Hydroxychloroquine retinopathy screening guidelines: a false positive

Abstract

Hydroxychloroquine sulphate (HCQ) is widely used for the treatment of a variety of rheumatological and dermatological conditions. Despite the advantages of HCQ as a treatment option, it is important to be aware of its potential retinal toxicity, which may be irreversible and progressive. In December 2020, The Royal College of Ophthalmologists published revised recommendations on monitoring HCQ retinopathy. Our case report highlights some of the shortcomings of blindly following their monitoring algorithm by presenting a case where apparent HCQ retinopathy resolved after Yttrium Aluminium Garnet (YAG) laser capsulotomy. The case reiterates the importance of thorough clinical examination. We suggest that while the acquisition of the spectral domain optical coherence tomography and fundus autofluorescence may be objective, their interpretation is subjective. Even with the use of artificial intelligence algorithms, false positives may be generated if the tests are confounded by copathology. There is no gold-standard test for detecting HCQ toxicity.

Background

Hydroxychloroquine sulphate (HCQ), an antimalarial, has found value as a treatment for a variety of rheumatological, dermatological and recently oncological conditions,.^{1 2} As understanding and research surrounding HCQ use increases, so too does the use of HCQ. The number of patients taking HCQ in the UK in 2016 was estimated at 166 673, with an additional 36 444 newly started in the same year,.³ Recent estimates suggest up to 320 000 people are currently taking HCQ in England, and many more worldwide.⁴ Hydroxychloroquine has a better side effect profile than other disease-modifying antirheumatic drugs and rarely causes renal or hepatic dysfunction, cardiotoxicity or musculoskeletal side effects at the recommended dosage. Despite all the positive attributes HCQ has as a treatment option, it is important to be aware of its potential retinal toxicity. Hydroxychloroquine retinopathy can continue to progress after cessation of HCQ therapy.⁵ In December 2020, The Royal College of Ophthalmologists published revised recommendations on monitoring hydroxychloroquine and chloroquine retinopathy.⁴ These recommendations include a diagnostic algorithm (figure 1). The stated aim of the monitoring algorithm is not to prevent HCQ retinopathy but to detect the earliest definitive signs of presymptomatic toxicity. This will facilitate an informed discussion between the patient and the prescribing physician on further treatment options.⁴ Delaying the discovery of HCQ’s toxic effects, to a stage of visible retinal pigment epithelial damage, risks an eventual loss of foveal structure and visual acuity.⁵ This case report highlights some of the shortcomings of blindly following the algorithm and reiterates the importance of thorough clinical examination and lateral thinking.

Figure 1.

December 2020 The Royal College of Ophthalmologists revised recommendations on monitoring hydroxychloroquine and chloroquine retinopathy flow chart. Pathway taken by our patient is shaded grey. HCQ, hydroxychloroquine sulphate; FAF, fundus auto fluorescence; mfERG, multifocal electroretinogram; SD-OCT, spectral domain optical coherence tomography;HVF, Humphrey visual fields; eGFR, estimated glomerular filtration rate

Case presentation

A Caucasian woman in her 60s presented for screening because she had been taking oral HCQ (2.8 mg/Kg) for 6 years for seronegative rheumatoid arthritis and nodular osteoarthritis. Her treatment also required subcutaneous methotrexate (25 mg one time per week). The patient weighed 70.2 kg (ideal body weight 48.76 kg) with a BMI of 28.85 kg/m², did not have a history of tamoxifen use and had normal renal function with an estimated glomerular filtration rate (eGFR) of 80–90 mL/min/1.73 m².

Investigations

As per the HCQ monitoring guidelines, figure 1, she underwent spectral domain optical coherence tomography (SD-OCT) with the Heidelberg Spectralis (Heidelberg Engineering GmbH, Heidelberg, Germany) and Humphrey (Carl Zeiss AG, Jena, Germany) 10–2 visual field (VF) testing. The 10–2 VF tests were reliable. Testing in the right eye showed 8/8 Early Treatment of Diabetic Retinopathy Study (ETDRS) sectors were below the normal threshold and 7/8 ETDRS sectors were below normal threshold in the left eye,⁶ figure 2. The SD-OCT revealed outer nuclear layer (ONL) thinning in 5/8 ETDRS sectors in the right eye and 4/8 ETDRS sectors in the left eye, figure 3.⁶ The patient was instructed to cease HCQ treatment until further follow-up in a joint rheumatology and ophthalmology clinic and was referred for multifocal electroretinogram (mfERG) at a tertiary centre.

Figure 2.

10–2 visual field test with average deviation per ETDRS sector pre right Yttrium Aluminium Garnet (YAG) laser capsulotomy. Sectors below age matched threshold in red. ETDRS Early Treatment of Diabetic Retinopathy Study;, SITA, Swedish Interactive Testing Algorithm; FL, Fixation loss; FN, False Negative; FP, False positive; MD, Mean deviation; PSD, Pattern standard deviation.

Figure 3.

10–2 visual field test with average deviation per ETDRS sector post right Yttrium Aluminium Garnet (YAG) laser capsulotomy. No ETDRS sectors below threshold. ETDRS, Early Treatment of Diabetic Retinopathy Study; OD, Right Eye; OS, Left eye.

The patient was seen in a joint rheumatology and ophthalmology clinic 6 weeks after the initial screening. The history revealed that the patient had undergone bilateral lens extraction and multifocal lens implantation at a private provider. On examination, it was noted that there was moderate right posterior capsular opacification.

Treatment

A right Yttrium Aluminium Garnet (YAG) laser capsulotomy was carried out with the aim of improving the quality of the imaging.

Outcome and follow-up

The uncorrected visual acuity (UCVA) in the right eye improved from 0.38 to 0.02 logMAR. In the left eye, the UCVA was −0.08 LogMAR. There was no evidence of cystoid macular oedema or peripheral retinal tears. The mfERG was performed, after capsulotomy, to examine retinal function centrally from fixation to 20° eccentricity. In the right eye, very central function appeared reduced and although there was no statistically significant reduction in peripheral concentric circles, analysis of the topography plots suggested some reduction in function in the temporal retina from 15 to 20°. This may be explained by leftover peripheral outer segment of the capsule, which is often found following YAG posterior capsulotomy. The suggested reduction in temporal function noted may be from media opacity as opposed to retinal toxicity. The current Royal College of Ophthalmology guidelines state that mfERG cannot be recommended as a gold-standard test on its own due to evidence showing a high sensitivity (0.9), but poor specificity (0.51).⁷ There was no evidence of left localised dysfunction from the mfERG.

Twelve weeks post-YAG capsulotomy, the patient underwent repeat 10–2 VF and SD-OCT testing for HCQ retinopathy. There was marked improvement in the 10–2 VF thresholds, figure 4. None of the ETDRS sectors were below threshold in either eye. The SD-OCT showed ONL thinning in 2/8 ETDRS sectors (5/8 previously) in the right eye and 3/8 ETDRS sectors (4/8 previously) in the left eye, figure 3, 19 February 2021. This no longer reaches our threshold for a diagnosis of HCQ retinopathy.⁶ ETDRS sector 4 (outer temporal) remained abnormal on SD-OCT after YAG laser capsulotomy, but 10–2 VF sector 4 reverted to normal after YAG laser capsulotomy. Unfortunately, we have no pre-HCQ SD-OCT scan with which to compare.

Figure 4.

SD-OCT thickness maps of the outer nuclear layer with average outer nuclear layer thickness and volume per ETDRS sector, pre 11 December 2020 (above) and post 19 February 2021 (below) right Yttrium Aluminium Garnet (YAG) laser capsulotomy. ETDRS, Early Treatment of Diabetic Retinopathy Study; SD-OCT, spectral domain optical coherence tomography.

Discussion

Hydroxychloroquine retinopathy is a well-documented adverse effect of long-term HCQ use. However, diagnosis is hampered by there being no gold-standard test. Interpretation of VF and SD-OCT results is a subjective process and can result in interobserver variation in diagnosing and monitoring of HCQ retinopathy.^{8 9} In the context of HCQ retinopathy, the most useful test of visual function is the 10–2 VF test. Many patients find that this test is difficult to perform and the reliability of the test maybe degraded by fixation losses, false positives and or false negatives sufficient for the test to be classified as unreliable. In a recent large review of patients taking HCQ for more than 5 years, 33.1% produced unreliable fields; some of these patients attended for repeat field testing.¹⁰ The 10–2 VF test is easier than the 24–2 test used for monitoring glaucoma. A learning effect is well recognised from glaucoma studies. The guidelines only call for VF testing if SD-OCT or fundus auto fluorescence (FAF) is abnormal, figure 1. Our patient’s first tests were reliable and produced a generalised reduction in threshold. In a retrospective review of 10 patients with HCQ retinopathy, there was evidence of toxicity in 50% of the 10 patients with prior VFs available for review. These five patients had 24 VF tests reviewed with 14 tests showing findings missed by the clinician.⁹ The 10–2 VF shown in the paper has excessive fixation losses but shows the typical ring scotoma of HCQ retinopathy with four out of eight ETDRS sectors below the normal threshold.^{6 9} In our patient’s follow-up test, there is a marked increase in the thresholds.

The monitoring guidelines put considerable weight on SD-OCT despite reports of early HCQ retinopathy detected using 10–2 VFs and mfERG with a normal SD-OCT.⁹ This suggests that the SD-OCT is not sensitive enough to be used without additional investigations.

Although the capture of the SD-OCT images is objective, the interpretation is subjective.

In HCQ retinopathy, there are thickness and volume changes that are most often localised to the parafovea. Hasan and colleagues found that the volume of the ONL was more useful metric than the average thickness.⁶ They used the Heidelberg Eye Explorer V.2.0 software (HEYEX) segmentation feature, on macular SD-OCT scans taken using a Spectralis OCT scanner (Heidelberg Engineering—Heidelberg, Germany), to calculate the ONL volume. The scanning protocol used was the 49-line high-definition mode that produced a 6×6 mm macular cube.⁶ The mean ONL volume for each of the macular segments (mm³) in their normal cohort was as follows: sector 1—superior inner (0.11±0.02), sector 2—superior outer (0.33±0.03), sector 3—temporal inner (0.12±0.01), sector 4—temporal outer (0.31±0.03), sector 5—inferior inner (0.10±0.02), sector 6—inferior outer (0.28±0.03), sector 7—nasal inner (0.11±0.02) and sector 8—nasal outer (0.30±0.04).⁶

The association between the inferior macular sector damage and development of toxicity has been documented in the literature.^{8 11} Hasan and colleagues also found certain ETDRS sectors of ONL volume reduction to be most indicative of toxicity, these are sectors 1 (superior inner), 2 (superior outer), 6 (inferior inner) and 7 (nasal inner).⁶

SD-OCT shows significant thinning of the retinal pigment epithelium to outer segments with age in the eight non-central ETDRS sectors. There is significant thinning of the ONL in the inner temporal sector 3 from 35 to 66 years of age.¹²

The mean subfoveal thicknesses of the ONL is significantly thinner (81.70±13.07 µm) in highly myopic eyes (≥−6.0D) compared with normal controls (90.84±11.87 µm) (p=0.002). In addition, several sectors (3 inner temporal, 4 outer temporal, 5 inner inferior, 6 outer inferior and 7 inner nasal) were thinner in the high myopes compared with controls (all p<0.05).¹³

Interpretation of FAF images is even more subjective. The Royal College of Ophthalmologists monitoring guidelines suggest that fundus autofluorescence is not as reliable as VF testing or SD-OCT. The guidelines say it may be a useful additional test; especially for identifying pericentral HCQ retinopathy.⁴ Good quality images usually require pupil dilation with its attendant comorbidity for the patient.

The current Royal College of Ophthalmologists guidelines recommend screening to begin 5 years after commencing HCQ therapy if no additional risk factors are present. If the patient has additional risk factors; concurrent use of tamoxifen, impaired renal function, a dosage over the maximum recommended (5 mg/Kg), or if the patient is taking chloroquine, then screening should commence 1 year after starting therapy. An advantage of carrying out a baseline assessment is to allow a comparison with SD-OCT scans that cannot be due to HCQ retinopathy.

This case report highlights the importance of a thorough examination and investigation of potential confounding causes of abnormal screening results, before escalating to more expensive and niche investigations (eg, mfERG) or exiting the patient from the screening programme,⁴ (figure 1). The consequences of a false-positive diagnosis may result in the patient being taken off an effective, well tolerated, easily administered and inexpensive drug (hydroxychloroquine) and being started on more expensive treatments such as rituximab, which may need hospital admission to be administered, and require regular blood tests for monitoring.

Patient’s perspective.

For myself, this was a roller coaster ride. I had been taking hydroxychloroquine within the recommended limits and had found it helpful. To then be told it might be affecting my retina and that it would need to be stopped raised some uncertainty about how my condition would be managed. When I was told that it was likely that the cause for the test results was the posterior capsular opacification, I felt very relieved. A relative had undergone successful YAG laser capsulotomy so I didn’t have any worries about the laser treatment. After the YAG laser capsulotomy I was pleased to have improved vision and was relieved that the tests had reverted to normal. I am hopeful that others will benefit from my experience.

Learning points.

• There is no gold-standard test for detecting hydroxychloroquine retinopathy.

• While the acquisition of the spectral domain optical coherence tomography and fundus auto fluorescence may be objective, their interpretation is subjective.

• The appearance of the classical ring scotoma on 10–2 visual field testing represents advanced retinopathy.

• Even with the use of artificial intelligence algorithms, false positives may be generated if the tests are confounded by copathology.

• Given the low incidence of hydroxychloroquine retinopathy patients with a suspected diagnosis should undergo clinical examination by an ophthalmologist to confirm the diagnosis before exiting the screening pathway.

Ethics statements

Patient consent for publication

Consent obtained directly from patient(s).