Intrasession Test–Retest Variability of Microperimetry in Age-Related Macular Degeneration

Abstract

Purpose.: To determine the intrasession test–retest variability of microperimetry in participants with age-related macular degeneration (AMD).

Methods.: This study consisted of two separate groups of subjects who had not performed microperimetry previously. In group 1, 30 AMD and 14 control participants performed three microperimetry examinations of a selected eye within one session (test 1 and 2, first pair; test 2 and 3, second pair). Follow-up examination at 6 months was available in 20 AMD participants in group 1, who performed two microperimetry examinations. In group 2, 71 AMD participants performed a short practice examination, then two microperimetry examinations of the right eye (test 1 and 2, first pair) and two of the left eye (test 3 and 4, second pair).

Results.: There was a significant improvement in average point-wise sensitivity (PWS) between the first pair of examination in both groups (P < 0.001), but not in the subsequent pair (P ≥ 0.774). This improvement was not observed at the follow-up visit in the subset of AMD participants in group 1 (P = 0.433). The PWS coefficient of repeatability (CoR) for the second pair of examinations was ±4.12 dB and ±4.37 dB for AMD participants for group 1 and 2 respectively.

Conclusions.: A significant increase in sensitivity between the first and second test, but not in the subsequent tests, was found for participants who had not performed microperimetry previously. Intrasession test–retest variability can therefore be minimized by discarding the first examination to avoid the influence of a learning effect.

Introduction

A major impediment to the initiation of novel interventional studies for the early stages of AMD is the lack of a specific, sensitive, and clinically applicable functional marker for disease severity and progression. Among the outcome measures used in clinical trials for AMD, best-corrected visual acuity (BCVA) remains the most widely used functional measure. ^1,2 However, in the earliest stages of AMD, BCVA often underestimates the disease extent and is a poor measure of progression, ³ and therefore new outcome measures are urgently needed to take advantage of emerging therapies to intervene early in the disease process. ⁴  

Retinal sensitivity determined on fundus-tracked perimetry (commonly referred to as “microperimetry”) has been recently used as such a measure, and has been shown to be effective at detecting functional deficits in the early stages of AMD. ^5–7 These functional deficits have been found to correspond with changes detected on new imaging modalities, including increased fundus autofluorescence (FAF), ⁸ thinning of the outer segment (OS) thickness, thickening and elevation of the RPE, ⁹ and disruption of the second hyperreflective band on spectral-domain optical coherence tomography (SD-OCT). ¹⁰ The real-time compensation of eye movements achieved through visualization of the fundus with microperimetry has allowed highly precise measurements of retinal sensitivity at individual retinal regions sampled. This feature is crucial when considering a measure of progression, as pathological changes in AMD frequently occur in localized regions. ¹¹  

However, as with all subjective measures of retinal function (especially in the setting of disease and especially in older individuals), measured values vary around the true value, and this variability can be determined by performing repeated measurements on the same subject. This test–retest variability is influenced by factors inherent in psychophysical testing, including natural fluctuations in perception and familiarity with the examination. 

In inexperienced subjects, an improvement in measured values is frequently observed on repeated examination, a phenomenon that is typically described as a learning effect. This systematic improvement produces artifactual defects on initial examination, and inevitably increases the test–retest variability in addition to the natural fluctuations of subjective responses. Although a learning effect is well-reported when using standard automated perimetry, ¹² all studies to date using microperimetry have reported that a learning effect is not present in subjects with and without retinal disease. ^13–16 These studies to date, however, have considered only changes in mean sensitivity (MS), which have not considered perimetric data as measurements at multiple locations on the same patient. As a result, learning effects that may be present can be lost through the process of averaging. ¹⁷  

Additionally, it is also not known whether test–retest variability decreases during a single session involving multiple examinations, even if there is no significant learning effect. Given that both eyes are typically tested once during a single visit in clinical trial and practice settings, this information is crucial to understand the influence of testing sequence, in order to determine whether the same estimates of test–retest variability can be applied to both eyes. This study aimed to investigate the intrasession test–retest variability when testing either one eye or both eyes in patients with AMD. 

Methods

This study was approved by the Human Ethics Committee of the Royal Victorian Eye and Ear Hospital and adhered to the Declaration of Helsinki. Written informed consent was obtained from all participants following an explanation of all tests involved. In this study, participants were recruited from the Macular Research Unit from the Centre for Eye Research Australia, where they were already involved in research projects; in group 1, participants were involved in a natural history study of AMD, and in group 2, participants were examined for eligibility for an early interventional study for AMD. These two groups were included in this study because participants in group 1 had only one eye tested, whereas participants in group 2 had both eyes tested. This allowed intrasession test–retest variability to be examined under both monocular and binocular testing conditions (details of the testing protocol for each group are described in the microperimetric examination section). 

Participants

Participants with BCVA of 20/60 or better and an age of 50 years or older were recruited for this study. The inclusion criteria for AMD participants were drusen greater than or equal to 125 μm or multiple drusen greater than or equal to 63 μm with or without pigmentary anomalies ¹⁸ or noncentral geographic atrophy (GA). In group 1, both AMD and control participants were recruited and only one eye was selected for the study. Control participants were spouses and friends of the AMD participants of a similar age range. When both eyes met the inclusion criteria in this series, the eye with the better visual acuity was chosen as the study eye. In group 2, only participants that met the inclusion criteria for AMD in both eyes were included. Exclusion criteria for any study eye included the presence of choroidal neovascularization (CNV), central GA, significant cataracts, any corneal pathology that could compromise vision, glaucoma, amblyopia, presence of diabetes, neurological or systemic disease affecting vision or if the participant was taking any medication known to affect retinal function (eg, hydroxychloroquine). Participants were also excluded if they had any physical and/or mental impairment preventing them from participating in this study or inability to sign a consent form. 

Procedures

All participants underwent standard measurements of BCVA (4-meter logMAR chart) and microperimetry examination, followed by multimodal imaging. This was composed of color fundus photography using a nonmydriatic fundus camera (Canon CR6-45NM; Canon, Saitama, Japan) and near-infrared reflectance (IR), short-wavelength fundus autofluorescence (SW-FAF) and high-resolution optical coherence tomography (OCT) line and volume scans on a Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Germany). All participants then underwent clinical examination by a senior retinal specialist (RHG). 

Microperimetric Examination

All participants underwent microperimetric examination using the Macular Integrity Assessment (MAIA; CenterVue, Padova, Italy) microperimeter prior to performing any tests that may significantly bleach the photoreceptors (such as FAF or fundus photography). Pupillary dilation was performed on all participants using 1 drop of 1% tropicamide and 1 drop of 2.5% phenylephrine to achieve a pupil diameter of at least 6 mm. Identical instructions were given to all participants regarding how to perform the microperimetry examination at any visit. Participants were not dark-adapted and the room illumination was switched off just before each examination commenced to ensure the same degree of light exposure before the examination. All participants were given a few minutes of rest between each test to minimize the effect of fatigue on the test. 

The MAIA performs fundus tracking using a line-scanning laser ophthalmoscope (SLO) with a super-luminescent diode illumination that has a central wavelength of 850 nm. It uses the entire fundus as a reference rather than retinal landmarks (as used in the Microperimeter-1 [MP-1]; Nidek Technologies, Padova, Italy) when performing fundus tracking, capturing the fundus images at 25 frames per second. A red circular fixation target of 1° diameter was used in this study, and Goldman III stimuli were presented against a background of 1.27 cd/m² using a 4-2 threshold strategy. The maximum stimulus luminance was 318 cd/m², creating a dynamic range of 36 dB. 

A customized stimulus grid was designed specifically for the assessment of the macular region in AMD. It consisted of 37 points located at 0°, 1°, 2.33°, 4°, and 6° from fixation (see examples in Fig. 3), designed in a manner to allow a relatively regular sampling density throughout the region, but with a slightly increased density towards the fovea. This foveal-weighted design attempts to balance between sampling the lesions of interest in AMD with adequate density while minimizing the number of stimuli, so as to shorten examination time. The outcome parameters used for analyses were the threshold of each individual test point (point-wise sensitivity [PWS]) and the average of all test points (MS). 

Test reliability was assessed by the frequency of false-positive responses, measured by presentations of suprathreshold stimuli to the optic nerve head (blind spot), which was manually located on the MAIA before the presentation of the first stimuli. Any participant with false-positive responses of more than 25% on any examination was excluded from this study. This cutoff was chosen because the MAIA presents a false-positive stimulus approximately every 1 minute, and there are typically only four to five false-positive stimuli presented in each test given the short duration of the examinations. 

In group 1, all participants performed three full examinations of the study eye within the same session, supervised by one experienced examiner. A subset of participants was reviewed again 6 months following the initial visit, and the same examiner performed two examinations in the same study eye determined at baseline within the same session. In group 2, all participants performed examination of both eyes within the same session supervised by a different experienced examiner. This involved performing one short practice examination (of approximately 1 minute) to familiarize the participants with the testing procedure, followed by two examinations of the right eye (tests 1 and 2) and then two examinations of the left eye (tests 3 and 4). 

Statistical Analysis

The average MS of each group for each examination was calculated by averaging the MS of all patients. The average MS were compared between each test for all groups using an ANOVA with post hoc Bonferroni tests when performing multiple comparisons. Bland-Altman plots were used to inspect test–retest characteristics (including any systematic change and test–retest limits) and if there were obvious associations between the difference and magnitude of MS, confirmed by calculating the rank correlation coefficient (Kendall τ). The average PWS of each group for each examination was determined using a linear mixed effects model, ¹⁹ considering test number as a fixed effect and stimuli points nested within patients as a random effect in group 1. Because repeated tests were performed on two different eyes in group 2, the fixed effect was considered as the interaction between the tested eye and test number, and the random effect was considered as stimuli points nested within the tested eye nested within patients. 

Coefficients of repeatability (CoRs) were calculated as outlined by Bland and Altman ¹⁷ previously, and represents a value for which 95% of the test–retest differences for the same subject are expected to lie. With microperimetric examination, CoR represents both the measurement error of the instrument and the subjective variability. Hence, a larger value of CoR would represent a greater degree of test–retest variability. All statistical analyses were performed using commercially available statistical software (IBM SPSS Statistics, software version 21; IBM/SPSS, Inc., Chicago, IL). 

Results

A total of 30 AMD (69.1 ± 10.5 years, range 50–88) and 14 control participants (66.7 ± 6.0 years, range 57–80) were included in group 1. Of the AMD participants examined, 20 participants were reviewed again (69.9 ± 7.8 years, range 57–88) at approximately 6 months. In group 2, a total of 71 AMD participants (69.6 ± 8.0 years, range 50–86) were included. The visual acuities ranged from 46 to 63 letters on a standard 4-meter logMAR chart. All participants in this study had not performed microperimetry previously. The duration for each examination ranged from 4.6 to 6.6 minutes, and was approximately 5.5 minutes on average. 

Intrasession Test–Retest Variability in One Eye

The average MS was not significantly different between all comparisons of examination pairs in the AMD and control participants of group 1 (P ≥ 0.529 and P ≥ 0.403, respectively). However, the CoR for MS of the first examination pair (between tests 1 and 2) was reduced compared with the second examination pair (tests 2 and 3), as shown in Table 1. 

Table 1

 

CoR for MS

	CoR for MS (dB), Mean (95% Confidence Interval)
	First Examination Pair*	Second Examination Pair†
Group 1
Control	2.01 (1.28–2.73)	1.10 (0.71–1.50)
AMD	2.32 (1.78–2.87)	1.08 (0.83–1.33)
AMD,§ 6 mo	1.00 (0.71–1.30)
Group 2
AMD	1.77 (1.51–2.03)	1.56 (1.33–1.80)

*  Group 1 = test 1 versus test 2 (study eye), group 2 = test 1 versus test 2 (right eye).

†  Group 1 = test 2 versus test 3 (study eye), group 2 = test 3 versus test 4 (left eye).

§  A subset of participants whose study eye was retested six-months following baseline.

Bland-Altman plots did not show significant correlations between the average and difference of MS of both groups (test 2 versus test 1, τ = −0.197, P = 0.129; test 3 versus test 2, τ = 0.102, P = 0.432). An example for AMD participants in group 1 is shown in Figure 1. 

Figure 1

 

Bland-Altman plots of MS for AMD participants of group 1, with horizontal dashed lines representing upper limits of 95% of the mean (+2 SD) and lower limits of 95% of the mean (−2 SD) from top to bottom, respectively, and the horizontal solid black line representing the mean. There was no relationship between the difference and magnitude of MS in both examination pairs.

Bland-Altman plots were then used to inspect the test–retest characteristics of both groups for PWS, and an example for the AMD participants in group 1 is shown in Figure 2. The test–retest variability appeared larger between the first examination pair (tests 1 and 2) than the second examination pair (tests 2 and 3). A positive increase in average PWS is also noted in the first examination pair, but not the second. 

Figure 2

 

Bland-Altman plots of PWS for AMD participants of group 1, with horizontal dashed lines representing upper limits of 95% of the mean (+2 SD) and lower limits of 95% of the mean (−2 SD) from top to bottom, respectively, and the horizontal solid black line representing the mean. This example illustrates the larger test–retest variability between the first examination pair (tests 1 and 2; left) than the second (tests 2 and 3; right). More points of positive test–retest difference can also be observed in the first examination pair, skewing the mean difference positively and indicating a learning effect. The scale bar represents the number of overlapping points for each point.

Results of the linear mixed-effects model revealed a significant improvement in average PWS between test 1 and test 2 in the AMD and control participants in group 1 (P < 0.001 for both the control and AMD group). However, there was no significant improvement in average PWS between test 2 and test 3 for both AMD and control participants in group 1 (P = 0.774 and P = 0.375, respectively) as shown in Figure 3. 

Figure 3

 

Results of the linear mixed-effects model of serial examinations in control participant of group 1 (top left), AMD participants of group 1 (top right), a subset of AMD participants in group 1 examined at 6 months after the baseline examination (bottom left) and AMD participants of group 2 (bottom right). For the subset of AMD participants (n = 20) in group 1 examined again at 6 months (bottom left), results are plotted for retinal sensitivity of these participants at baseline (gray) and 6 months (black). There was a significant increase (P < 0.001) in average PWS between first examination pair (test 1 and 2) in all groups except the subset of AMD participants examined after 6 months in group 1. Error bars represent the 95% confidence interval.

PWS CoR could not be determined for the first examination pair in control and AMD participants of group 1 since there was a significant increase in PWS between the examinations. The PWS CoR was greater for AMD than control participants of group 1 in the second examination pair (Table 2). 

Table 2

 

CoR for PWS

	CoR for PWS (dB), Mean (95% Confidence Interval)
	First Examination Pair*	Second Examination Pair†
Group 1
Control	N/A	3.74 (3.37–4.11)
AMD	N/A	4.12 (3.92–4.32)
AMD,§ 6 mo	4.18 (3.72–4.65)
Group 2
AMD	N/A	4.37 (4.26–4.48)

 N/A, not applicable, because there was a statistically significant change between the examinations.

*  Group 1 = test 1 versus test 2 (study eye), group 2 = test 1 versus test 2 (right eye).

†  Group 1 = test 2 versus test 3 (study eye), group 2 = test 3 versus test 4 (left eye).

§  A subset of participants whose study eye was retested 6 months following baseline.

Intrasession Test–Retest Variability in One Eye at 6 Months

A total of 20 AMD participants in group 1 performed microperimetry on the same study eye again at 6.4 ± 0.9 months following baseline. The average MS was not significantly different between test 1 and 2 (P = 0.933). There was also no significant association between the difference and magnitude of MS on Bland-Altman plots (test 2 versus test 1, τ = −0.042, P = 0.795). The average PWS was also not significantly different between the first and second tests (P = 0.433), showing no systematic improvement within the same session at 6 months. This is in contrast to findings at baseline, where a significant improvement was noted for the same cohort of participants examined (Fig. 3). The average PWS at 6-month follow-up visit appeared lower compared with that at baseline when considering test 2 (where no learning effect is present at both visits), suggesting a decline in sensitivity over 6 months (Fig. 3). 

Intrasession Test–Retest Variability in Two Eyes

For participants in group 2, the average MS was not significantly different between pairs of examinations for the right and left eyes (P = 0.364 and P = 0.787, respectively). Examination of Bland-Altman plots also did not reveal any significant relationship between the difference and magnitude of MS (test 2 versus test 1, τ = −0.049, P = 0.550; test 4 versus test 3, τ = 0.006, P = 0.938). The CoR for MS was similar between the first examination pair (test 1 versus test 2, right eye) and the second examination pair (test 3 versus test 4, left eye). However, there was a significant increase in average PWS between the first examination pair (P < 0.001), but not the second examination pair (P = 0.141); therefore, PWS CoR could be determined only for the second examination pair (Table 2). 

Examples of Study Implications

To illustrate the implications of these findings in monitoring disease progression, two examples are shown in Figure 4; the CoRs for MS and PWS are assumed to be values obtained from AMD participants from group 1 for illustrative purposes. 

Figure 4

 

Two examples of AMD participants illustrating the findings of this study. In the first example (A, B), a −0.38-dB change in MS occurs in the presence of a −6-dB change overlying an area of noncentral GA over 6 months. This localized change occurring with disease progression falls outside the 95% confidence interval of point-wise test–retest limits, whereas the change in MS falls within its test–retest limits. This example illustrates how localized pathological changes can be missed when considering MS only. In the second example (C, D), a change of −2.49 dB in MS is considered to be a significant change, but occurs with up to seven points showing a loss of 6 dB or more over 6 months. This example illustrates that by the time a significant change is noted on MS, there has already been a marked decline in retinal sensitivity when considering PWS.

In the first example (Figs. 4A, 4B), a −0.38-dB change in MS occurs in the presence of a localized change of −6 dB in an area that developed GA over 6 months. Even assuming the lowest CoR for MS of ±1.08 dB (Table 1), this defect is not considered significant. However, this point is considered to have a significantly changed even when considering the highest CoR for PWS of ±4.37 dB (Table 2). This illustrates that a localized defect can be easily missed when considering MS. 

In the second example (Figs. 4C, 4D), a −2.49-dB change in MS over 6 months falls just outside of the CoR for MS of ±2.32 dB, if the participant performed microperimetry for the first time at baseline. However, up to seven points are considered to have displayed a significant decline in retinal sensitivity (loss of ≥6 dB, outside the CoR for PWS of ±4.37 dB; Table 2). This example illustrates that marked loss in retinal sensitivity has occurred by the time that a significant change is picked up on MS, especially if the participant has not performed microperimetry before. 

Discussion

This study demonstrates a significant intrasession learning effect, but only between the first and second examinations when performing microperimetry for the first time. This learning effect was not present in a subset of participants examined after 6 months. Therefore, intrasession test–retest variability can be minimized by discarding the first examination to avoid the influence of this learning effect. These findings have important implications for design of clinical trials and testing in clinical practice when using microperimetry. 

To determine the test–retest variability within a single session involving multiple examinations, it is crucial first to determine whether a significant systematic change, such as a learning effect, is present. A systematic change inevitably increases the test–retest variability of the examinations considered. Previous studies have consistently reported an absence of a significant learning effect using microperimetry in healthy subjects ^14,15 and subjects with macular disease ²⁰ when considering MS. Similarly, our findings confirm the lack of any significant difference in MS between any pairs of examination in any group. 

However, we found a consistent improvement between the first and second examinations during the baseline examination session, and not subsequent examinations within the same session, when considering analysis of PWS. This improvement is also observed in participants who have performed a short practice test (approximately 1 minute in duration). A recent report by Cideciyan and colleagues ²¹ also examined whether there was a learning effect by PWS analysis when performing dark-adapted, red-on-red microperimetry in participants with inherited macular degenerations. They did not observe a significant learning effect in their study, and this discrepancy may be attributed to the different testing parameters used (such as background luminance, and size, color, number, and location of the test stimuli). Additionally, the discrepancy could also have arisen from the statistical analysis applied, since a variance component analysis incorporating test–retest as a random effect was used in that study, rather than considering test sequence as a fixed effect. ¹⁹ Applying the same method of analysis to our data, we also did not find test–retest as a significant effect (data not shown), even though all groups displayed a consistent learning effect when using a linear mixed-effects model that considered test–retest as a fixed effect. When a factor is considered as a random effect, the factor is considered as being randomly sampled from a population, and that the quantity of this factor will depend on chance. However, factors considered as fixed effects are factors that are measured without error, being manipulated during an experiment to determine its effect on an entire population, such as the test sequence in this study. ²²  

We did not detect a significant learning effect between the second and third examinations of the same eye in all participants of group 1, or between the third and fourth examinations of the session of the second eye in participants of group 2. On the whole, these results suggest that no significant systematic changes are likely to occur following the first test, even when testing the fellow eye for the first time. This suggests that the results from examinations following the first examination are likely to represent the true value of retinal sensitivity of participants, with inherent variability included. 

A previous report that found higher retinal sensitivities in the second eye tested within the same session after assessment of both eyes suggested that adaptation to the low background luminance of the microperimeter (1.27 cd/m²) may be responsible for this observation (Notaroberto NF, et al. IOVS 2012;53:ARVO E-Abstract 4828). To avoid the effects of adaptation, we ensured that each test was performed under identical settings by turning on the room lighting following each test. Therefore, the systematic increases in our study and previous studies are likely to be a result of a learning effect, rather than adaptation. 

Additionally, there was also no significant learning effect present between the first and second examination when a subset of participants in group 1 were reviewed approximately 6 months following the initial examination. This suggests that results obtained from the first examination during this visit are likely to represent the true value of retinal sensitivity with inherent variability included. However, what is not known is whether there is a significant intersession learning effect, and this could not be determined from this study because clinical progression and functional decline were observed in this subset of participants. Although it is possible that there may be some learning between sessions, it is likely to be small if the true value can be accurately established at each visit. Further studies are required to examine and confirm this, and also whether the intrasession learning effect remains extinguished if participants are retested at longer review intervals. 

Our results of test–retest variability are comparable to findings from previous studies. Chen and colleagues ²⁰ reported a PWS CoR of ±5.56 dB using the MP-1 with 68 test stimuli. This was similar to the PWS CoR of ±4.76 dB performed under similar conditions (group 2; brief training before commencing the examination with an experienced clinician) in eyes with AMD using 37 test stimuli. The larger CoR obtained by Chen and colleagues ²⁰ may be attributed to a greater number of test stimuli used in their study, resulting in longer test durations of approximately 12 minutes, as compared with approximately 5.5 minutes in this study. 

In addition to determining the intrasession test–retest variability under different conditions, this study also highlights the benefits of performing point-wise analysis rather than averaging multiple measurements with microperimetry. Averaged values are useful in providing an overall representation of the retinal sensitivity over the area averaged, and reduce the test–retest confidence limits in a similar way to obtaining multiple measurements over the sampled area. Depending on the spatial density and size of the stimulus, averaged values may underestimate localized pathological changes that occur in AMD, ^{6,8–10,23–27} such as areas that subsequently develop GA ¹¹ or the expansion of a slow-progressing atrophic lesion. ²¹ This is particularly important when the distance between each test stimulus is larger than the size of the lesion of interest, or the expected rate of progression. Point-wise analysis of a region of interest can allow localized changes occurring at each test stimulus to be better represented during analyses. 

The findings of this study have several implications for the design and analysis of clinical trials that use microperimetry as an outcome measure. First, the learning effect evident between the first and second examination suggests that functional defects observed on the first test may be falsely identified. This was especially evident in healthy participants, in whom functional defects (which were not expected in eyes without pathology) were present during the first examination, but disappeared consistently on subsequent examinations. Second, discarding the first examination during the first session can reduce test–retest variability influenced by a learning effect and allow the test–retest confidence limits to be more accurately applied to both eyes. We also noted in this study that a short practice examination (of approximately 1 minute in duration) is not sufficient in eliminating this learning effect under these test conditions, and performing a full examination as training would be prudent. Although the results of this study suggest that a learning effect is absent on repeat examination 6 months following the initial examination, it is not known whether this is the case with longer intervals between follow-up examinations. Therefore, discarding the first examination of a session for subsequent visits may still be beneficial in minimizing the test–retest variability, although future studies are required to examine intersession changes. However, caution must be made when applying these findings to microperimetric examinations with different test parameters and/or protocols, as the extent of learning and fatigue under those conditions is unknown. The findings of this study should not be generalized to subjects with poor central visual acuity, those with more advanced disease, including central GA or CNV, or other macular diseases; the generalizability of these findings remain to be investigated. We also have observed in our clinic that test–retest variability was larger for examinations supervised by less experienced examiners, and therefore recommend supervision by experienced examiners to minimize the variability. Finally, this study also underscores the benefits of performing point-wise analysis to better reflect the functional changes measured by microperimetry, which averaged values may underestimate. 

In summary, we report that the intrasession test–retest variability represented by PWS CoR to be ±4.37 dB or less between subsequent examination pairs of the same or fellow eye within the same session. There was a significant learning effect between the first and second examination within the same session, but not the subsequent examination pairs of the same or second eye. The learning effect was extinguished when a subset of participants was retested 6 months after the initial examination. These findings suggest that the test–retest variability influenced by the learning effect can be minimized by discarding the first examination of participants performing microperimetry for the first time, and these findings have important implications for both the design and analysis for clinical trials and also clinical practice using microperimetry. 

Acknowledgments

Supported by National Health and Medical Research Council (NH&MRC) Project Grant 1027624, NH&MRC Practitioner Fellowship 529905 (RHG), a Macular Disease Foundation Research grant, the Bupa Health Foundation (Australia), a William Angliss (Victoria) Charitable Fund Research grant, and Murray Redvers and Rodney Alastair Brownless Perpetual Charitable Trust funding administered by the Trust Company. The Centre for Eye Research Australia receives operational infrastructure support from the Victorian Government and is supported by NH&MRC Centre for Clinical Research Excellence Award 529923. 

Disclosure: Z. Wu, None; L.N. Ayton, None; R.H. Guymer, None; C.D. Luu, None 

Footnotes

 RHG and CDL are joint senior authors. RHG and CDL contributed equally to the work presented here and should therefore be regarded as equivalent authors.