Identification of Usher Syndrome Subtypes by ERG Implicit Time

Abstract

purpose. Usher syndrome (US) is a recessively inherited disorder combining retinitis pigmentosa (RP) and a sensorineural hearing loss. The classification in subtypes is based mainly on auditory tests. The purpose of this study was to analyze implicit time (IT) differences in the electroretinogram (ERG) between RP alone, US I, and US II.

methods. The data of 15 control subjects and of 15 patients with US I, 15 with US II, and 15 with RP with nonzero 33-Hz flicker ERG responses were analyzed. The ITs of three signal peaks (P1–P3) were evaluated. Sensitivity and specificity of a test to distinguish between US I and II based on timing differences were determined. Multifocal (mf)ERGs were used to assess differences in disease topography.

results. Despite the similar amplitude loss with retinal eccentricity in the mfERG in all three groups, the peak delay in US I was negligible compared with that in US II and RP. In the flicker ERG data, US I and control subjects had almost identical peak times, and the same was true for subjects with US II and RP. Because of the slight overlap between US I and II, the diagnostic test achieved a sensitivity of 100% and a specificity of 93.3%.

conclusions. Substantial timing differences between US I and II and their usefulness for a diagnostic test were demonstrated. This finding may also be the basis for further investigations regarding the structural differences of retinal impairment between US I and II on a cellular level.

Usher syndrome (US) is characterized by a combination of progressive visual loss of the type that occurs in retinitis pigmentosa (RP) and a stationary, congenital sensorineural hearing loss (SNHL), which is moderate in type II and profound in type I. ^{1 2} Typically, patients with type I also show an absent vestibular response during caloric testing. In addition to the two classic clinical subtypes, there is a rare third type featuring progressive hearing loss. ³  

US is a both phenotypically and genotypically heterogeneous group of diseases with autosomal recessive heredity. ⁴ Genetically, there are five known loci for type I (USH1A-E), ^{5 6 7 8 9 10} one for type II (USH2A), ¹¹ and one for type III (USH3A). ³ Responsible genes known so far are a defective myosin VIIA gene that has been linked to US type IB, ⁴ a gene for usherin ¹² in type IIA, and one for otocadherin in type ID. ¹³  

The pathophysiological processes leading to US are still not resolved. The presence of myosin at the base of photoreceptor-connecting cilia has led to the hypothesis that US type IB may be the result of a primary cytoskeletal defect. Indeed, it has been found that myosin VIIA is involved in the transport of rhodopsin between inner and outer photoreceptor segments. ¹⁴ It has therefore been speculated that the cilia of cochlear and vestibular hair cells are the primary target structure in the inner ear and that the nasal ciliary cells may also be defective. However, it remains unclear how a similar ciliary defect may cause a stationary SNHL (with the exception of type III), a progressive tapetoretinal degeneration, and no quantitative and qualitative olfactory loss. ¹⁵ The functions of usherin and otocadherin have yet to be identified. 

Several studies have shown that there are rather minor clinical differences between the ocular manifestations of US I and II, ^{1 15} and only a few features, such as age of onset of RP and secondary cataract formation, reach borderline statistical significance. Because these have a low discriminative power, it is obvious why the diagnostic separation between US types I and II has been based so far mainly on otologic test results. ¹⁶  

Electroretinography (ERG) is a very sensitive method for the detection of disorders of the RP type ^{17 18} in general and US ^{1 16 17 19} in particular, even in the earlier stages. However, because the ERG responses decrease during the course of the degeneration, discrimination between different forms of RP becomes difficult in the advanced stages. Strategies to optimize the diagnostic value of the ERG in such cases include the use of strong, repetitive stimuli in the Ganzfeld ERG (to maximize response amplitude) and the multifocal (mf)ERG, which provides a topographical distribution of the electrical activity across the retina. ²⁰ In RP and US II, the latter method has been shown to reveal a typical pattern of amplitude loss ^{21 22 23 24} and implicit time (IT) delay ^{20 25 26} from the center to the periphery. 

After initial evidence that mfERGs in US I did not show this typical delay pattern, the purpose of this study was to analyze IT differences between US I, US II, and RP alone. 

Patients and Methods

Patients

The patients’ data in this retrospective study were taken from files recorded between 1995 and 1997 (more recent records were not used to avoid bias due to changed equipment). For the evaluation of the 33-Hz narrow-band flicker described in the Ganzfeld ERG section that follows, the ERGs of 15 patients with US I, 15 with US II, and 15 with RP with informative responses were analyzed and compared with those of 15 normal control subjects. The results of the better eye were used. The diagnosis of US had been made based on results of otologic and ophthalmic examinations. In the US I group, the age median was 25 years (8.1–43.4; data in parentheses represent the range between the 5% and 95% quantiles), the visual acuity median 0.8 (0.1–1.0), and the visual field radius median using a Goldmann III4e stimulus was 11.4° (4.1–45). Auditory loss was quantified as previously described. ¹⁵ In all patients with US I, both pure tone and speech-recognition audiography yielded a loss of 100%. The age median in the US II group was 31 years (16–51.3), the visual acuity median 0.6 (0.3–1.0), and the visual field radius median 9.6° (4.3–37). Pure-tone audiography showed a loss of 45% (4.5–100) and speech recognition a loss of 60% (9–85.5). Finally, the age median in the RP group was 25.2 years (14.5–47.9), the visual acuity median 0.8 (0.29–1.2), and the visual field radius median 10° (3.7–29). Visual fields were input into a computer system and the remaining area calculated as a spatial angle. ²⁷ The age median in the control group was 30 years (21.7–60.1). Results of the genetic analysis were available in three of the patients with US I (two type IB, one type ID). 

As many of the patients tested with a narrow-band flicker did not have an mfERG in their records, a different but overlapping set of patients with US I, US II, or RP was obtained from the files based on the quality of their mfERGs to attain a satisfactory number. A patient’s records were included if a clear response at least in the most central hexagon was detectable. This set of 8 patients with US I, 11 with US II, and 22 with RP was used to calculate ring statistics and group average traces as described in the mfERG section that follows. The research followed the tenets of the Declaration of Helsinki. 

Ganzfeld ERG

ERG responses were obtained with a Ganzfield setup (Spirit; Nicolet, Madison, WI) according to the International Society for Clinical Electrophysiology of Vision’s (ISCEV) standard protocol. ²⁸ Specifically, patients underwent a dark adaptation of 30 minutes before scotopic recordings, and photopic recordings were preceded by a light adaptation of 10 minutes at an illumination of 30 candelas (cd)/m². An additional 33-Hz microflicker was used at the end of the photopic session, which is an implementation of narrow-band filtering ²⁹ on standard ERG equipment. The protocol differs from the ISCEV standard in frequency and filter settings (33-Hz frequency, band-pass 30–100 Hz, notch on, 50 averages, time base 75 msec, background 10 foot Lambert (ftL), flash intensity 1.0 log). 

Patients tested with another system (UTAS2000; LKC Technologies, Gaithersburg, MD), which provides an 32-Hz microflicker showed similar results but were not used in this study, to avoid potential problems during analysis. 

Multifocal ERG

mfERGs were recorded on a visual evoked response imaging system (VERIS) system implementing a multi-input stimulation technique first introduced by Sutter and Tran. ³⁰ Measurements were performed approximately 15 minutes after the end of the photopic Ganzfeld session, and the same Dawson-Trick-Litzkow (DTL) ³¹ electrodes were used to record the evoked field potentials from the cornea. The setup has been described in detail previously. ¹³ In short, the stimulating 61 hexagonal elements were presented on a 20-in. monitor (75-Hz frame rate, 100-cd/m² for white, mean luminance 51.8-cd/m²; Sony, Tokyo, Japan), projecting to a visual field area of 30° radius. Recording at 16 samples per frame yielded a temporal resolution of 0.83 msec. Signals were subsequently amplified (×200,000) and filtered (10–100 Hz; model 12; Grass Quincy, MA). 

Ring averages were calculated by averaging responses from five ring-shaped areas of the same eccentricity (see Fig. 1 ). Group average traces were obtained with the “combination” feature of the VERIS software, which combines raw data from a group of patients in one record that can be evaluated the same as a single patient record. 

Statistical Work-up

The microflicker waveforms featured three size-variant peaks (P1–P3), for which ITs were separately evaluated. Because of the repetitive nature of the flicker stimulus, all peaks appeared periodically within a trace (at least twice with the chosen settings). To reduce noise influence, the values for P1–P3 from the first occurrence in the ERG trace were averaged with those of the second occurrence after subtraction of the time interval between two flashes of the 33-Hz flicker (30.3 msec). 

Differences between groups were analyzed using the nonparametric Mann-Whitney test and were considered significant at P < 0.01. If the probability was between 0.05 and 0.01, the respective difference was classified as borderline significant. 

A Clinical Test

The sum score S was calculated as the sum of the ITs of all three consecutive peaks: P1, P2, and P3 (each previously averaged as described earlier). A test using S to separate US I and II could then be performed as follows: If S in a given patient is below a threshold t (in milliseconds), the patient is classified as having US I, and if S is above t, as having US II. Sensitivity (Se) and specificity (Sp) were determined by  

\[Se{=}\ \frac{\mathrm{no.\ of\ correctly\ classified\ US\ I\ patients}}{\mathrm{no.\ of\ all\ US\ I\ patients}}\]

 

\[Sp{=}\ \frac{\mathrm{no.\ of\ correctly\ classified\ US\ II\ patients}}{\mathrm{no.\ of\ all\ US\ II\ patients}}\]

The test performance can then be optimized by variation of the cutoff threshold t, because this value influences the number of correctly classified patients. Usually, maximally high values of both sensitivity and specificity are desired. 

An important measure for the practical relevance of a test is the positive predictive value (Pv+). It is an estimate of the probability to make a correct diagnosis (i.e., a Pv+ of 0.9 means that, on average, disease in 9 of 10 patients is correctly diagnosed with a given test). In contrast to sensitivity and specificity, it is also influenced by the prevalence. It is calculated by  

\[Pv\mathrm{{+}}{=}\ \frac{\mathrm{no.\ of\ correctly\ classified\ US\ I\ patients}}{\mathrm{no.\ of\ all\ patients\ with\ a\ positive\ test\ result}}\]

The prevalence refers in this case to the fraction of the patients with US I within all patients with US examined at a given institution. In a recent prospective study in our clinic, ¹⁴ this fraction was 8 of 39 (approximately 20%). 

Results

Multifocal ERG

The mfERG in US I and II showed an RP-like amplitude loss with retinal eccentricity, but no or only a slight IT delay in US I compared with US II and RP. The trace arrays in the left column of Figure 1 illustrate a similar amplitude topography in US I, US II, and RP. However, the normalized ring averages of typical patients as well as group averages (center and right column, respectively) illustrate the IT differences between US I and both US II and RP. In the latter two, there was a growing delay with eccentricity of 10 msec and more in the peripheral ring (Figs. 1C 1D) , whereas the US I records did not show any or only a minor delay, even in heavily affected regions with small ERGs close to noise level (Fig. 1B , center and right). This difference was also apparent from an analysis of mfERG ITs by eccentricity (Fig. 2) . ITs were identical within the central two rings for all groups. In ring 3, the records of some patients with RP or patients with US II began to exhibit delays, whereas those of patients with US I remained unchanged. In the most peripheral rings 4 and 5, RP and US II ITs were markedly delayed, but the US I traces had no or minor delays, which were below 5 msec in all cases observed in the study. 

Microflicker Ganzfeld ERG

The IT delay detected in the mfERG was also found in 33-Hz Ganzfeld ERG. This was not only a confirmation of the mfERG results, but also was important for the design of a diagnostic test, because Ganzfeld ERGs are available to many more ophthalmologists. However, because of the early onset and rapid course of retinal changes, the fraction of patients with US seen in our clinic (particularly US I) who had residual responses that permitted determination of IT was relatively small. The 33-Hz microflicker ERG was used in this study because, as a result of the strong, repetitive stimulation, it had the highest probability of detecting response peaks that could be evaluated for IT. It was found that the 33-Hz waveform featured three peaks of variable size that were present over a wide range of amplitudes (Fig. 3 , top two traces). Because it was outside the scope of this study to determine the origin of these components, they were neutrally labeled in the order of their appearance P1, P2, and P3 (Fig. 3) . The time lag (or phase shift) in the patients with US II or RP (bottom two traces) relative to the patient with US I and the control subject (top two traces) is clearly visible (Fig. 3 , arrows in bottom trace). The bottom two traces were also chosen to illustrate potential limitations that were not specific for any type of disease. First, especially P1 was sometimes visible as a shoulder only (Fig. 3 , third trace from the top), and second, if the signal became very small, the increased influence of noise induced extra variability (Fig. 3 , bottom trace). In the first case, it may be helpful to put a template from a control subject on top of the patient’s record to determine the order of peaks. In the second case, it turned out that the averaging process equalized most of the variability if repetitive peaks were discernible at all. 

Statistical Work-up

The comparison of IT data, averaged from two cycles of the 33-Hz flicker ERG, between control subjects and patients with US I, US II, or RP is shown in Figure 4 . For each peak, a clear difference between the IT level of the control group and the US I group and between that of the U II group and RP group was apparent. A detailed statistical evaluation (Table 1) confirms this finding. It also shows that some differences between control subjects and patients with US I and between patients with US II and those with RP are borderline significant. In the first case, this suggests that peak P3 and possibly P2 were somewhat delayed in US I. In the second case of only one borderline-significant difference for P2 and none for P1 and P3, it remains uncertain whether the slight difference between US II and RP data is real or not. To exclude two potential sources of bias, group differences in age and visual field area were analyzed (Table 2) . It turned out that although the US I group had a lower age median than, for example, the US II group, none of these differences was even borderline significant and can thus not explain the large IT difference found. Similarly, there were no significant differences in visual field area between the disease groups. 

Clinical Test

The clear IT difference between US I and II made it appear possible to separate these patients diagnostically on the basis of 33-Hz flicker ERG IT. A test was designed using a score (S) arrived at by summing the three peak ITs. The distribution of S is shown in Figure 5 (top) in the form of a histogram and a probability density function calculated from the mean ± SD of each group. The latter gives a good overview but is based on the (unproven) assumption of a normal distribution. The low degree of overlap between the distribution of S of US I and II (Fig. 5 , top) explains why the proposed test achieved a specificity of 100% and a sensitivity of 93.3% for the optimal cutoff threshold setting at 72 msec (Fig. 5 , bottom). This means, if the 15 patients with US I and the 15 with US II in this study had been classified only on the basis of this test, 14 would have been correctly recognized as having US I, whereas one patient with US I and the 15 patients with correctly diagnosed US II would have been classified as having US II. The positive predictive value (Pv+) in the case of 100% specificity is also 100% (i.e., if the test result diagnosed US I, it was always true), so that the prevalence of US I in the clinic population of US patients (estimated 20% in our clinic) may not have a major influence. 

Discussion

ERG IT is an important feature in the differential diagnosis of many hereditary and acquired retinal disorders. In this study, it was shown that US I and II are very different in this regard. First, the topographical distribution of changes was studied with the mfERG. Whereas patients with RP, US I, or US II all displayed typical loss of amplitude with eccentricity, only the US I group had normal to moderately delayed ITs in peripheral areas. In contrast, patients with US II had a marked delay similar to that of patients with RP. This effect was unexpectedly strong—so much so that the difference between US I and II was not only statistically significant but also useful for the design of a clinical test with high sensitivity and specificity. The problem of strongly reduced signals even relatively early during the course of the disease (probably a reason that this effect has not been described yet) could be overcome in many cases by the use of a special stimulation mode available on the equipment of at least two major manufacturers. Thus, the proposed test may be useful for many clinicians without the need for special equipment or complicated methods of data evaluation. 

Another striking finding is that the US I group, despite the numerous genetic subtypes, was very homogeneous with regard to implicit time. Although unlikely because of the homogeneous US I distribution, genetic US I subgroups may differ from one another in IT. Three of the 15 patients have so far been genetically classified (two subtype IB, one subtype ID), but there was no perceivable difference in timing. Possible explanations are that there is a common defect resulting from all these mutations or that one or more similar mutations are exceedingly frequent. A recent report of similar ERG amplitudes in patients with and without myosin VIIA mutations in US I unfortunately does not provide ITs. ³²  

A review of IT data may also disclose whether there is a perfect match of the otologic- and the IT-based decisions. It may be that some patients with US II (by implicit time) were classified as type I due to complete deafness, whereas some patients with US I (by implicit time) may not be completely deaf. 

The present study may also be helpful in further work undertaken to answer the long-standing question of why cone ERGs are delayed in some disorders. It appears very probable from this result that there is a principal structural difference between US I and II on a cellular level. In particular, it could be speculated that the site of action in US I is at a relatively early step in the visual pathway. For example, if cell contacts were specifically impaired in US I but not in US II and RP (as may be deduced from a defective cadherin in type I D ¹³ ), this might lead to changes in IT due to changed electrical spread in the photoreceptor network. ^{33 34}  

In summary, in the present study there was no substantial IT delay in US I compared with US II and RP. It was further shown that this IT difference could be used to diagnostically separate US I from US II. In addition, this finding may be the basis for further investigations regarding the structural differences of retinal impairment between US I and II on a cellular level. 

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2000, and the 38th meeting of the International Society for Clinical Electrophysiology of Vision, Sydney, Australia.

 Supported by Deutsche Forsgemeinschaft Grants SFB 430 C2, Zr 1/10, Se837/1, and fortuene Grant 517.

 Submitted for publication July 6, 2000; revised January 29 and July 20, 2001; accepted August 1, 2001.

 Commercial relationships policy: N.

 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 1.

 

mfERG results in patients with US. Left: topographical distribution of ERG activity (trace array) within a visual field of 30° radius; center and right: ring averages, starting with the central response alone (trace 1), the average of the six surrounding hexagons (trace 2), and so on (see sketch in top right corner). Trace 5 is the average of the 24 most peripheral hexagons. All ring averages were normalized (i.e., they have the same size, regardless of their actual amplitude) for better visibility of ITs. (A) Normal control; (B, left and center) patient with US I. The RP-like topographical distribution of ERG signals corresponded to the tubular vision. However, there was no IT delay in the periphery. Right: Combined data of patients with US I. (C, D, left and center) Patients with US II and RP, respectively. Despite a similar topographical distribution, there was a substantial IT delay of approximately 10 msec in the peripheral regions of the 30° area. Right: Combined data from patients with US II and those with RP.

Figure 2.

 

mfERG IT analysis by eccentricity groups (rings). Shown are median (filled symbols), 5% (bottom bar), and 95% quantile (top bar) of individual patient ITs grouped by eccentricity (rings 1–5). Even in the periphery, patients with US I showed normal to slightly delayed (5 msec) ITs, whereas the average delays in US II and RP were approximately 10 msec.

Figure 3.

 

Ganzfeld 33-Hz flicker waveforms in patients with US. Note the difference in amplitude between records. Traces from top to bottom, respectively: waveform of a normal subject, waveform of an patient with US I, waveform of a patient with US II, and waveform of a patient with RP.

Figure 4.

 

Evaluation of peak ITs. Peak ITs of the three patient groups and the normal control subjects plotted for P1, P2, and P3 separately. The symbols in each left column identify individual patient data, and the symbol in each right column indicates the group median and the whiskers the 5% and 95% quantiles.

Table 1.

 

Statistical Evaluation of Group Differences in Implicit Time

	U I	U II	RP
Peak 1
Control	0.81	0.000012	0.0000023
U I	—	0.0000046	0.0000036
U II	—	—	0.34
Peak 2
Control	0.031	0.000014	0.0000012
U I	—	0.000016	0.0000080
U II	—	—	0.047
Peak 3
Control	0.00077	0.0000046	0.0000021
U I	—	0.000017	0.0000027
U II	—	—	0.12

 The group differences were evaluated for each peak separately. Shown are the probabilities obtained with the nonparametric Mann-Whitney test. P 0.05 to 0.01 was defined as borderline significant, and P < 0.01 as significant.

Table 2.

 

Statistical Evaluation of Potential Sources of Bias

	U I	U II	RP
Age
Control	0.15	0.71	0.37
U I	—	0.093	0.33
U II	—	—	0.46
Visual field
Control	—	—	—
U I	—	0.26	0.075
U II	—	—	0.53

 To exclude potential sources of bias, the group differences regarding age and visual field area were also evaluated. Both factors did not explain the IT differences found. See Table 1 for a description of probabilities.

Figure 5.

 

Performance of the proposed diagnostic test. Top: the sum score values from the US I (gray) and II (black) groups. For comparison, probability density functions of both groups are overlaid. There was a low degree of overlap between groups. The bottom plot illustrates the changes of sensitivity and specificity with the cutoff threshold of the test. The optimum cutoff in these data was at S = 72 msec (indicated by the dashed line).