Multifocal Electroretinographic Evaluation of Long-term HydroxychloroquineUsers

Objectives  To observe the long-term effects of hydroxychloroquine sulfate on retinalelectrical activity by multifocal electroretinography (mfERG) and to evaluatethe regional variation of retinal dysfunction in subjects with hydroxychloroquineretinopathy.

Methods  Multifocal ERG with 103-hexagon stimulation was performed on 19 patients(36 eyes) treated with hydroxychloroquine for systemic lupus erythematosus,rheumatoid arthritis, or localized atypical scleroderma. Visual acuity testing,Amsler grid testing, and Ishihara color vision testing were also performed.In 2 of the patients, hydroxychloroquine was discontinued due to concernsabout toxicity. Both of these patients had additional mfERG performed afterdiscontinuation of medication.

Results  Twelve patients (19 eyes) had a normal response density in one or botheyes, including 6 patients (12 eyes) with a low lifetime dose (≤438 g)of hydroxychloroquine who had normal response densities in both eyes. Elevenpatients (17 eyes) had abnormal response densities in one or both eyes, and2 of these patients (4 eyes) had significant attenuation of response densitiesin almost the whole tested field; 4 patients had a normal mfERG result forone eye but had a slight decrease of response densities for the other eye.There were 4 patterns of abnormal mfERG amplitude change observed: (1) paracentralloss, (2) foveal loss, (3) peripheral loss, and (4) generalized loss. Implicittimes were abnormal for pericentral responses in 3 patients. The results ofcolor vision and Amsler grid testing were normal, except for one patient witha generalized loss pattern. In 2 subjects in whom hydroxychloroquine toxicitywas suspected, response densities improved after termination of hydroxychloroquine.

Conclusions  Long-term hydroxychloroquine use may be associated with mfERG abnormalities.The mfERG appears to detect retinal physiological change earlier than visualacuity testing, color vision testing, or Amsler grid testing can. The greatestvalue of the mfERG is in differentiating a retinal cause and, hence, providingimportant evidence for hydroxychloroquine toxicity, for whatever visual fieldloss is apparent on perimetry.

Although chloroquine and hydroxychloroquine sulfate are 2 of severalantiprotozoal drugs that have been used for many years in the treatment ofmalaria, their preferred usage in the United States is for treatment of variousrheumatic diseases, particularly systemic lupus erythematosus and rheumatoidarthritis. However, long-term hydroxychloroquine and chloroquine use can causesevere visual loss due to a toxic effect on the outer retina and the retinalpigment epithelium. Chloroquine retinopathy was first reported by Hobbs etal¹ in 1959, and hydroxychloroquine retinopathywas first reported by Shearer and Dubois² in1967. Ophthalmoscopy, slitlamp examination, some psychophysical methods (Amslergrid, perimetry, color vision testing, and visual field testing), and electrophysiology(full-field electroretinography [ERG]) have been used for the early detectionof these diseases.^3,4 Becauseit is a sum response of the whole retina, full-field ERG cannot sensitivelydetect the local variation of the retinal function in subjects with earlyhydroxychloroquine retinopathy.⁵ Focal ERGis more sensitive than full-field ERG for the detection of hydroxychloroquinetoxicity in the retina⁵; however, focal ERGcannot test the function in multiple areas of the retina or provide topographicalinformation. Kellner et al⁶ observed characteristicabnormalities in the multifocal ERG (mfERG) response in 2 patients with differentdegrees of chloroquine retinopathy and found the mfERG to be more sensitivethan Goldmann perimetry, visual acuity testing, and the full-field ERG indetecting early chloroquine retinopathy. One of us (R.K.M.) previously reporteda case⁷ of hydroxychloroquine retinopathy,where the response densities of the mfERG were attenuated, while the full-fieldERG was normal. The patient had a normal full-field ERG result. We postulatedthat the mfERG could be an objective and sensitive method for the early detectionof hydroxychloroquine retinopathy. In this study, we retrospectively evaluatedthe mfERG results of asymptomatic and symptomatic patients who took hydroxychloroquineand identified 4 patterns of topographic changes in response density. Theparacentral loss of amplitude with prolonged implicit times was the most specificfor hydroxychloroquine toxicity.

Multifocal ERG was performed on 19 patients (36 eyes; both eyes weretested in 17 patients and only one eye was tested in 2 patients) treated forsystemic lupus erythematosus, rheumatoid arthritis, or localized atypicalscleroderma with hydroxychloroquine sulfate (Plaquenil; Sterling Winthrop,Inc, New York, NY). Visual acuity testing and, in some patients, color visiontesting, Amsler grid testing, Humphrey static perimetry (30-2), and fluoresceinangiography were performed. Eyes with concomitant diseases (eg, age-relatedmacular degeneration and histoplasmosis scars) were not included in the analysis.The response densities of the first-order kernel of mfERGs were analyzed ringby ring at each eccentricity from fixation and compared with those of ourcontrol values for response density and implicit time. Control values wereobtained from 20 healthy age-matched subjects (20 eyes; mean ± SD age,50.8 ± 14.5 years; compared with the ages of hydroxychloroquine userswith a group t test, P>.05)with a normal ophthalmoscopic appearance of the fundi and normal visual acuities(20/20 or better).

For the mfERG, the test was performed monocularly as the subjects viewedthe monitor screen through a video camera system that allowed monitoring ofthe eye position. The stimulus was presented on a 48-cm high-intensity monochromaticmonitor, consisted of 103 hexagon elements, and covered the central field44° horizontally. The fixation target was a small cross in the centerhexagon. The optics of the video camera corrected the refractive error ofeach eye for the test distance. The hexagons were modulated between a light(400 candelas [cd]/m²) and a dark (1 cd/m²) state accordingto a binary pseudorandom m-sequence. The refresh rate was 75 Hz. The pupilswere dilated, and the corneas were topically anesthetized before contact lensplacement. The mfERG results were recorded with bipolar contact lens electrodes(Burian-Allen model; Hansen Instruments Inc, Iowa City, Iowa). The amplifiergain was 100 000, and the bandpass was 10 to 300 Hz. The total recordingtime was 7 minutes 17 seconds, divided into 16 segments.

The 103 traces were grouped from center to periphery into 6 rings (ring1, foveal patch, 1 arc degree in diameter; ring 2, 1-6 arc degrees in diameter;ring 3, 6-12 arc degrees in diameter; ring 4, 12-21 arc degrees in diameter;ring 5, 21-31 arc degrees in diameter; and ring 6, 31-44 arc degrees in diameter).In each group, the averaged first-order kernel was analyzed. The N1-P1 (N1indicates first negative wave; P1, first positive wave) response densities(amplitude/area, measured in nanovolts per degree squared) were measured fromthe N1 trough to the P1 peak. The normal ranges for these response densitiesand implicit times were defined by calculation of mean ± 95% confidenceintervals of 20 age-matched control eyes.

Table 1 provides the clinicalfindings in all 19 patients, ordered according to total cumulative dose (ingrams). Table 1 describes theages of the patients, the daily dose adjusted for weight, the duration ofmedication use, the fundus findings, and the results of color vision testing,Amsler grid tests, and fluorescein angiography. The mean ± SD age ofthe patients analyzed was 54.4 ± 14.9 years. The daily dose rangedfrom 200 to 400 mg/d, and the duration of treatment ranged from 1 to 20 years.The daily dose adjusted for body weight ranged from 2.0 to 9.0 mg/kg per day.The cumulative dose ranged from 110 to 2920 g.

Table 2 contains the visualacuities and results of Humphrey static perimetry, where available. The correctedvisual acuities ranged from 20/20 to 20/50. Also in the table are Humphreyvisual fields for 15 subjects (30 eyes), and a description of the responsedensities.

In the 15 eyes in which 500 g or less hydroxychloroquine was used andfor which mfERG and Humphrey visual field data are available, 6 had normalmfERG findings and normal automated perimetry findings, 8 had normal mfERGfindings and abnormal automated perimetry findings, and 1 had abnormal mfERGfindings and normal automated perimetry findings. In the 14 eyes in whichmore than 500 g of hydroxychloroquine was used and for which mfERG and Humphreyvisual field data are available, 2 had normal mfERG findings and normal automatedperimetry findings, 3 had normal mfERG findings and abnormal automated perimetryfindings, 3 had abnormal mfERG findings and normal automated perimetry findings,and 6 had abnormal mfERG findings and abnormal automated perimetry findings.

Table 3 provides the controlvalues of N1-P1 response density and P1 implicit time from the 20 controlsubjects. Table 4 shows the resultsof the mfERGs of the 19 patients (36 eyes). Twelve patients (19 eyes) hada normal response density in one or both eyes (cumulative dose, 110-2336 g;maximum, 16 years of use). Eleven patients (17 eyes) had abnormal responsedensities (cumulative dose, 438-2920 g; hydroxychloroquine administrationfrom the age of 3-20 years) in one or both eyes; 2 of these patients (4 eyes)had significant attenuation of response densities in almost the whole testedfield (cumulative dose, 1533-2482 g; 11-17 years of hydroxychloroquine use);and 4 of these patients had normal mfERG results in one eye but had a slightdecrease of response densities in another eye (cumulative dose, 438-2336 g;6-16 years of hydroxychloroquine use). We identified 4 types of abnormal mfERGresults (Figure 1, Figure 2, Figure 3, and Figure 4). (1) Some results showed a decreasein N1-P1 response density in the paracentral area and/or prolongations ofcorresponding P1 implicit times (2 patients [4 eyes]). In patient 13, theresponse densities and the implicit times showed paracentral change (cumulativedose, 1095 g; 7.5 years of hydroxychloroquine use) (Figure 1). In patient 18, the response densities showed generalizedloss but the P1 latencies were prolonged for paracentral responses (cumulativedosage, 2482 g; 17 years of hydroxychloroquine use) (Figure 4). (2) Some results showed a decrease in the central areaalone (7 patients [9 eyes]) (cumulative dose, 438-2336 g; 3-16 years of hydroxychloroquineuse) (Figure 2). (3) Some resultsshowed a decrease in the peripheral area alone (1 patient [2 eyes]) (cumulativedose, 2920 g; 20 years of hydroxychloroquine use) (Figure 3). (4) Some results showed a generalized decrease in theentire tested field (2 patients [4 eyes]) (cumulative dose, 1533-2482 g; 11-17years of hydroxychloroquine use) (Figure 4). The implicit times for the P1 peak of the mfERG responses werenormal when analyzed in the aggregate and in all individuals except for patients13 (Figure 1, Figure 5, and Table 5),18 (Figure 4), and 15 (rings 4 and5 in the left eye only). The results of color vision testing were normal inall patients. The results of Amsler grid testing were normal for all eyes,except for 1 patient (2 eyes) with 17 years of hydroxychloroquine use whodemonstrated a generalized pattern loss (patient 18). A decrease in visualacuity, when present, was usually because of anterior segment findings.

In 2 of the patients (patients 9 and 13) who underwent a longitudinalmfERG study, the response densities demonstrated a trend toward improvementafter the cessation of hydroxychloroquine use. Figure 6 shows the change of their averaged response density inrings 1 and 2 relative to the period of termination of hydroxychloroquineuse. Significant prolongations of implicit times occurred for patient 13,particularly for those responses that correspond to rings 2, 3, and 4. Atthis time, early hydroxychloroquine retinopathy was clinically recognizedon this patient (Figure 1C and Figure 5 and Table 5). The patient stopped drug treatment and follow-up mfERGscontinued to demonstrate prolonged implicit times. The visual acuities forthese 2 patients improved mildly in the same period. An additional patient(patient 18) with a generalized loss pattern stopped using hydroxychloroquinedue to mfERG findings and central visual field loss. Follow-up mfERG testresults were not available for this patient.

Although hydroxychloroquine originally was thought to be a safer drugthan chloroquine, retinal toxicity from hydroxychloroquine can occur. Thelong-term safety of hydroxychloroquine and most effective means of screeningfor retinal toxicity are still subject to debate. In the study by Johnsonand Vine,⁸ patients receiving dosages up to400 mg/d (≤6.5 mg/kg per day) seemed to tolerate massive cumulative doses(1054-3923 g) of hydroxychloroquine without developing abnormalities in theirvisual acuity, Amsler grid result, color vision testing, and visual fieldtesting. Toxicity has been shown when a cumulative dose of more than 800 gof the drug is ingested.⁹ Also, a toxic reactionhas been reported for those undergoing long-term therapy (10 years, for atotal cumulative dose of 1460 g) in whom the adjusted daily dose was nevergreater than 6.3 mg/kg per day.¹⁰ This suggeststhat other factors, possibly genetic and acquired, may influence which patientsdevelop retinal toxicity. For example, Shroyer et al¹¹ havesuggested that the carrier state for mutation of the gene ABCA4 (the gene that is defective in Stargardt disease) may increasesusceptibility to hydroxychloroquine and chloroquine retinal toxicity. Furthermore,because hydroxychloroquine is cleared through renal and hepatic functions,those with severe kidney or liver disease may be at greater risk of toxicity.

Most of our patients (10 of 11) who had abnormal mfERG results had takenhydroxychloroquine for at least 5 years. This implies that, in the first 5years of hydroxychloroquine use, the incidence of retinopathy caused by toxicityof the drug is low or nonexistent. This finding supports the recommendationfrom the American Academy of Ophthalmology¹² that,for individuals using less than 6.5 mg/kg of hydroxychloroquine per day, screeningcan be modified for the first 5 years to account for the minimal risk of atoxic reaction at this dose.

Three of the patients with abnormal mfERG results (patients 13, 18,and 19) were ingesting higher dosages (8.0, 8.0, and 9.0 mg/kg per day, respectively)than recommended at the time of our study. Both subjects with paracentralloss were taking a higher daily dose by weight, as were the 2 subjects withthe highest cumulative doses. However, overall, the incidence of retinal dysfunctiondid not correlate significantly with the daily dose adjusted for weight. Althoughthe daily dose adjusted for weight may be an important risk factor for toxicity,we found foveal loss or generalized loss of response density for 8 patientstaking less than 6.5 mg/kg per day. Hydroxychloroquine is not retained infatty tissues.¹² The daily dose adjusted forweight is, thus, relatively lower for obese patients, and this index willnot reflect the real dose in lean weight. Thus, we conclude that the durationof hydroxychloroquine use and the total cumulative dose are still importantrisk factors for hydroxychloroquine retinopathy.

All of the patients in this study had visual acuities between 20/20and 20/50 and normal color vision results, as tested. We believe that colorvision testing and fluorescein angiography show findings indicative of toxicchanges typically late in the course of the disease. Thus, color vision testing,although useful in defining visual function in cases of suspected drug toxicity,may not have a significant benefit as a screening tool for early disease.One subject had a change on fluorescein angiography. Amsler grid testing hasbeen suggested to show change earlier in the disease process and has beensuggested for home screening.¹² However, onlyone subject in our series had an Amsler grid change.

Humphrey visual field testing was somewhat useful as a method for evaluatingearly drug toxicity. Our study, in most cases, did not incorporate retestingof visual fields or mfERG to evaluate the reproducibility of abnormal findings.However, many cases of abnormal visual fields can be noted for patients inwhom no hydroxychloroquine toxicity is expected. In some cases, repeat testingof visual fields (data not shown) gave results that were still abnormal butdifficult to interpret. In the lower-dose hydroxychloroquine users in whomretinal toxicity was not expected, we found many instances in which perimetryshowed scattered defects. In these instances when the Humphrey visual fieldwas abnormal and the mfERG was normal, we were confidently able to reassurepatients that the retina was still functioning well and that they could continuehydroxychloroquine therapy despite the abnormal visual field.

In the higher-dose group of patients, we found 3 eyes in which perimetryfound abnormalities while the results of mfERG testing were normal. In 2 ofthese eyes (patient 14), the perimetry results showed scattered abnormalitieswith no other clinical findings of hydroxychloroquine toxicity. We believedthat the patient displayed no signs of hydroxychloroquine toxicity. In thethird eye (the right eye of patient 17), it is possible that the mfERG didnot detect early hydroxychloroquine changes that were apparent on perimetry(the findings on both tests showed abnormalities in the fellow eye). Thus,while it is possible for perimetry to demonstrate changes associated withhydroxychloroquine use earlier than mfERG, in most cases, mfERG supplementedand confirmed abnormalities found on visual field testing. In addition, inthe patient for whom we were concerned about toxicity (patient 19), mfERGshowed abnormalities that were not detected on visual field testing. The greatestvalue of the mfERG is in differentiating a retinal cause and, hence, providingimportant evidence supportive of hydroxychloroquine toxicity, whenever visualfield loss is apparent on perimetry.

We believe that mfERG testing can detect change in retinal functionfar sooner than any other electrophysiological modality in use. However, theclinician is required to interpret these changes and to determine whetherthe mfERG abnormalities are significant enough to warrant discontinuationof hydroxychloroquine therapy. In this retrospective study, 3 patients (patients9, 13, and 18) stopped using the medication for various reasons. Abnormalitieswere seen in the mfERG for each of these patients. Two of these patients (whosemedication duration ranged from 6-17 years) returned for subsequent mfERGs,which, for both, showed an improvement in response density. Serial repeatstudies in these 2 hydroxychloroquine users who stopped using the drug showeda trend toward electrophysiological recovery of mfERG response (Figure 6). This suggests that in early cases, the loss of retinalresponse densities associated with hydroxychloroquine use may be reversible.We did not observe significant reduction in the prolonged implicit times inthe subjects. Implicit time recovery, however, if it is to occur at all forhydroxychloroquine toxicity, may not take place within the same time frameas recovery of response density, and the observation period in our patientsmay have been too short. Other retinal degenerations often show lack of correlationof latency prolongations and changes in amplitudes.¹³

We were not able to determine whether a particular pattern of mfERGabnormality preceded another pattern, except that foveal loss may be an earlyfeature and generalized decrease in response density is the final pattern.Paracentral loss of amplitude (rings 2-4), particularly with prolonged implicittimes, however, is the most specific pattern for hydroxychloroquine toxicity.We have not encountered this latter specific pattern in more than 800 mfERGsthat we have obtained (for various diseases) in the past few years. Seeligeret al¹⁴ reported that the latencies for P1show low topographical variability in control subjects. Thus, prolongationsin latencies for the pericentral responses seem to be a particularly diagnosticfeature of focal retinal dysfunction. The fact that the most characteristicloss on mfERG is located in the paracentral rings would suggest that, forscreening for hydroxychloroquine toxicity, the Humphrey visual field 10-2test may be a better program than the 30-2 test.

The abnormalities in retinal electrophysiology, as detected by the mfERG,occurred earlier than any morphologic fundus change using ophthalmoscopy.We believe that the mfERG may be the most sensitive objective test for theearly detection of hydroxychloroquine retinopathy. Patients who take hydroxychloroquinein higher than the recommended dose (>6.5 mg/kg per day) or for a long duration(>5 years) should be considered for periodic mfERG testing for the early detectionof hydroxychloroquine retinopathy.

In summary, the evidence supports periodic testing with mfERG, whenpossible and available, in patients undergoing long-term hydroxychloroquinetherapy, particularly if there is any clinical suggestion of drug toxicity.An mfERG is also an excellent choice for confirming the absence of retinaltoxicity when perimetry or other tests detect abnormalities. Further studyis necessary to evaluate the sensitivity of the mfERG in early detection ofhydroxychloroquine retinopathy and to determine the role of mfERG in screening.

Correspondence: Raj K. Maturi, MD, Midwest Eye Institute, 201 PennsylvaniaPkwy, Indianapolis, IN 46280 (rmaturi@indyretina.com).

Submitted for publication September 5, 2002; final revision receivedNovember 24, 2003; accepted January 7, 2004.

This study was supported by the Methodist Research Institute and MidwestEye Foundation, Indianapolis, Ind; The Foundation Fighting Blindness, OwingsMills, Md; and Research to Prevent Blindness, New York, NY.