Screening for Hydroxychloroquine Retinopathy—Can We Do Better?

Screening for hydroxychloroquine-induced retinal toxicity presents a challenge to the screening ophthalmologist, and even to the retina specialist. The difficulty lies in the lack of consensus of a single screening test or criterion that has demonstrated ideal sensitivity and specificity. With the aim of investigating a testing modality that has promise of detecting early changes in macular sensitivity reflecting toxicity, the authors, Iftikhar et al,¹ investigate the potential of microperimetry testing to be used as a screening test, and they compare the testing results to results obtained with multifocal electroretinogram (mfERG) testing.

As hydroxychloroquine is the treatment of choice, and even the cornerstone of treatment for several autoimmune diseases,² the goal of screening is to survey for evidence of damage enabling confident determination of toxicity, which, if found, would justify recommendations to discontinue a systemic medication that is otherwise likely benefitting the patient.³ With prevalence rates of retinal damage difficult to calculate but estimated in the range of 7.5% for those patients using hydroxychloroquine for longer than 5 years,⁴ it is critical to develop objective criteria for the evaluation for toxicity in the tests performed. False positives may guard the patient against developing a drug-related retinal toxicity, but other potential systemic treatments for underlying autoimmune diseases may be less desirable because they often carry with them some additional systemic risk.⁵ With evidence that eyes with moderate and severe toxicity have changes that are not only irreversible but progressive,^6–8 the goal of screening is to preserve visual function and therefore focuses on the detection of definite but early toxicity.

With an array of available screening tests, the Academy of Ophthalmology has enumerated tests it recommends and others which it recommends against.^5,9,10 These recommendations are updated as data accrue and newer testing modalities evolve. Most recently, in 2016,⁵ the recommendations include the use of a functional test, specifically a Humphrey visual field (HVF) 10–2 with a white test spot, and imaging with spectral domain optical coherence tomography. In addition, screening ophthalmologists should evaluate for risk factors including the following: dose, duration, renal function, and tamoxifen use. Knowledge of the patient’s ethnicity is also useful to ensure that appropriate testing is targeted to the area of potential damage.^11,12

Structural testing and functional testing provide complimentary data useful in evaluating for potential toxicity. Documentation of decreased function provides evidence of a detrimental effect and justifies recommendations to stop the medication. However, although functional testing, such as visual fields, can be reassuring when reliable and normal, they are notoriously variable and, in many cases, difficult to interpret. Even when reliable, the results of visual field testing are an indirect measure of retinal function. Multifocal electroretinograms are historically considered to be the “gold standard” in definitively diagnosing retinal toxicity from hydroxychloroquine. Although the mfERG has its limitations and drawbacks—its limited availability, need for expert technical administration, good patient fixation and cooperation, and variability—it does have the positive attribute of being a measure of function that records directly from the retina while maintaining spatial information. Characteristic abnormalities on mfERG testing provide convincing evidence of retinal dysfunction in patterns consistent with drug toxicity, and published quantitative analyses provide guidelines for analysis.^13–16

Imaging modalities are attractive screening tests because they provide insights into the anatomy of the retina, which can reveal characteristic changes of toxicity and require less patient cooperation and have good reproducibility. Optical coherence tomography has added considerably to the screening armamentarium and has been demonstrated to be highly useful in the detection of structural changes that accompany hydroxychloroquine toxicity.^4,5,17,18 However, the clinical use of optical coherence tomography has been limited to a qualitative inspection of images. The findings with consensus agreement that are deemed to be pathognomonic for toxicity are typically found in overt and advanced toxicity.⁵ As the goal of screening is to detect early changes, the lack of a consensus agreement for identification of the earliest stages by optical coherence tomography imaging leaves evaluations subject to individual interpretation. Qualitatively deciphering between normal variation and effects of signal strength and directional effect of acquisition from the subtle findings of early structural changes that accompany pathology is not straightforward and currently requires interpretation by trained and expert graders.¹⁹ The establishment of quantitative criteria would remove the need for expert inspection and reduce the potential for false-positive and false-negative errors.

With this background, we can understand the motivation for the authors to explore the modality of microperimetry, which measures functional thresholds while imaging and tracking the fundus. Establishing the utility of a new modality or method of screening would necessitate agreement with a gold-standard/consensus test and, ideally, demonstration that longitudinal changes on the screening test support the progressive nature of the findings—from early findings to a time point where there is evidence of definitive toxicity. To this end, the authors explore testing using microperimetry and correlate the findings with performance on “gold-standard” mfERG testing.^14,16

As HVF testing has long been a recommended screening test, the differences between microperimetry testing and HVF testing should be considered, and the authors acknowledge omitting HVF as a limitation of their study. Although both tests use a very similar 10° pattern over the macula, the same stimulus size, and even similar repeatability parameters, there are also several differences.²⁰ One apparent difference is the ability of microperimetry to accommodate noncentral fixators with fundus imaging and tracking. However, as patients with even moderate severity hydroxychloroquine toxicity have excellent central visual acuity and fixation, it is not clear whether this feature is used or necessary in hydroxychloroquine screening. Another significant difference includes the nature of the background illumination—with microperimetry testing against a mesopic background at 1.27 cd/m², versus HVFs, which use a photopic background at 10 cd/m².²⁰ The dynamic range of testing varies greatly between the two instruments with microperimetry testing over two log units and HVF testing over five log units. Both the differences in dynamic range and the background sensitivities could affect the results of these visual field tests. The limited dynamic range of the MP1 microperimeter leads to floor and ceiling effects, which can limit its utility at extreme ends of disease processes (i.e., very subtle or very severe). And, the difference in backgrounds could lead to measures of different underlying cellular physiology. As rod and cones have different but overlapping light ranges of sensitivities, testing in different ranges of background illumination could lead to evaluation of both rod and cone function under mesopic conditions but only evaluate cone functioning under photopic conditions where rods are saturated.²¹ Although both microperimetry and HVF are perimetric tests of the central macula, and the authors state that they are clinically equivalent—they are not truly equivalent tests, and differences between the tests may be revealing.

The results demonstrated in this article indicate that microperimetry—under the objective parameters defined in this study—was not as sensitive as mfERG criteria. To have had nearly 200 participants tested on both specialized tests is commendable and does provide a unique data set. With microperimetry allowing for registration of visual field threshold testing while imaging the fundus, structure and function correlations are possible. These types of analyses could be insightful in understanding hydroxychloroquine toxicity but were beyond the scope of this report. Perhaps, future planned analyses may address some of the spatial information that could be extracted from this perimetry test.

As a research tool, microperimetry may be a modality enabling the exploration of structure and function correlations and further understanding of retinal physiology. Although the purpose of this study is set forth to assess the sensitivity and specificity of microperimetry as a screening test, even with a positive study result, the barriers to using microperimetry would be quite high, and the authors acknowledge that this modality is unlikely to be suitable for primary screening.

However, the study by Iftikhar et al¹ is evidence of the absence of a conclusive screening test for hydroxychloroquine toxicity. Although hydroxychloroquine has been used widely for decades, and we have progressed in our knowledge of toxicity, we still have gaps in our ability to confidently identify early changes. Promise may lie in the development of objective and quantitative evaluation of optical coherence tomography images,^16,22–24 as this imaging modality is ubiquitous in even general ophthalmology offices and requires little patient cooperation. The development of quantitative criteria for this screening test would alleviate the reliance on domain expertise and bring to attention those eyes with a suspicion of abnormality for further specialized evaluation, leading to a greater preservation of visual function in patients taking this drug.