Improving Detection of Mild Loss of Retinal Light Increment Sensitivity at the Posterior Pole With the Microperimeter MP1

Abstract

Purpose.: With “standard” stimuli (white, Goldmann size III, 200 ms), the Nidek Microperimeter MP1 underestimates retinal light increment sensitivity (LIS). We thoroughly analyze this problem, suggest alternative settings to improve sensitivity to detect dysfunction, and provide true normal values.

Methods.: LIS was tested at 55 positions in the macular region using a 4-2-1 staircase strategy with 200 ms white or red stimuli on a 1.3 cd m⁻² background. Stimulus size was Goldmann III and I, and additionally II in the healthy subjects. All participants underwent a complete ophthalmologic examination, spectral domain optical coherence tomography (OCT), and fundus autofluorescence (FAF).

Results.: In normals, distributions of LIS for white Goldmann sizes II and III within the central 6° to 10° were clipped off at 20 dB—the MP1 cannot attenuate them any further. When the stimulus size was reduced to Goldmann I or the color changed to red, median LIS in the fovea (∼15 dB) was approximately 5 dB higher than at 10° eccentricity. Estimated from these results, central LIS for white Goldmann sizes II and III stimuli were 21 and 27 dB, respectively. In four patients with either focal or diffuse macular pathology, as confirmed by funduscopy, OCT, or FAF, reduced LIS was detected clearly with Goldmann size I stimuli, but not III.

Conclusions.: In all subjects reported here, standard central LIS was above the technical limit of the MP1. To measure true thresholds in healthy subjects, either smaller (Goldmann size I) or dimmer stimuli (red) must be used.

Introduction

Conventional static perimetry, that is measurement of light increment sensitivity (LIS) at different retinal positions, can be confounded seriously by involuntary shifts of fixation of the subject. Detection of even small scotomata has been improved greatly with the advent of positional control via infrared fundus observation. Instruments developed for this purpose were termed “fundus perimeter” or “microperimeter.” ^1–5 In the Microperimeter MP1 (Nidek Technology, Padova, Italy), the software localizes a reference area in an infrared fundus image taken in real-time (eye-tracking). Any shift (but not rotation) is compensated by correspondingly moving the stimulus on the liquid crystal display (LCD). If the fundus pattern cannot be matched, the threshold procedure pauses. 

Basing on the results of a recent large multicenter trial, Midena et al. recommended standard settings for the MP1, which are used in most of the publications and trials. ⁶ The present report confirmed and investigated earlier notions ^7–11 that these settings underestimate true LIS because the MP1 is limited to light increments from 0 to 20 dB (also referred to as bottom and ceiling effect). Several lines of evidence are presented: analysis of “threshold” distribution, measurements under different stimulus conditions, and analysis in the context of psychophysical literature and other publications. The results from four patients whose retinal dysfunction would have been overlooked or underscored with the settings recommended by Midena et al. ⁶ underline the practical importance of the diagnostic gap when using “standard” parameters. 

Sensitivity of the MP1 is improved when using either smaller white or large red stimuli. The psychophysical foundation of this choice is illustrated, and means to estimate true thresholds for different stimulus settings are presented. When used with appropriate stimulus parameters, the MP1 (or other microperimeters) is a valuable tool to match function and morphology and allows measurement of retinal LIS comparable to the literature on central visual fields. 

Methods

We included 22 healthy Caucasians (14 female/8 male) aged 16 to 49 years (median 24 years) in the study. Ethical approval was obtained from the local institutional ethics committee. All subjects and patients gave written informed consent prior to their participation. The study was conducted in accordance with the tenets of the Declaration of Helsinki. 

No subject reported any history of ocular or systemic disease. All subjects underwent a full comprehensive eye examination with refraction (Oculus/Nidek AR-310A; Nidek Technology), best-corrected visual acuity determination by Early Treatment Diabetic Retinopathy Study (ETDRS) charts and Landolt C. Refractive errors were within ±6 diopters (D), astigmatism ≤2.0 D, and best-corrected visual acuity ≥20/20. Slit-lamp anterior segment evaluation, and 90 D lens fundus examination focusing specifically on the optic disc and macula were normal for all subjects. All subjects underwent spectral-domain optical coherence-tomography (SD-OCT) and fundus autofluorescence imaging (FAF; wavelength 488 nm) with the Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Germany), or fundus photography and FAF with the Optos 200Tx (FAF wavelength 532 nm; Optos, Dunfermline, Scotland). 

Two further subjects originally perceived to be suitable as healthy subjects had to be excluded from the normal group because of either visible discrete pathology, or changes obvious with OCT and FAF. Additionally, two patients with complaints consistent with some macular pathology underwent further testing with the Oculus Twinfield perimeter (Oculus, Wetzlar, Germany), color vision Lanthony Desaturated Panel D-15 (Luneau Opthtalmologie, Paris, France), electroretinography (ERG; Nicolet Spirit, Madison, WI), multifocal electroretinography (mfERG, Veris; EDI, Redwood City, CA), and electrooculography (EOG, Espion e2; Diagnosys LLC, Lowell, MA), the latter three according to the standards of the International Society for Electrophysiology in Vision (ISCEV). ¹² Together, these four patients formed a separately considered case series (see also Figs. 5, 6). 

The MP1 (Nidek Technology) images the fundus in real time with an infrared (IR) fundus camera (768 by 576 pixels resolution; 45° field of view) similar to a photographic fundus camera. It allows the examiner to view the current picture on a computer monitor. Fixation target and stimuli appear on an LCD within the MP1 and are projected inversely onto the retina. The software allows stimuli to be either white (composition of red, green, and blue) or red. 

For all conditions, background luminance according to the manufacturer was 1.27 cd m⁻², which would be equivalent, for example, to the Octopus 101 perimeter, but significantly lower than the 10 cd m⁻² in the Humphrey Field Analyzer (HFA). Test targets were attenuated in steps of 1 dB (0.1 log) from maximum brightness (127 cd m⁻², defined as 0 dB for white) up to 20 dB, that is the background luminance. Differences in the absolute definition of 0 dB, in the threshold procedure, and possibly in background intensity require consideration when comparing results with other perimeters. The MP1 also incorporates an automated tracking system to compensate for eye movement during examination. Immediately before examination, an infrared reference image of the fundus was captured to define an area of high contrast (e.g., large vessels, disc margin, or pigmented lesions) to be used for tracking. This reference landmark was tracked every 40 ms (25 Hz) to allow correction of the stimulus' horizontal and vertical position (but not torsion) on the internal LCD to maintain the same test locations on the fundus. 

In our study, stimuli were presented at 55 positions of a customized pattern covering the posterior pole of the retina up to 10° eccentricity (Fig. 1). The pupil was undilated. The examination was performed in a darkened room after approximately 10 minutes of adaptation and only on the dominant eye of the subjects. Starting with 14 dB, LIS threshold was determined with a 4-2-1 double staircase strategy and the last intensity seen reported. Results from the left eye were mirrored along the vertical axis to be comparable to those from the right eye. Subjects were asked to fixate a custom-made red X 10° in diameter (Fig. 1). Gaps within the X allowed us to project stimuli in the center of fixation or at pericentral positions without interfering with the fixation target. Initially, fixation stability derived from tracking was continuously recorded for 5 seconds. Thereafter, fixation was recorded during stimulus presentation and mapped. The diameter of the stimuli ranged from Goldmann I (6.45 min arc) to III (25.7 min arc). Due to the resolution limits of the LCD, the Goldmann II (Goldmann size II stimulus; abbreviated denotation used to improve readability) was, in fact, a 2-by-2 square matrix (12.9 min arc) and Goldmann I was a single dot. 

Figure 1

 

Customized stimulation pattern covering 55 positions on the posterior pole of the retina: the fovea, 4 parafoveal positions, and rings of ≥8 spots each at 3, 4, 6, 8, and 10° radius. The customized red fixation cross (black) is interrupted to allow stimulus presentation without interference.

Results

Healthy Subjects

With “standard” settings (i.e., 200 ms white, Goldmann III), LIS obtained with the MP1 were 20 dB for any position except four in the upper periphery, resulting in a plateau (Fig. 2A). This value is the minimal light increment the MP1 can present. Had retinal LIS been higher, the threshold procedure would still have yielded the same result. 

Figure 2

 

Median retinal LIS of the posterior pole in 22 normal subjects for white 200 ms stimuli of Goldmann size III (A), II (B), and I (C) presented on a white background of 1.27 cd m⁻². Left eyes were mirrored along the vertical meridian. Only values ≤19 dB can be considered to be within the measurement range. Hence, large areas with “20” in (A) and (B) result from technical limitations (see Fig. 3 A). In (C), higher LIS in the lower versus higher retina is apparent, but no significant difference of temporal to nasal side.

Figure 3

 

(A) Ring medians of data in Figure 2 as a function of eccentricity. Data obtained with Goldmann size III (not shown) were exactly 20 dB for any eccentricity due to physical limitations of the MP1. Equally, distributions for Goldmann size II stimuli (black boxes) still are capped up to the median (horizontal line) for eccentricities up to 6°. Only those distributions at 8° and 10° eccentricity, and those for Goldmann size II (white boxes) are within the measurement range and shift to lower LIS with eccentricity. At 8° and 10° eccentricity, median LIS for Goldmann size I and II differs by 7 dB as predicted by complete spatial summation (Ricco's Law, see text). (B) Repetition of experiment in (A) with red Goldmann size III stimuli. Small boxes in the center contain 25%, large boxes 50%, and whiskers 97% of the individual ring medians.

In a second experiment (Fig. 2C), we reduced the visibility of the stimuli by making them smaller (Goldmann I, 16 times smaller area). In Goldmann kinetic perimetry, for comparison, this causes the isopters to become smaller. In the MP1, the plateau changed to a hill with the LIS in the center being approximately 6 dB higher than the LIS at 10° eccentricity. Furthermore, reducing the size had less effect on the central LIS (5 dB lower) than in the periphery (11 dB lower). We ran an additional experiment using Goldmann II (Fig. 2B). Again, LIS in the center was higher than in the periphery, but formed a plateau of 20 dB in the central third of the field. 

The problem becomes clearer in Figure 3A when analyzing the distribution of ring medians. As the isopters of sensitivity were nigh on concentric, median sensitivities for similar distances to the fovea were calculated for each subject and distributions of these medians displayed as box plots. The number of retinal positions in each group beyond the first and second ring was almost equal. For Goldmann III, all subjects reached exactly 20 dB for all ring medians (data not plotted). For Goldmann II (black boxes), the upper parts of the distributions were cut off at 20 dB at all eccentricities up to 6°. Only with stimuli of Goldmann I were the distributions entirely within the measurement range. Median LIS was highest in the center and declined by 5 dB at 10° eccentricity. 

The results were confirmed by a second line of experiments (Fig. 3B). Stimuli were made less perceptible not by reducing their size, but by reducing their spectral emission. In contrast to “white” stimuli, red stimuli in the MP1 lack the green and blue emission spectrum of the LCD. Figure 3B demonstrates the distribution of ring median LIS for red stimuli size Goldmann III. Effectively, this too reduces the number of absorbed photons in a given area and thresholds fell within the measurement range of the MP1. This is apparent when comparing the LIS for red Goldmann III (Fig. 3B) with “white” Goldmann I stimuli (Fig. 3A), which are just slightly lower. Thus, for a given stimulus size, red is approximately 1/10th of the brightness of white and yields approximately 11 dB lower LIS. 

The use of red stimuli on a white background, however, exposed a software problem: when red stimuli reached values of 17 dB or more, stimulus presentation appeared darker on the display compared to the surroundings, that is the stimulus was not added correctly to the background illumination. As such, the stimulus was not identified as being brighter, but darker than the surroundings. This inverse paradigm spoils the threshold algorithm, since the contrast became more obvious the higher the dB value or damping was set. Until this issue is solved, such a setting cannot be recommended for clinical use. 

Clinical Case Series

The four cases presented below demonstrated that using white Goldmann III as a standard in clinical practice unnecessarily reduces sensitivity in the identification of retinal dysfunction, and that the problem illustrated above is highly relevant in clinical practice. In all cases, perimetry with the MP1 was performed with Goldmann III and I in close succession. 

Case 1 refers to a 49-year-old woman with no visual complaints, unremarkable ophthalmologic history, and normal visual acuity of 20/20. She was originally recruited to participate in the study as a healthy subject. Sensitivity was 20 dB for Goldmann III at the entire posterior pole. Results with Goldmann I are illustrated in Figure 4A. At one location just nasal to the center LIS dipped to 4 dB. Figure 5A provides the differences of the patient's results to the median of the reference group, with numbers marked in red indicating a significant reduction. SD-OCT of the posterior pole at the corresponding area identified a localized elevation of the retinal pigment epithelium (Fig. 4A, inset). Furthermore, FAF was increased in the same area (Fig. 6A). Changes highlighted by SD-OCT and FAF were far more obvious than the fundus appearance, especially with undilated pupils. The diagnosis was a retinal scar following central serous chorioretinopathy. 

Figure 4

 

Correlation of functional testing and tomographic retinal imaging. Threshold sensitivity is presented according to the scale below in 1 dB steps. Normal LIS is indicated by green color according to 20 dB, decreased LIS by red color according to 0 dB. This figure shows a functional map measured on the MP1 Microperimeter with Goldmann I of 200 ms superimposed on a near-infrared ophthalmoscopy, and an optical coherence tomography slice suitable to the most representative part of the measurements. (A) Case 1. (B) Case 2. (C) Case 3. (D) Case 4.

Case 2, a 27-year-old female, was very similar to case 1 in that the history and vision were unremarkable and that testing with Goldmann III with the MP1 did not reveal any changes. As in case 1, she had been recruited as a healthy subject. With Goldmann I, LIS in the mid-peripheral field was reduced by 11 dB (Figs. 4B, 5B) in an area composed of a number of positions with reduced FAF superior to the fovea (Fig. 6B), as well as defects in the photoreceptor layer in SD-OCT (Fig. 5B). Again, these changes appear to result from a retinal scar after inflammation. 

Figure 5

 

Differential plot of the four clinical cases. (A) Case 1. (B) Case 2. (C) Case 3. (D) Case 4. Every measured position of the pattern was compared to associated positions in the compilation of the 22 healthy subjects. The resulting deviation of the data then was compared to corresponding box plots. The amount of the deviation is marked with green (within 50% around the median), yellow (within 97% around the median), and red (significantly lower than healthy subjects).

Case 3 concerned a 26-year-old female complaining of variable binocular problems of central vision without any other ophthalmologic history. Visual acuity was 20/20. Conventional perimetry was unaffected. LIS with the MP1 and Goldmann III was 20 dB at all locations of the posterior pole. For Goldmann I the macular region was less sensitive than the surrounding retina (Fig. 4C), and significantly lower by 4 to 6 dB in 7 locations in comparison with healthy subjects (Fig. 5C). Multifocal ERG revealed only slightly decreased central amplitudes, and color vision testing with the Lanthony desaturated Panel D15 contained just a few confusions; both tests were not significantly affected. The functional changes in the MP1 were not paralleled by changes of the retinal layers in the SD-OCT depicted in Figure 4C. Fundus autofluorescence was unchanged (Fig. 6C). Findings were similar in both eyes. There was no family history of eye disease, and no signs of inflammation or vascular events. The underlying disease was not identified. 

Figure 6

 

(A–D) Fundus autofluorescence of the four clinical cases, 1 to 4, respectively ([A] and [B] with Optos 200Tx; [C] and [D] with Spectralis). Case 4 (C) has normal FAF, Case 1 (A) has a single spot with minute change that corresponds to significant deviation in Figure 5A. Multifocal changes in Case 2 (B) and perimacular in Case 4 (D) are quite obvious.

Case 4 depicted a 48-year-old female referred for pigmentary changes of the macula. She had seen an ophthalmologist because of metamorphopsia and changing refraction. Furthermore, she recalled a similar event during a pregnancy. Her best corrected vision was 20/20 in the right eye and 18/20 in the left eye. Conventional automated perimetry was normal. Multifocal ERGs were slightly reduced in the central area. With the MP1 Goldmann III, LIS was subnormal at three positions only, whereas with Goldmann I the whole macular region exhibited reduction in LIS from 5 to 14 dB (Figs. 4D, 5D). The functional defects in the MP1 correlated well with the retinal changes in the SD-OCT (Fig. 4D) and in the FAF (Fig. 6D). Yellow discoloring on funduscopy and normal EOG led to the diagnosis of pseudovitelliform macular degeneration. 

All patients presented were more sensitive in the normal regions of their central retinae than with more peripheral regions similar to the normal group above. In cases 1 to 3, sensitivity losses with the MP1 were unapparent with Goldmann III but manifested when tested with Goldmann I. In case 4, LIS with Goldmann III was reduced only at those three positions for which the defect depth was approximately 13.5 dB for Goldmann I. Adjacent functional losses were only measurable with Goldmann size I stimuli. Areas of functional loss clearly demonstrated a changed morphology using FAF or OCT, except in case 3. 

Discussion

A fundus-controlled perimeter, such as the MP1, measures LIS at defined, reproducible retinal locations. This is important especially in those areas of the retina for which even a small misplacement causes a large change in threshold, as is the case with pathologies of small extent or with maculopathy. Similarly, the use of coarse grids or large stimuli may mask significant pathology. However, an additional factor important in the choice of stimulus parameters is the operating range of the device. Our work with the MP1 illustrates that the choice of stimuli and placement is crucial to the successful detection of mild retinal dysfunction at the posterior pole with this equipment. 

The influence of stimulus size on threshold is known as spatial summation, regularly observed in Goldmann kinetic perimetry in which isopters for Goldmann I or even 0 surround a smaller area than those for Goldmann III. Although a final conclusion on the physiologic background has not been drawn to date, one explanation is that information from several photoreceptors converges on single ganglion cells.^13–15 In the central retina, summation under photopic conditions is complete up to a critical area, that is, the product of light sensitivity I and area A is constant:  

 

When stimuli are increased in size, some summation can still be observed according to Piper's Law.  

 

The effect of summation is directly visible in Figure 3, comparing results for Goldmann II and I at 8° and 10° eccentricity: if the stimulus area is reduced by a factor of 4 (from 12.9 to 6.45 min arc diameter), LIS according to Ricco's Law ought to drop by the same factor, that is by 6 dB, and this is indeed the case approximately. For Goldmann III versus II, the difference at 10° eccentricity is slightly smaller (5 dB). The difference is smaller than expected because the distributions are distorted by the limited operating range of the MP1—the MP1 cannot present stimuli with contrast lower than 20 dB, and hence, this reading should be referred to correctly as “≥20 dB.” This causes the box plots to appear cut off and the median to move upwards within. Ultimately, with respect to sensitivities well above the operational range, only “20 dB” is obtained in all subjects as recorded for Goldmann III in the first experiment (not shown in Fig. 3A). Similar cutoffs were reported by Midena et al.⁶ in their Figure 5 and Table 3, but were left uncommented.  

True retinal LIS in the center for Goldmann II and III were estimated from the results with Goldmann I assuming full validity of Ricco's Law, yielding 20.5 and 26.5 dB, respectively. Consistent with this assumption, the box plot for central presentations of Goldmann II measures half the height (top half cut off) and the median lies at the upper operating limit of the MP1. Subjects perceive the testing procedure with Goldmann III as much easier than with I because stimuli 6.5 dB or 4.5 times above threshold are much more easily discerned from background noise. The problem was first briefly mentioned by Springer et al., ¹⁹ explicitly by Shpak et al., ^7,8 and by others later. ^9,11 Some reports caution that threshold reliability was higher with larger, blurry stimuli. ^20,21 According to a recent extensive study by Wallis et al., ²² though with a different two-interval forced-choice procedure, this is not due to the slope of the psychometric function. 

An issue not furthur explored by us (or others to our knowledge) is the precision of the provided sizes for Goldmann I, II, and III, and the influence of the fact that these stimuli are composed of separate, rounded rectangles (pixels; see also fourth approach). Differences will mainly cause relative shifts of threshold at all measurement positions, but some dependence on eccentricity due to changes in, for example, receptive field size and spatial summation may be expected. 

We provided a second line of evidence that the above interpretation regarding the operating range is valid and not only a theoretical consideration, but an important practical problem. Keeping the Goldmann III stimuli, we effectively dimmed them by choosing red stimuli, an option in the settings dialog, when the green and blue light on the LCD are switched off. This yields comparable results (Fig. 3B) to using smaller white stimuli (Goldmann I instead of III). Hence, we estimated the brightness of red versus white for these background conditions to be approximately 10% (11 dB), similar to an estimation by others obtained with a dimmer, red background. ²³ The attenuation appears reasonable, since the photopic sensitivity curve peaks at 555 nm, that is green light is by far more effective in terms of photopic luminance than red light. 

However, we have encountered a potentially serious software problem in that requests for red stimuli apparently are handled incorrectly for values ≥17 dB. Instead of rejecting the request, the call for an increment stimulus is converted to a perceivable decrement stimulus (red within a matrix of red-blue-green), thus inverting the threshold paradigm from “bright on background” to “dark on background.” Unless subjects are instructed specifically not to respond to “dark spots,” the algorithm would decrease the stimulus intensity further. Thus, it would increase the contrast of the decrement until it reaches “20 dB.” Hopefully, the company will provide a software patch to take care of this problem soon. 

Red stimuli had been used previously, ^10,23 but on a dimmer, red background, which enhances the influence of rod vision, but at the same time eliminating the above software dilemma. Since both stimulus and background were changed, this should have left the Weber contrast unaltered. However, this (and even the white background used in our study) only light-adapts the system to a mesopic state. The question as to whether in that state changing the relative emphasis of retinal pathways affects the detection of different forms of disease remains complex and currently unsolved. ²⁴  

A third possibility, not explored further in this study, is to shorten stimulation time. As formulated in Bloch's Law, ^25–27 temporal summation allows changes in duration to be exchangeable with changes in intensity and vice versa for stimulation times of approximately ≤100 ms. Hence, white Goldmann III would have to be shortened to approximately 20 ms to move within the upper operating range of the MP1 for normal subjects. Such values are currently not available in the software for the MP1. 

A fourth possibility to diminish visibility of the stimulus, again not explored further here, is to reduce its edge contrast with an opaque filter. At the same time, the opaque filter would even out the visible matrix of the LCD. However, if such a filter were placed in the common light path, it would also affect the contrast of the infrared retinal image used for tracking. Hence, it would be preferable to place the filter in the projection light path of the LCD, which requires modification of the MP1. 

Using any of the above changes to the stimulus parameters to “tune” the operating range, as shown for stimulus size and stimulus emission spectrum (first and second approaches, respectively), the physiologic phenomenon of the hill of vision becomes apparent, that is, LIS is highest in the center of the retina. The hill of vision is a basic psychophysical feature of the light-adapted retina. It is well known for Goldmann kinetic perimetry ^28,29 in which smaller or dimmer targets yield smaller isopters. Furthermore, the phenomenon has long been shown to be present in static perimetry normative data obtained with conventional apparatus, ^30–32 as well as with other scanning-laser ophthalmoscopes. ^3,5 Others have compared the relative measurements obtained with the MP1, for example with the Octopus 101, ^19,33 the HFA, ^9,11,34 or the Opko OCT/SLO. ¹¹ However, in all of these studies except that of Seiple et al., ¹¹ only a Goldmann III was used in the MP1, which according to the data presented here for the central retina is very problematic. Using 2-color–threshold perimetry, ^35,36 it is possible to demonstrate with the MP1 that thresholds in the normal retina are determined by cones (Friedburg C, et al. IOVS 2012;53:ARVO E-Abstract 4843). Their density ¹³ and those of the connecting cone bipolar and ganglion cells ¹⁴ are highest in the fovea and quickly decrease toward the periphery. The same stimulus placed in the fovea will thus illuminate and activate more cones and ganglion cells than a peripheral stimulus. Thus, a plateau of sensitivity in the center measured with Goldmann I, as depicted in Figure 4C for case 3 or described by Midena et al. ⁶ for Goldmann III, indicates macular dysfunction. 

Our data, thus, conclusively illustrated that normal retinal sensitivity measured with a modified test protocol on the MP1 is significantly and, in practice, relevantly higher in the macular region than reported previously by Midena et al., ⁶ Springer et al., ¹⁹ Ratra et al., ³⁴ Shah et al., ³⁷ and Sabates et al. ³⁸ A staircase procedure that is limited operationally to an intensity well above threshold produces highly consistent (as shown for Goldmann II at 1.2° eccentricity in Fig. 3A) and highly repeatable data. However, the data are distributed in a skewed fashion, finally compressing to a single value. Data may appear to be in the operating range below 20 dB when raw data are not examined closely. The same can be said for when average values are calculated for data samples that do not follow a normal distribution. For example, had we plotted the distribution of the mean individual sensitivity at given retinal eccentricities in Figure 3A, some of the box plots with a median of 20 would have appeared to be in the operational range below 20 dB. In essence, no median sensitivity of 20 dB may be trusted (since the MP1 would have had no chance to test any lower contrast). All of these problems would be prevented if the MP1 software were to mark the results of staircase procedures finishing with values of 20 dB due to the ceiling effect as “above-range.” Similarly those results with 0 dB due to the floor effect should be labeled “below-range.” This would expand the usable range of the MP1 by 2 dB. 

To summarize, optimizing stimulus parameters is essential to obtain true LIS from the MP1 without leaving its operating range. Otherwise serious dysfunction may be overlooked or remain unquantified. Though using larger stimuli may be advantageous, our second approach (Fig. 3B) cannot be recommended for a standard clinical setting until the company provides a software update. Hence, in normal subjects, a reasonable choice currently is to use Goldmann I with a fixation mark that does not interfere with the stimulus pattern. Unfortunately, we are left with the conclusion that some important, recent studies on “norm” values for retinal sensitivity ^{6,19,34,37,38} are confounded by this problem. 

Acknowledgments

Presented in part as a poster at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2012 (abstract 4843), and as a poster at the annual meeting of the German Ophthalmological Society (DOG), Berlin, Germany, September 2012 (abstract 687). 

Supported in part by a grant of the Deutsche Forschungsgemeinschaft (German Research Council DFG LO457/10-1). Fundus autofluorescence pictures in Figure 6A and 6B with financial support by Optos. The authors alone are responsible for the content and writing of this paper. 

Disclosure: W. Bowl, None; B. Lorenz, Optos (F); M. Jäger, None; C. Friedburg, None