrees of the central retina and to correlate with local perimetric sensitivity. ¹⁷  

Several studies suggest that central cone function decays less rapidly than global cone function. ^3,8,12,16 Yet, when considering published ERG studies (Fig. 8A), this issue was unclear, because of the large discrepancies across existing estimates. Earlier studies, based on small cohorts and limited follow-up have provided the first crucial estimates for clinical trials. ^2,3,7,8 As larger patient databases and longer disease follow-ups were collected, slower decay rates were obtained than estimated originally. ^2,13–16 These discrepancies had biologic and technical explanations deriving from the genetic heterogeneity of the patient cohort analyzed, the higher accuracy of decay estimates provided by long-term follow-up studies when compared to cross-sectional or short-term analysis, ¹³ the sample size, and the use of censoring. 

Figure 8. 

 

( A, B) Summary of estimated decay rates for central and peripheral ERG in RP. Estimated annual decay rates of whole field (light gray) and central cone ERG (dark gray) reported in the literature for mixed patient cohorts (A), as well as for selected inheritance patterns (B). When indicated in the study, the 95% CI is illustrated. Asterisks indicate studies using censoring. (a) 3-year follow-up, 90 patients, 3 examinations/patient. ³ (b) 5.2-year mean follow-up, 90 patients, 5 examinations/patient (censored: high amplitude control group). ² (c) 4-year mean follow-up, 64 patients, 4 examinations/patient. ⁷ (d) 9-year mean follow-up, 26 patients, 9 examinations/patient. ⁸ (e) 8.9-year mean follow-up, 134 patients, 6.2 examinations/patient. ¹³ (f) 10.4-year mean follow-up, 125 patients, 7.4 examinations/patient, censored. ¹⁵ (g) 6.3-year mean follow-up, 23 patients, 2.8 examinations/patient, ¹⁵ multifocal ERG (light gray indicates data from the inner 6° ring, dark gray indicates data from the 3 outermost rings [18°–66°]). (P) Present study: 43 patients, 9.5 year follow-up, 10 examinations/patient, censored.

As a consequence, different studies (and the estimates they provide) are more or less comparable on the basis of their methodologic similarity, in terms of the temporal analysis used, the size and frequency of sampling, and the use of censoring. With this in mind, the studies to which the present analysis is closest methodologically are those concerning large single inheritance pattern cohorts reported, ^13–15 as well as the analysis of the mixed high amplitude cohort used as a control in a large clinical trial testing Vitamin A. ² These studies present the same temporal long-term follow-up as our study, and use large samples enabling the same censoring criteria to which we informed our study. A practical demonstration of this methodologic similarity is that our estimates of VA and VF size decay rate are very similar to those reported in the aforementioned studies. 

When we restrict the comparison to studies involving large samples, long-term follow-up and censoring (indicated by asterisks in Figs. 8A, 8B), it becomes very clear that central cone function decays more slowly (our study) than does global field ERG. ^2,13,15 In particular, for a patient cohort including three main inheritance patterns and a syndromic subtype, central cone function annual decay rate is 4% to 7% (95% CI of our study), while global field ERG annual decay is in the 9% to 11% range. ² Thus, we believe that, beyond technical and biologic issues, our study provides strong evidence for a slower central cone functional decay when compared to global field visual decay in RP patients. 

Recent studies in animal models and patients support the expectation that central cone function decays less rapidly than overall visual function in RP. Cone death in RP is secondary to rod degeneration. Different reasons have been proposed for this “bystander effect,” including loss of trophic support supplied by rods, release of toxic agents by dying cells, cone metabolic dysregulation, and/or oxidative stress induced by rod loss (reviewed by Punzo et al. ¹¹ ). All these hypotheses suggest that the deleterious effect of rod loss relates to the local rod–cone ratio and, thus, is lower in the central retina. In accordance with this expectation, cell loss is related inversely to cell density in animal models of RP, ^22–24 and islands of functionally normal cones are found in the central retina of a significant fraction of RP patients. ¹²  

Although decaying more slowly than global cone function, fERG amplitude appeared unrelated to global visual loss, as determined by residual VF measures. In particular, this meant that fERG amplitude already could be significantly lower than normal in patients still retaining a considerable extent of VF. This agrees with previous studies showing that the onset of central cone function decay can occur early in the course of the disease. ^4,5  

A characteristic feature of all patients' fERG time course was persistent short-term variability. This certainly partly reflects technical aspects, such as small differences in electrode coupling from visit to visit. Yet, the same range of variability has been observed using different techniques, ^3,19–21 and appears to have psychosocial components, ²⁵ suggesting the contribution of biologic factors, such as, for example, the variable performance of cones that still are viable, but already are functionally impaired. Short-term variability accounts for fERG variations up to 40% to 80%. Understanding its biologic components could allow designing new therapeutic strategies to slow down central cone degeneration in RP. 

Acknowledgments

We thank the study patients and their families for their cooperation and trust; and Adriana Fiorentini, Nicoletta Berardi, Enrica Strettoi, and Maria Cristina for critical reading of the manuscript. 

Footnotes

 Supported by the intramural Research Grant “Linea D1” (BF), and the PNR-CNR Aging Program 2012-201 to the IN CNR (LGR).

Footnotes

⁷  These authors contributed equally to this work presented here and therefore should be regarded as equivalent authors.

Footnotes

 Disclosure: B. Falsini, None; L. Galli-Resta, None; A. Fadda, None; L. Ziccardi, None; M. Piccardi, None; G. Iarossi, None; G. Resta, None