Plaquenil Dose and Retinopathy Risk Calculator

What changed in the 2025 AAO revision

The American Academy of Ophthalmology replaced its 2016 hydroxychloroquine screening recommendations in October 2025 (Marmor, Ahn, Ehlers, Melles, Mieler, Sarraf, and Yusuf; Ophthalmology 2026;133:439–450). The daily dose threshold is unchanged, but the screening protocol itself changed materially. The five differences that alter what happens in the exam lane:

• OCT and wide-pattern FAF are the primary tests. Visual fields are explicitly demoted: “Since the VF is subjective and of variable reliability, it is no longer recommended as a primary method of screening.” Fields and mfERG are now confirmatory tests for equivocal findings.
• Ethnicity no longer selects a visual field size. Both parafoveal and pericentral regions are screened in every patient, because a moderate percentage of patients in each heritage group show the contrary or a mixed pattern. The newer 24-2C algorithm covers both regions in a single test.
• Severe obesity now carries a ceiling. The 5 mg/kg rule is validated only to a BMI of 35. Above that, no dosing formula predicts a safe level, and a 400 mg/day maximum applies unless medical need justifies more with careful monitoring.
• Two new risk factors: starting hydroxychloroquine at an older age, and female sex.
• Cumulative dose is declined as a risk indicator. The 1000 g threshold from the pre-2016 guideline is not carried forward, because patients use different doses depending on weight and other factors.

The revision also narrows its scope. Where the 2016 document covered chloroquine and hydroxychloroquine together, the 2025 revision applies primarily to hydroxychloroquine, with chloroquine addressed in a single short section.

Dosing: 5 mg/kg real weight, and the BMI 35 ceiling

The recommended daily dosage remains ≤ 5.0 mg/kg of real body weight, unchanged from 2016. In practice, a 400 mg/day dose stays within that threshold only in patients who weigh roughly 80 kg (176 lb) or more. Below that weight, 400 mg/day exceeds the recommendation and is the most common scenario worth raising with the prescriber. Doses near 6.0 mg/kg/day markedly accelerate risk, and patients on twice the recommended level can show toxicity within the first year or two.

What is new is the upper bound. Risk at ≤ 5.0 mg/kg/day has been shown to be relatively stable only up to a BMI of 35, and the guideline states plainly that no dosing formula predicts safe levels at higher BMI. It advises that patients with severe obesity start at a lower calculated dose, with a maximum of 400 mg/day that may be raised for medical need under careful monitoring. The mg/kg calculation alone does not license scaling the dose upward indefinitely with weight, and it may also overdose short obese patients within the validated range.

A common strategy in lupus is to start above 5 mg/kg while disease activity is highest and reduce later. Asked about this directly in published correspondence, the authors of the Annals of Internal Medicine 2023 cohort urged caution: average doses of 5–6 and above 6 mg/kg across the first five years significantly raised the risk of retinopathy later, so the early period is not a free window. They also noted that of the eight patients who developed severe retinopathy, only two had continuously exceeded 5 mg/kg — a reminder that severe cases are not confined to the persistently over-dosed.

Hydroxychloroquine is only manufactured in 200 mg tablets, so intermediate doses are reached by splitting tablets or omitting a tablet on certain days. Systemic levels stabilize over weeks, so variable daily dosing averages out. Blood levels are not a useful safety check — studies show them to be highly variable and unreliable as predictors of retinopathy, useful only when very low as an indicator of non-compliance.

Risk factors for hydroxychloroquine retinopathy

Five of the factors below are listed in Table 1 of the guideline, and any one of them moves a patient from deferred monitoring to annual screening from the start of therapy: dose, renal disease, tamoxifen, macular disease and older age at initiation. Duration and sex raise risk without changing that schedule — duration is what the five-year deferral is measured against, and sex is discussed in the text but not listed in Table 1.

• Daily dose: > 5 mg/kg of real body weight. The single most important modifiable factor.
• Duration of use: risk rises with years of exposure. This drives screening timing rather than per-year risk at an appropriate dose.
• Renal disease: hydroxychloroquine is cleared largely by the kidney, so reduced function raises circulating drug levels. A GFR reduced by 50% roughly doubles the risk of retinopathy, and CKD stage 3 or greater (eGFR below 60) carries an adjusted hazard ratio of 1.95 (95% CI 1.25–3.04).
• Tamoxifen: a retinal toxin in its own right, with possible adverse metabolic synergy. Reported at an odds ratio of 4.59 in the 2014 cohort and an adjusted hazard ratio of 3.43 (95% CI 1.08–10.89) in 2024 — though that second estimate rests on only 18 users and 3 events, so the confidence interval is very wide. Case reports raise similar concern for other aromatase inhibitors.
• Age at initiation: risk rises steadily with age at the start of therapy rather than at a single cutoff. Relative to initiation before 45, adjusted hazard ratios are 2.48 for ages 45–54, 3.82 for 55–64 and 5.68 for 65 and over, with 15-year cumulative incidences of 1.9%, 5.2%, 7.6% and 10.6% across those bands. This is not a reason to withhold the drug, but adherence to screening matters more the later therapy begins.
• Sex: female sex carries an adjusted hazard ratio of 3.83 (95% CI 1.86–7.89) versus male at comparable dose and duration — a 15-year cumulative incidence of 6.1% against 2.6%. This is one of the larger effects in the literature and is easy to overlook, precisely because rheumatic disease populations are roughly 83% female and essentially all published risk figures derive from them. It is not, however, grounds on its own for screening earlier: the guideline does not list it in Table 1, and treating it as a trigger would cancel the deferral for almost every patient.
• Macular disease, qualified in Table 1 of the guideline as counting only if it affects vision or the ability to screen: no specific data confirm higher toxicity risk, but maculopathy interferes with interpreting screening tests. Extensive retinal atrophy or large confluent drusen with associated field loss are a clear contraindication to starting, because early toxicity would be undetectable. Sparse macular drusen with intact overlying photoreceptors and an intact ellipsoid zone should not interfere with screening.

Several commonly assumed factors are not supported. Liver disease shows no demonstrated association despite the liver's role in metabolism (adjusted HR 1.51, 95% CI 0.56–4.11). Neither does diabetes (0.72, 0.31–1.63) or use of CYP2D6, CYP2C8 or CYP3A4 inhibitors (0.67, 0.39–1.13), despite the plausible metabolic mechanism. And no genetic variant has been consistently linked to hydroxychloroquine retinopathy.

Screening tests: OCT and wide-pattern FAF

The 2025 revision names macular OCT and wide-pattern fundus autofluorescence as the primary and objective screening tests. Automated visual fields and multifocal ERG are important adjuncts but are not advised as primary screening tools — the field is subjective and of variable reliability, and the mfERG requires calibrated equipment and experienced personnel that most general practices do not have.

Wide-pattern FAF is the part most practices are not yet doing. Scans should extend beyond the arcades — roughly 30 degrees, and ideally out to 50 to 55 degrees. The first sign of toxicity in either pattern is an area of increased autofluorescence, most often in the inferotemporal quadrant. Dark or black areas indicate more severe damage, where photoreceptors are gone and the RPE is damaged or atrophic. The goal of screening is to identify retinopathy before areas of increased autofluorescence go dark. Comparison against the baseline study matters, since isolated images are not always easy to interpret.

When a visual field is used for confirmation, white SITA testing with pattern deviation plots is recommended over the standard grey scale. A 10-2 has high resolution for parafoveal toxicity, but wider patterns (24-2 or 30-2) are needed to show pericentral damage — and since heritage does not reliably predict which pattern a given patient will show, effective field screening requires both central and peripheral testing. Table 2 of the guideline specifies a 10-2 plus a 24-2 or 30-2, or a combined algorithm such as 24-2C that adds central points to a wider grid and accomplishes both in a single test. The most frequent early defects are superonasal, corresponding to the inferotemporal damage seen on OCT and FAF. Uncertain changes should trigger re-testing within a few months or comparison against objective tests.

Patterns of toxicity by heritage

Heritage tells you which pattern to expect first. It no longer tells you which test to run.

• Parafoveal: White patients of European descent, and also those of South Asian or Middle Eastern descent — India, Pakistan and the Middle East.
• Pericentral: patients of East Asian descent, including both Northeast and Southeast Asia, with initial damage in a more peripheral extramacular distribution near the arcades.
• Black patients appear to have a greater tendency toward pericentral disease than White patients, on evidence drawn from African American cohorts. Limited data from the Indian subcontinent suggest a mix.

The effect sizes are large. Relative to non-Hispanic White patients, the adjusted hazard ratio for a pericentral pattern is 15.02 (95% CI 4.82–46.87) in Asian patients and 5.51 (1.22–24.97) in Black patients, though the latter rests on few cases and awaits replication. In one cohort, 54% of cases in Asian patients were pericentral against 5% in non-Hispanic White patients.

Worth noting separately: Asian patients also showed an elevated risk of retinopathy overall, not merely a different pattern (adjusted HR 1.76, 95% CI 1.10–2.81). Hispanic patients differed significantly from non-Hispanic White patients in neither pattern nor overall risk.

A moderate percentage of patients in every group show a contrary or mixed pattern, which is why the guideline directs that screening examinations look for both. There is no known difference in dose sensitivity or outer retinal layer predilection between the two patterns. Age, sex and renal disease were associated with the parafoveal pattern but not significantly with the pericentral one — likely a function of how few pericentral cases exist to analyse.

OCT findings and the inflection point

Toxicity is not a continuous evolution of damage from day one. The retina remains largely normal for many years at a proper dose, and then reaches an unpredictable point at which outer retinal thinning appears and progresses to obvious retinopathy within just a few years. Recognizing that inflection is the practical purpose of annual screening.

The sequence on OCT:

• Initial thinning of the outer retina, detectable on ETDRS grid thickness values or sequential en face thickness images — often several years before ellipsoid zone change is visible on a B-scan.
• Loss of the interdigitation zone line from focal thinning of the EZ-to-RPE space.
• The “bright line” sign under the fovea — not true hyperreflectivity, but preserved central EZ and IZ signal relative to attenuated adjacent areas.
• Flying saucer or sombrero appearance as parafoveal outer retinal thinning continues.
• Progressive EZ damage: granular fragmentation, then focal loss, then large confluent regions with outer retinal atrophy.
• ELM disruption or loss marks a more advanced stage with considerable risk of progression even after the drug is stopped.
• RPE disruption and atrophy is the latest stage, and the point at which bull's eye maculopathy becomes visible on ophthalmoscopy.

The pericentral pattern is harder to catch because OCT studies are often limited to the central macula, and normal outer retinal thinning makes the marker lines harder to identify with increasing eccentricity. It can usually be identified on wider scans as marked thinning and loss of marker lines beyond the parafovea, confirmed by wide-pattern FAF.

AAO 2025 screening schedule

• Baseline: a full exam soon after hydroxychloroquine is begun, including fundus examination, OCT, and wide-pattern FAF. This is an upgrade from 2016, which required only a fundus exam at baseline with objective testing reserved for suspected abnormalities. The baseline serves four purposes: ruling out lupus retinopathy and other unrecognized disease, identifying contraindications, recognizing pre-existing findings that could mimic toxicity, and establishing a reference for later comparison.
• Years 1–5 with no significant risk factors: deferring annual screening is reasonable — but only “as long as screening will be started reliably after 5 years.”
• Year 5 and beyond: annual screening with OCT and wide-pattern FAF.
• Any significant risk factor: annual screening begins at initiation of therapy.
• More frequently for patients with strong risk factors that could accelerate toxicity, or when ambiguous signs are detected.

Annual is considered sufficient because toxicity develops slowly, leaving time to repeat or add tests when results are suspicious but not definitive. At every visit, check dosage against current weight and ask about changes in systemic status — kidney disease, tamoxifen, or major weight loss.

How likely is retinopathy, really?

The best current figures come from Melles, Jorge, Marmor and colleagues (Annals of Internal Medicine 2023, n = 3,325), a contemporary cohort with guideline-recommended serial screening and central adjudication of every OCT, graded for severity. Fifteen-year cumulative incidence by weight-based dose in the first five years:

Two things stand out. First, a correctly dosed patient faces roughly a 2.7% risk at fifteen years, and about 1% for anything worse than mild — considerably more reassuring than older figures that circulated widely. Of 81 cases in that cohort, 56 were mild, 17 moderate and 8 severe. Overall incidence across all doses was 2.5% at ten years and 8.6% at fifteen.

Second, dose matters enormously: above 6 mg/kg/day the risk is roughly eight times that of a correctly dosed patient. Jorge and colleagues (JAMA Network Open 2024, n = 4,677) put the adjusted hazard ratios at 3.23 for 5–6 mg/kg/day and 5.30 for above 6, relative to 5 mg/kg/day or less, on an independent cohort whose incidences ran 2.1%, 7.0% and 14.0%.

That paper also found the dose-response curve to be close to linear, and its authors argue explicitly against reading 5 mg/kg as a safe threshold: prescribers should use the lowest dose that achieves disease control rather than treating the guideline figure as a line below which nothing happens.

One caveat on all of these figures, raised in published correspondence and conceded by the authors: dose was measured from pharmacy dispensing records, not from what was written on the prescription. Adherence to hydroxychloroquine is often poor, so a patient's dispensed exposure can sit well below their prescribed dose. The 5 mg/kg figure is therefore best read as a limit on the dose actually taken. For a patient you know to be reliably adherent, the prescribed number and the studied number are the same; for one who is not, the calculated mg/kg overstates their true exposure.

Population risk versus risk after a normal exam

The figures above are cumulative population risks. The 2025 revision makes a distinction that changes how they should be used in the chair:

Population risks do not apply once an individual's screening exam shows no evidence of toxicity. What matters for counseling is the risk of toxicity developing in the year after a normal exam, and that figure is always much lower than the cumulative population figure. Even after 20 years of use at a proper dose, the incremental risk in the next year is under 5%.

The corollary is that screening intervals matter more than cumulative risk arithmetic. In the Jorge cohort, the mean interval between a patient's last normal screening study and their first abnormal one was 1.6 years — which is the empirical case for an annual rather than a longer interval, in a single number.

Tests not recommended for screening

The guideline is explicit about what does not belong in a screening protocol:

• Ophthalmoscopy and fundus photography: visible damage implies RPE disruption, which is already an advanced stage.
• OCT angiography: vascular changes have not correlated consistently with toxicity until damage is late and severe.
• Time-domain OCT: resolution is insufficient for early change.
• Fluorescein angiography: reveals only late RPE defects.
• Full-field ERG: abnormal only in very late diffuse toxicity affecting the entire retina.
• Amsler grid: not consistent enough for reliable detection of subtle scotomas.
• Color vision testing: errors occur late and are neither sensitive nor specific.
• Electro-oculogram: not validated as a screening test.

Microperimetry sits outside that list. It is conceptually more reliable than automated perimetry, but long examination times, patient fatigue, and sparse availability have kept it from proving more revealing in practice, and different test patterns are needed for European and East Asian eyes.

Chloroquine

Chloroquine is presumed to share a mechanism with hydroxychloroquine and may be more toxic, though no recent studies confirm it. Older literature equated 3.0 mg of chloroquine with 6.5 mg of hydroxychloroquine, which makes 2.3 mg/kg/day of chloroquine equivalent to 5.0 mg/kg/day of hydroxychloroquine — and that is the recommended ceiling. The practical consequence is worth knowing: the smallest chloroquine tablet is 250 mg, so one tablet daily is an overdose for anyone weighing less than 108 kg (238 lb). Dose is adjusted the same way as hydroxychloroquine, by splitting tablets or omitting them on certain days.

Progression after stopping, and the decision to stop

Whether retinopathy progresses after discontinuation depends entirely on how early it was caught:

• Mild OCT change with an intact ELM: typically only mild, limited progression for a year or so, then stabilization, long before foveal damage. Minimal if any further risk to vision.
• Extensive EZ loss or ELM disruption: retinopathy may continue to worsen for years despite cessation.
• RPE involvement and visible bull's eye: progression can be relentless, advancing to foveal thinning and eventual loss of central acuity.

Some recent data suggest that halving the daily dose may halt the acute progression of thinning and EZ loss, but the duration of that effect is unknown and long-term data are not available. No diet or medical therapy has been shown to prevent or treat the retinopathy — cessation or dose reduction are the only proven measures.

The decision to stop belongs to the patient and the prescribing physician, not the eye care provider alone. Patients with rheumatoid arthritis often have effective alternatives and will choose to stop. Patients with active lupus may not, and some will elect to continue under close observation, possibly at a lower dose, until signs are more definitive. As detection gets earlier, that conversation gets harder and more important.

How other national guidelines differ

The AAO recommendations are the most widely cited, but they are not the only ones, and the differences are substantive rather than cosmetic. If you practise outside the United States, check which applies to you.

• Royal College of Ophthalmologists (UK), 2020, reviewed 2023. Reached the OCT-plus-widefield-FAF position five years before the AAO did, and goes further in two respects. The visual field is conditional, performed only when SD-OCT or FAF is abnormal, where the AAO makes it confirmatory but does not gate it. And the RCOphth does not recommend a baseline examination at all, citing UK audit data showing under 4% discontinuation — a direct contradiction of the AAO 2025 position that a full baseline is strongly advised. It also offers the tidiest mnemonic in the field, the “5 × 5 rule”: keep the dose under 5 mg/kg/day, monitor after 5 years.
• RANZCO (Australia and New Zealand), 2025. Still built on the 2016 framework. It retains the 1000 g cumulative dose as a high-risk criterion, lists dilated fundus examination, 10-2 visual field and SD-OCT as the minimum battery with FAF and mfERG as adjuncts, and selects field patterns by ethnicity. A 2025-dated document that the AAO revision has now overtaken.

Where they agree is more striking than where they differ: every current guideline uses 5 mg/kg of real body weight, and every one treats renal impairment, tamoxifen and pre-existing macular disease as grounds for earlier monitoring.

Clinical pearls

• Bull's eye maculopathy should no longer be encountered. The guideline states this directly: if dosing recommendations are followed and regular screening is performed, this late finding should not appear. Seeing one is a failure of the screening pathway, not a diagnostic success.
• A normal visual field does not rule out toxicity — and under the 2025 revision it was never the right primary test. OCT thinning precedes definitive field, FAF, and mfERG change.
• Screen both patterns in every patient. Heritage shifts the prior probability; it does not narrow the protocol.
• Recheck mg/kg at every visit. Weight changes, and so does renal function. A patient dosed correctly three years ago may not be dosed correctly today.
• Corneal verticillata are not a marker of retinopathy. More common with chloroquine, unrelated to retinal damage, and usually reversible.
• Most patients with early toxicity are asymptomatic. Central acuity is retained until severe, late damage in either pattern. A few notice paracentral scotomas while reading.
• Compliance is the weak link. Studies consistently show that a significant subset of patients exceed the recommended dose or miss baseline and annual screening entirely. Sharing the dosing rationale with the patient and the prescriber is part of the job.
• AI and advanced imaging are not ready. Adaptive optics, automated layer segmentation, en face EZ mapping, and FLIO all show promise, but none are validated or widely available. AI analysis of OCT is expected to become clinical within several years.