Hydroxychloroquine Blood Levels Predict Hydroxychloroquine Retinopathy

Abstract

Objective

In 2016, the American Academy of Ophthalmology (AAO) changed the recommended dosing of hydroxychloroquine from 6.5 mg/kg to less than 5 mg/kg. However, it is not clear that the lower prescribed dose of hydroxychloroquine will have the same efficacy for SLE activity or the same protective role against cardiovascular risk factors and thrombosis. We addressed the frequency of hydroxychloroquine retinopathy and the role of hydroxychloroquine blood levels to identify those at greater future risk of retinopathy.

Methods

537 SLE patients in a large clinical cohort were repeatedly assessed for hydroxychloroquine blood levels and tested for hydroxychloroquine retinopathy. We assessed the risk of retinopathy by clinical characteristics and levels of hydroxychloroquine in the blood.

Results

The overall frequency of retinopathy was 4.3% (23/537). There was 1% risk of retinopathy in the first 5 years, 1.8% from 6 to 10 years, 3.3% from 11 to 15 years, 11.5% from 16 to 20 years, and 8.0% after 21 years of use. We found that older patients (p<0.0001), higher body mass index (p=0.0160 for trend) and longer duration of hydroxychloroquine intake (p=0.0024 and p=0.0006 for trend) were associated with higher risk of hydroxychloroquine toxicity. Higher hydroxychloroquine blood levels predicted later hydroxychloroquine retinopathy (p=0.0124 and p=0.0340 for mean and maximum HCQ blood levels respectively).

Conclusion

Our data prove the utility of hydroxychloroquine blood levels in predicting retinopathy. This would allow clinicians to either decrease dose or increase monitoring in those with high blood levels.

Introduction

Hydroxychloroquine is one of only four medications approved for use in SLE by the US Food and Drug Administration (1,2). It is the only medication proven to improve survival (2–4). It reduces SLE flares by half (5). Hydroxychloroquine is particularly effective in the treatment of cutaneous disease (6) and arthritis (7). It has antithrombotic (8–10), anti-diabetic (11,12) and lipid lowering effects (13,14). In lupus nephritis, hydroxychloroquine is an independent predictor of complete renal remission in lupus patients treated with mycophenolate mofetil (15). Hydroxychloroquine has been shown to improve SLE pregnancy outcomes (16,17), and to reduce the risk of congenital heart block in women with positive anti-Ro (SSa) antibody (18,19). It has been shown to delay the onset of SLE in those with undifferentiated connective tissue disease (20). All these benefits of hydroxychloroquine were proven using previous dosing regimens (21,22).

The most important long term side effect is hydroxychloroquine retinopathy, involving the photoreceptors with secondary disruption of the retinal pigment epithelium (23), reviewed in Jorge et al (24). The hydroxyl group in hydroxychloroquine reduces its capability of crossing the blood retinal barrier, which explains the lower retinal toxicity compared to chloroquine (25). Risk factors which increase the risk of hydroxychloroquine retinopathy include: daily dose >400 mg, or >6.5 mg/kg ideal/lean body weight for short individuals; cumulative dose >1000 g; duration of use >5 years; renal or hepatic dysfunction; obesity; age >60 years; and pre-existing retinal disease or maculopathy (21).

The American Academy of Ophthalmology (AAO) 2011 recommendations advised hydroxychloroquine weight based dosing of 6.5 mg/kg with a maximum dose of 400 mg/day. Exceptions were individuals of short stature and obese patients, for whom the AAO advised calculating dosage based on ideal body weight (21). The incidence of hydroxychloroquine retinopathy in clinical practice was originally reported in several studies as little or no toxicity among thousands of patients (1,26). In a systematic review of 6 studies including 2043 patients taking hydroxychloroquine for a mean of more than 10 years, only 2 (0.1%) patients were diagnosed with definite retinal toxicity and 6 (0.3%) patients as having probable hydroxychloroquine retinal toxicity (1). A meta-analysis by Yam and Kwok of retinopathy series between 1960 and 2005 showed that hydroxychloroquine retinopathy was rare, with only 12 out of 4415 patients developing retinopathy (27). Wolfe and Marmor, in a study of 3,995 patients with rheumatoid arthritis or SLE taking hydroxychloroquine, found definite or probable toxicity of hydroxychloroquine in 0.65%, but in 1% after 5–7 years of use (28).

However, a later study by Melles and Marmor, a retrospective review of 2361 Kaiser Permanente patients who had used hydroxychloroquine, reported toxicity at 5 years in under 1%, up to 10 years in under 2%, but almost 20% after 20 years (29). This led, in 2016, to revised AAO recommendations that the maximum daily hydroxychloroquine should be less than 5.0 mg/kg real weight, which they thought might correlate better with risk than ideal weight (22).

These dosing recommendations were made without evidence that the lower dose of hydroxychloroquine would have the same efficacy for SLE activity or the same protective role against cardiovascular risk factors and thrombosis. The Kaiser Permanente Study did not analyze prescribed dose, but pharmacy-dispensed medication. While pharmacy-dispensed medication likely gives a better estimate of actual dosing than prescribed dosing, the issue of non-adherence confounds our ability to truly know hydroxychloroquine intake (30). Genetic differences in hydroxychloroquine metabolism might also contribute to our inability to estimate effective dosing (31–33).

Hydroxychloroquine blood levels can be quantified by high performance liquid chromatography (34), which is a widely available means of measuring drug concentrations. It is also possible to measure serum levels, but for reasons of pharmacokinetic stability and reliability, whole blood is preferable (35). Monitoring hydroxychloroquine blood levels has proven to be an effective tool to improve medication adherence in patients with SLE (30). We present the first prospective study of hydroxychloroquine blood levels and hydroxychloroquine retinopathy.

Patients and Methods

SLE patients were enrolled from the Hopkins Lupus Cohort, a prospective study of predictors of flare, atherosclerosis, and health status in SLE. All patients gave informed written consent to participate in the study. Enrolled subjects were followed quarterly or more frequently if clinically necessary. The Hopkins Lupus Cohort was approved yearly by the Johns Hopkins University School of Medicine Institutional Review Board and complied with the Helsinki Declaration. SLE patients met the SLICC SLE Classification Criteria (36).

Beginning in 2013, blood levels of hydroxychloroquine were measured at each visit for cohort patients who had been prescribed hydroxychloroquine. Hydroxychloroquine blood levels were measured by liquid chromatography-tandem mass spectrometry as described by Füzéry et al (34). For patients with retinopathy, all HCQ blood levels measured prior to the diagnosis date were included. For those without retinopathy, HCQ blood levels prior to the final retina assessment were included to calculate the mean and maximum HCQ blood levels.

SLE patients on hydroxychloroquine therapy were examined by retina specialists at the Johns Hopkins Wilmer Eye Institute with a fundus examination and one or more of the newer retinal screening tests: spectral-domain optical coherence tomography (SD-OCT), multifocal electroretinogram (mfERG), microperimetry (MP1) and fundus autofluorescence (FAF).

In our analysis, hydroxychloroquine toxicity was defined dichotomously by the retina specialist: all those with a value of “No” or “Possible” were categorized as not having hydroxychloroquine toxicity, and those who had a “Yes” were categorized as having it. The risk of hydroxychloroquine toxicity was then assessed in tertiles defined by the mean or maximum HCQ blood levels.

Statistical Methods:

The risk of retinopathy was calculated in subgroups defined by demographic and clinical variables, and by tertiles of hydroxychloroquine blood concentration. P-values were calculated using Fisher’s exact test, chi-squared test or Cochran-Armitage test for trend. The intraclass correlation coefficient (ICC) was calculated using the “proc mixed” procedure in SAS 9.4.

Results

Our study included 537 SLE patients with retinal testing, 494 (92%) were female. The majority were Caucasian (46.9%) or African American (41.5%). Figure 1 shows the distribution of hydoxychloroquine dosing in the cohort.

Figure 1 –

Distribution of hydroxychloroquine dosage (mg/kg/day).

Hydroxychloroquine blood levels were measured on 492 of the 537 SLE patients, with between 1 and 25 measurements per patient (median of 7 measurements). Figure 2 shows the relationship between prescribed dosing and hydroxychloroquine blood levels at the same visit. Patients showed a variation in their hydroxychloroquine blood levels over time. The ICC compares the variability of different measurements of hydroxychloroquine blood levels of the same patient to the total variation across all measurements and all subjects. The ICC for hydroxychloroquine blood levels was 0.51. To provide a frame of reference, the ICC of systolic blood pressure and weight were also calculated as 0.47 and 0.94, respectively. Weight was the most consistent between visits within a patient while hydroxychloroquine blood levels showed more variation, similar to that of systolic blood pressure.

Figure 2 –

Hydroxychloroquine dosage (mg/kg/day) versus hydroxychloroquine blood levels (ng/mL).

The overall prevalence of confirmed hydroxychloroquine toxicity was 4.3% (23 patients). Table 1 shows the risk of hydroxychloroquine toxicity by patient demographic and clinical subsets. Men exhibited a higher frequency of toxicity than women (9.3% vs 3.8%, p=0.1029). Caucasians had higher frequency of toxicity (5.6%). Greater age was associated with hydroxychloroquine toxicity. Of patients less than 45 years of age, 0.5% had hydroxychloroquine toxicity, while for those aged 45–59 years and those 60 years and greater the prevalence of hydroxychloroquine toxicity was 4.4% and 10.1%, respectively. Patients with a high body mass index had a higher risk of a hydroxychloroquine toxicity (p=0.0160 for trend test).

Table 1:

Risk of Hydroxychloroquine Toxicity by Demographic and Clinical Variables

	No Toxicity n (%) (n=514)	Toxicity n (%) (n=23)	P-value	P for Trend
Gender			0.1029
Female	475 (96.2%)	19 (3.8%)
Male	39 (90.7%)	4 (9.3%)
Ethnicity			0.3804
Caucasian	238 (94.4%)	14 (5.6%)
African American	215 (96.4%)	8 (3.6%)
Other	61 (98.4%)	1 (1.6%)
Age			<0.0001	<0.0001
< 45 yrs.	215 (99.5%)	1 (0.5%)
45 – 59 yrs.	175 (95.6%)	8 (4.4%)
60+ yrs.	124 (89.9%)	14 (10.1%)
Body Mass Index			0.1701	0.0160
<20	50 (98.0%)	1 (2.0%)
20–25	171 (97.7%)	4 (2.3%)
25–30	159 (95.2%)	8 (4.8%)
30–35	76 (95.0%)	4 (5.0%)
35+	58 (90.6%)	6 (9.4%)
Smoking			0.6463
Ever	154 (95.1%)	8 (4.9%)
Never	357 (96.0%)	15 (4.0%)
Hypertension Ever			0.0020
Yes	276 (93.2%)	20 (6.8%)
No	238 (98.8%)	3 (1.2%)

Table 2 shows the risk of hydroxychloroquine toxicity and duration of hydroxychloroquine intake. The duration of hydroxychloroquine usage ranged from 0 to 48 years, and was categorized into five categories by distribution. Patients with longer duration of hydroxychloroquine intake had more hydroxychloroquine toxicity (p=0.0024 and p=0.0006 for trend) after 16 years of use.

Table 2:

Hydroxychloroquine Toxicity by Duration of Use (Years)

Duration of use (years)	Total	N	Percent	P-value	P-value for Trend
5 or less	103	1	0.97%	0.0024	0.0006
6 to 10	109	2	1.83%
11 to 15	91	3	3.30%
16 to 20	96	11	11.46%
21 or more	75	6	8.00%

Both the mean (p=0.0124) and maximum (p=0.034) hydroxychloroquine blood levels predicted later hydroxychloroquine retinopathy (Table 3).

Table 3 –

Hydroxychloroquine Toxicity in Each Hydroxychloroquine Blood Level Tertile

	No Toxicity n (%)	Toxicity n (%)	P-value	P-value for Trend
HCQ Maximum Tertiles (ng/mL)			0.0340	0.0143
0 – 1182	161 (98.8%)	2 (1.2%)
1183 – 1752	157 (95.2%)	8 (4.8%)
1753 – 6281	153 (93.3%)	11 (6.7%)
HCQ Mean Tertiles (ng/mL)			0.0124	0.0027
0 – 741	162 (98.8%)	2 (1.2%)
741.5 – 1176.5	158 (96.3%)	6 (3.7%)
1177– 3513	151 (92.1%)	13 (7.9%)

There was no difference in the risk of hydroxychloroquine toxicity between Caucasians and African-Americans. Although African-Americans had a significantly lower mean dosage of hydroxychloroquine compared to Caucasians (4.46 vs. 4.84, p=0.0022 by t-test), this did not result in significant differences in blood levels or rate of retinal toxicity (Table 4).

Table 4 –

Comparison of African-American and Caucasian Hydroxychloroquine Blood Levels and Rate of Hydroxychloroquine Retinopathy

	African-American N (%)	Caucasian N (%)	p-value
HCQ blood level mean tertiles			0.1815
0–741	82 (39.2%)	71 (30.9%)
741.5–1176.5	63 (30.2%)	77 (33.5%)
1177+	64 (30.6%)	82 (35.6%)
HCQ blood level maximum tertiles			0.3621
0–1182	74 (35.4%)	73 (31.7%)
1183–1752	66 (31.6%)	66 (28.7%)
1753+	69 (33.0%)	91 (39.6%)
HCQ retinopathy			0.3084
Yes	8 (3.6%)	14 (5.6%)
No	215 (96.4%)	238 (94.4%)

Some patients in our study were screened multiple times with the newer screening tests, often with more than one test at the same visit. A total of 926 retina assessments were performed, up to 8 retina exams per patient with a median of 1. Three hundred had one retina exam, 149 had two exams, and 88 had three or more exams. Whether we looked at the prevalence of hydroxychloroquine retinopathy by number of ophthalmology visits done per patient, or by number of tests done at their last assessment, there was no association between number of exams and toxicity (p-value was not significant).

Discussion

This study represents the first prospective cohort study of hydroxychloroquine blood levels and retinopathy. The prevalence of hydroxychloroquine retinopathy was 4.3% using fundus examination and one (usually Optical Coherence Tomography) or more of the newer retinal screening tests [Spectral-Domain Optical Coherence Tomography (SD-OCT), multifocal Electroretinogram (mf ERG), fundus autofluorescence (FAF) and/or microperimetry (MP1)]. Earlier studies on the prevalence of hydroxychloroquine retinopathy were retrospective and depended on severe toxicity, bull’s-eye maculopathy, when apparent on fundus examination (1,26–28).

The recent large retrospective study by Melles and Marmor reported the overall prevalence of hydroxychloroquine retinopathy was 7.5% evaluated by Spectral-Domain Optical Coherence Tomography (SD-OCT) and visual field testing. The risk of toxicity was under 1% in the first 5 years, under 2% in 10 years, but it increased to almost 20% after 20 years (29). We found that the frequency of hydroxychloroquine retinopathy was less than that, with the overall prevalence of hydroxychloroquine retinopathy was 4.3%. The risk of retinopathy was 1% in the first 5 years, 1.8% from 6 to 10 years, 3.3% from 11 to 15 years, 11.5% from 16 to 20 years, and 8.0% after 21 years of use.

Our data agree that the prevalence of hydroxychloroquine retinopathy, using newer screening technologies, is much higher than previously reported. Importantly, our data show that the risk increases after 16 years of use. Even after 20 years, however, the risk was less than half of that reported in the Kaiser Permanente study (29). This may reflect our local practice, where a dosing limit of 400 mg maximum daily was used, and that reductions were made for renal insufficiency and in the elderly (30).

We found that elderly patients, high body mass index and duration of hydroxychloroquine intake were associated with a higher risk of hydroxychloroquine retinopathy. We found the results on high body mass index particularly concerning, as we capped the dose of hydroxychloroquine at 400 mg daily, no matter how high the body mass index. In accordance with our results, the American Academy of Ophthalmology (AAO) considered the duration of hydroxychloroquine usage and age as risk factors for development of hydroxychloroquine retinopathy (22). However, the AAO went far beyond emphasizing the need to adopt more sensitive screening tests. It recommended a cap at below 5 mg/kg. Figure 2 shows that this cap, in fact, still leads to a broad distribution of hydroxychloroquine blood levels.

The Kaiser-Permanente study found that age and chronic kidney disease were predictors of retinopathy risk (29). We also found age to be a predictor. Many causes of retinopathy, including common ones such as macular degeneration, which is diagnosed after age 55, increase with age. Diabetic retinopathy increases with duration of diabetes (and thus with age). One of the stronger predictors in the Kaiser-Permanente study (29) was chronic kidney disease. As we always reduce the dose in chronic kidney disease (as well as following blood levels), it is not surprising that we have never found an association with chronic kidney disease.

The equal proportion of Caucasians and African-Americans in our cohort allowed us to examine ethnic differences. We did not find any differences between the African-Americans and Caucasians with respect to blood levels or retinal toxicity, despite the lower mean dosing in African-Americans.

Monitoring hydroxychloroquine blood levels is an important step to improve medication adherence in patients with SLE (30). We now introduce the concept of hydroxychloroquine blood level monitoring to reduce over-dosage. Although both the mean and maximum blood levels predict risk, we believe mean blood levels provide a better measure than maximum level to assess risk due to the issues of patient non-adherence (30,37,38) and due to the variation we observed even when the patient is adherent. The correlation between hydroxychloroquine blood levels with skin pigmentation (39) and gastrointestinal side effects has been reported (35). Figure 2 show the large variance between hydroxychloroquine blood levels and prescribed dosage. It should not be surprising that blood levels vary. Genetic polymorphisms are known to affect hydroxychloroquine levels (31–33). However, the demonstrated non-adherence of patients on hydroxychloroquine (30) and changes in adherence over time (37), are the major reason for variations in blood levels. Monitoring serum/blood level targets is routine in other areas of rheumatology (such as monitoring uric acid levels, rather than allopurinol doses, in gout) and in other specialties (such as monitoring anti-epileptic levels, rather than anti-epileptic doses, in neurology). We think our data clearly indicate the benefit of this in SLE management.

We found the prevalence of hydroxychloroquine retinopathy was 4.3% using fundus examination and one or more of the newer retinal screening tests. The risk of retinopathy increased after year 16, but remained much lower than that reported in the retrospective Kaiser Permanente study (29). The risk of retinopathy was numerically higher in men and Caucasians, and statistically higher in older patients and those with higher body mass index. Importantly, no patients had blindness. For the first time, our data show the utility of hydroxychloroquine blood levels in predicting retinopathy risk. Other prospective studies are needed to substantiate our results. The clinical impact of our study is that practitioners would be able to either decrease hydroxychloroquine dose or increase monitoring in patients with the highest tertile of blood levels.