Drs. Munster and Gibbs contributed equally to this work.
Hydroxychloroquine concentration–response relationships in patients with rheumatoid arthritis
Article first published online: 6 JUN 2002
Copyright © 2002 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 46, Issue 6, pages 1460–1469, June 2002
How to Cite
Munster, T., Gibbs, J. P., Shen, D., Baethge, B. A., Botstein, G. R., Caldwell, J., Dietz, F., Ettlinger, R., Golden, H. E., Lindsley, H., McLaughlin, G. E., Moreland, L. W., Roberts, W. N., Rooney, T. W., Rothschild, B., Sack, M., Sebba, A. I., Weisman, M., Welch, K. E., Yocum, D. and Furst, D. E. (2002), Hydroxychloroquine concentration–response relationships in patients with rheumatoid arthritis. Arthritis & Rheumatism, 46: 1460–1469. doi: 10.1002/art.10307
- Issue published online: 6 JUN 2002
- Article first published online: 6 JUN 2002
- Manuscript Accepted: 7 FEB 2002
- Manuscript Received: 1 SEP 2000
- NIAID. Grant Number: AI34361
- Sanofi-Synthelabo, Inc.
- Rasmuson Center for Arthritis, Orthopaedics and Musculoskeletal Diseases
A dose–response relationship for hydroxychloroquine (HCQ), in terms of the proportion of patients achieving the Paulus 20% criteria for improvement, had previously been observed in patients with rheumatoid arthritis (RA) receiving a 6-week loading regimen of 400, 800, or 1,200 mg HCQ daily. This present retrospective analysis was performed to investigate possible relationships between the blood HCQ and HCQ-metabolite concentrations and measures of efficacy and toxicity. In addition, we sought to ascertain whether further investigation of HCQ/HCQ-metabolite levels might lead to testing of one of these substances as a new antirheumatic drug.
Patients with active RA (n = 212) began a 6-week, double-blind trial comparing 3 different doses of HCQ at 400, 800, or 1,200 mg/day, followed by 18 weeks of open-label HCQ treatment at 400 mg/day. Patients were repeatedly evaluated for treatment efficacy and toxicity. Blood samples were available from 123 patients for analysis of HCQ, desethylhydroxychloroquine (DHCQ), desethylchloroquine (DCQ), and bisdesethylchloroquine (BDCQ) levels using high-performance liquid chromatography. Achievement of the modified Paulus 20% improvement criteria for response in RA was used as the primary efficacy parameter. Spontaneously reported adverse events were categorized and analyzed as toxicity outcome variables. The relationship between response (efficacy and toxicity) and drug levels was evaluated using logistic regression analysis.
The subset of patients with blood concentration data was equivalent to the larger study population in all demographic and outcome characteristics. The mean HCQ, DHCQ, and DCQ elimination half-lives were 123, 161, and 180 hours, respectively. There was a positive correlation between the Paulus 20% improvement criteria response and blood DHCQ concentrations during weeks 1–6 (P < 0.001). A potential relationship between ocular adverse events and BDCQ levels was found (P = 0.036). Logistic regression analysis of adverse events data showed that adverse gastrointestinal events were associated with higher HCQ levels (P = 0.001–0.021) during weeks 1, 2, and 3.
There is a weak, but predictable, relationship between blood DHCQ concentrations and efficacy of treatment with HCQ. In addition, there is an association between gastrointestinal adverse events and elevated blood HCQ concentrations. Further investigation of these relationships is warranted to see if DHCQ may be introduced as a new antirheumatic drug.
Hydroxychloroquine (HCQ) is generally regarded as a safe and reasonably effective treatment for patients with rheumatoid arthritis (RA), with a recommended daily dose of <6.4 mg/kg/day and a maximum dose of 400 mg/day. Approximately 70% of patients receiving HCQ have clinical improvement in their disease (1, 2). However, the onset of action after initiation of HCQ therapy is slow, requiring 3–6 months for full efficacy to be achieved (3, 4). It was hypothesized that dose-loading over a short period of time may quickly increase HCQ concentrations to effective levels in the tissues, and thereby shorten the time to onset of effect.
To evaluate the feasibility of a loading regimen, a 24-week, double-blind study was conducted with either 400, 800, or 1,200 mg/day of HCQ being given during the first 6 weeks, followed by 18 weeks of open-label treatment with 400 mg HCQ daily (5). A statistically significant dose–response relationship was evident at the sixth week of HCQ therapy, suggesting that a blood concentration–response relationship might also exist. Several studies have examined the relationship between blood, serum, or plasma HCQ concentration and treatment effect (2, 6, 7). Unfortunately, these earlier studies all have the drawback of presenting data on only a small number of patients whose disease progression varied and who had previously received drug treatment. In addition, the possible contribution of potentially active metabolites of HCQ has not been assessed.
The most feared complication of HCQ treatment is retinopathy (8, 9), whereas gastrointestinal (GI) side effects are the most frequent problem reported by patients taking HCQ. This latter complication limits the use of HCQ in some individuals.
The purpose of this study was to investigate whether the efficacy and toxicity of HCQ correlate with the blood concentrations of HCQ and its oxidative metabolites, including desethylhydroxychloroquine (DHCQ), desethylchloroquine (DCQ), and bisdesethylchloroquine (BDCQ), in patients with well-defined disease severity and concomitant drug treatment. For adverse events, we focused on ophthalmic and GI side effects.
PATIENTS AND METHODS
This analysis was performed on a subset of patients from an earlier clinical trial (5). We have utilized blood concentration data collected prospectively (and analyzed retrospectively) in all 3 phases of the trial. In phase 1, patients discontinued treatment with nonsteroidal antiinflammatory drugs (NSAIDs) and began taking naproxen when flare-ups occurred. During phase 2, arthritis response and adverse events were measured in a double-blind study using 400, 800, or 1,200 mg HCQ daily for 6 weeks. Phase 3 was an open-label 18-week continuation study during which all patients received 400 mg HCQ daily.
The clinical aspects of this study were previously published (5). Two hundred twelve patients who were ≥18 years of age and who met the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) criteria for RA (10) were included, and relevant blood samples were available from 123 of these patients (see Results for comparison of this subset with the group as a whole; no differences were found) (5). Inclusion criteria of the study included carefully defined active disease, predefined flare-ups within 2 weeks of withdrawing baseline treatment with the NSAID, age ≥18 years, predefined limitations on the use of disease-modifying antirheumatic drugs (DMARDs), no concurrent use of DMARDs, washout of any previous DMARDs, and minimal (if any) previous use of HCQ; a previously stable regimen of ≤10 mg/day prednisone was allowed and treatment with 1,000 mg/day naproxen was required. Exclusion criteria included those characteristics listed in the 1958 exclusion criteria of the ACR (11). Patients were also excluded if they were pregnant or had active peptic ulcer disease or active inflammatory GI disease, levels of alanine or aspartate aminotransferase higher than twice the upper limit of normal, creatinine levels >1.5 mg/dl or creatinine clearance <50 ml/minute, a history of psoriasis or porphyria, known changes in visual field measurement or only 1 functional eye, cataracts interfering with retinal examination, a platelet count <100,000/mm3, white blood cell count <3,500/mm3, active alcoholism within the previous 2 years, significant malignancy within the previous 5 years, disease disability categorized as ACR functional class IV, a disease duration >5 years, other concomitant or recent investigational drug use, prior or current use of intraarticular or intramuscular steroids, prior or current use of intramuscular gold salts, a weight <45 kg, or previous failure to respond to naproxen therapy. Although no intraarticular steroids or other injections were allowed during phases 1 or 2, a single injection of 1 joint was allowed during phase 3.
Study design and treatment regimens.
A total of 18 investigators enrolled patients in the protocol. The study was completely described previously and was a multicenter, double-blind, parallel, randomized trial evaluating 400, 800, or 1,200 mg HCQ daily in the treatment of RA (5).
Briefly, prior to administering the HCQ, previous NSAIDs were washed out, the disease was allowed to flare up briefly, and 1,000 mg/day naproxen was then instituted. After stabilization was achieved, a double-blind, 6-week clinical and pharmacokinetic trial comparing daily doses of 400 mg, 800 mg, and 1,200 mg HCQ ensued. Thereafter, 400 mg HCQ daily was given to all patients and they were followed up systematically to 24 weeks (5).
Standardized, validated estimates of disease activity were calculated regularly (14 times) and included the tender joint count (66 joints were measured), swollen joint count (64 joints were assessed), duration of morning stiffness, global assessment of overall disease activity by physician and patient using a 10-cm visual analog scale (VAS), and patient evaluation of pain at time of visit using a 10-cm VAS. Laboratory tests of efficacy included the erythrocyte sedimentation rate (ESR) and C-reactive protein level, which were also obtained regularly (5).
As outlined in the clinical study, investigators regularly performed ophthalmologic examinations, Amsler grid evaluations, and laboratory examinations, and questions relating to adverse events were addressed (5). Chest radiograph, electrocardiogram, G6P dehydrogenase, and porphyria screenings were done at baseline. Pregnancy tests were done at screening, baseline, and as appropriate.
HCQ, DHCQ, DCQ, BDCQ, and chloroquine were kindly donated by Sanofi-Winthrop (New York, NY). Water, methanol, ethanol, and ether were high-performance liquid chromatography (HPLC) grade. All glassware coming into contact with HCQ and its metabolites was silanized using AquaSil (Pierce, Rockford, IL).
Collection and storage of blood samples.
Blood samples (10 ml) were collected in silanized vacutainers containing 125 units of heparin and stored at − 20°C. Blood samples were collected (within 1–2 hours of administering the doses of HCQ) during routine clinic visits, which occurred 1–6, 8, 10, 16, 20, and 24 weeks after the initiation of HCQ therapy. At the baseline, 6-week, and 24-week visits, multiple samples were collected over 24 hours (Furst DE, et al: unpublished observations).
HPLC analysis of blood samples.
We measured concentrations of HCQ and its metabolites in the blood, rather than in the plasma. Whole blood assay allows for a greater test sensitivity, due to the extensive accumulation of the drug in the cells and because of the technical difficulty of reproducibly separating blood cells and platelets from plasma (12, 13). Distilled water (1.0 ml), internal standard (200 μg chloroquine), and 150 μl of 10N NaOH were added to the whole blood samples (1.0 ml). Alkalinized blood was shaken with 3.0 ml of ether and centrifuged for 10 minutes at 2,000g. The aqueous phase was frozen and the ether layer was transferred to an evaporation tube. Ether extraction was repeated and the combined extracts were evaporated to dryness. The residue was reconstituted in 500 μl of mobile phase, which was then filtered using Gelman 0.45μ polyvinylidene difluoride syringe filters and placed in capped polypropylene autosample vials, pending analysis.
HCQ, DHCQ, BDCQ, and DCQ in the blood extracts were analyzed by HPLC according to the method of Tett et al (14) using a Waters 510 isocratic HPLC pump, Thermo-Separation Products AS-1000 autosampler, Waters 470 fluorescence detector, and a Gilson FC204 fraction collector. The internal standard for the HPLC assay was chloroquine. Fluorescence was monitored at an excitation wavelength of 337 nm and emission wavelength of 405 nm. Chromatograms were recorded and analyzed using the PE Nelson (Cupertino, CA) Turbochrom (version 4) computer data processing system. Two standard curves were constructed, covering low (50–500 ng/ml) and high (100–2,000 ng/ml) drug concentration ranges. Initially, samples were analyzed using the high-range standards. Samples with concentrations <500 ng/ml were reanalyzed using the low-range standard curve.
The analytes were separated on a polystyrene divinylbenzene stationary phase column (150 × 4.1 mm with 5 μm particle size). The mobile phase consisted of 70% methanol, 30% water, and triethylamine (2% volume/volume). The pH was adjusted to pH 9 with 6N HCl. The mobile phase was recycled in a closed system under helium pressure to prevent the evaporation of triethylamine.
The limit of detection for HCQ, DHCQ, BDCQ, and DCQ was 1.0 ng/ml, with a limit of quantitation of 5 ng/ml except for BDCQ, which had a limit of 10 ng/ml. The inter- and intraday coefficients of variation for all compounds were not more than 8.4% and 4.0%, respectively.
Statistical and kinetic analysis.
The relationships between drug and metabolite concentrations and either efficacy or adverse events were assessed by logistic regression analysis. For the dependent variable in the efficacy analysis, we calculated the proportion of patients achieving the modified Paulus criteria, a composite index of response (15), at the 20% improvement level. To meet the requirements for achieving this predefined level of response, patients must have improved in 4 of the following 6 criteria: 20% improvement in either the tender joint count or swollen joint count, the Westergren ESR, or morning stiffness, and at least a 40% improvement in patient or physician assessments of global disease activity. Baseline values of individual patient variables were calculated from the mean of the values from the screening and baseline visits.
Spontaneously reported adverse events were the primary toxicity outcome variables. Adverse events were categorized by body system, comprising the GI, neurologic, ophthalmic, genitourinary, allergic, cardiovascular, musculoskeletal, infectious disease, psychological, and hematologic systems. Individual patient response was categorized as either having (1) or not having (0) an adverse event within a single body system at each clinical visit.
Logistic regression analysis was performed using SPSS (version 8.0; SPSS, Chicago, IL) to assess the relationship of efficacy or toxicity measures to the pooled blood concentrations of HCQ, DHCQ, BDCQ, and DCQ from weeks 5 and 6, which was at the end of the dose-loading phase of the trial when blood concentrations of parent drug and metabolites had reached a plateau. A forward conditional approach was used in the selection of predictor variables. For the univariate logistic regression analysis, we set a conservative level of significance at P < 0.01 to account for multiple testing. Preliminary analysis indicated that patient demographic variables, including sex, race, weight, age, and duration of disease, did not influence HCQ efficacy or toxicity. Therefore, these variables were not included in the multiple logistic regression analysis. Pharmacokinetic analysis was done using SPSS.
The subset population utilized for this analysis, which was chosen because they were the patients who had blood samples available at all relevant time points, was similar in all patient demographic and outcome characteristics to those reported for the entire trial population (data not shown) (5). The proportion of patients in this subset achieving the modified Paulus 20% improvement criteria at week 6 was 48% (n = 42), 60% (n = 40), and 64% (n = 36) of the patients receiving 400, 800, and 1,200 mg HCQ daily, respectively (as reported in the original publication ). By comparison, in the whole trial analysis, the proportion of patients achieving the Paulus 20% improvement criteria at week 6 was 48%, 58%, and 64% for patients receiving 400, 800, and 1,200 mg HCQ daily, respectively (5). As indicated in the original report, other response measures (i.e., tender joint count, swollen joint count, ESR, morning stiffness, patient and physician global evaluations) were not dose-dependent at week 6.
In the present subset of RA patients (n = 123), the mean clearance rate (clearance of HCQ divided by bioavailability of HCQ) was 15.0 liters per hour (SD 4.41). The mean volume of distribution of HCQ at steady-state was 2,283 liters (SD 1,019). The mean elimination half-life of HCQ was 123.4 hours (SD 44.9). The mean elimination half-lives of DCQ and DHCQ were 180 hours (SD 63) and 161 hours (SD 41), respectively. At steady-state, the mean ratio of HCQ to DCQ was 7.20 (SD 1.88), while that of HCQ to DHCQ was 1.75 (SD 0.37). The concentrations of BDCQ were so low that they were usually at or below the sensitivity of the measurement, and so the ultimate variability of the BDCQ concentrations did not allow reliable pharmacokinetic estimates, although correlations between concentration and effect were still possible to calculate.
The steady-state relationship between blood HCQ, DHCQ, DCQ, and BDCQ concentrations and the HCQ dose (expressed relative to body weight) was examined using blood concentration data obtained during weeks 5 and 6. For illustrative purposes, the relationships between blood HCQ and DHCQ concentrations and HCQ dose (expressed relative to body weight) are shown in Figures 1A and B, respectively. The highest coefficient of variation was observed between HCQ dose (in mg/kg) and the HCQ blood concentration (in ng/ml) (r2 = 0.45). Lower coefficients of variation were found between HCQ dose and the metabolite concentrations (r2 range = 0.04–0.16). Correlations between HCQ blood concentrations and all 3 metabolite concentrations were also sought; there were statistically significant coefficients of variation between HCQ blood concentrations and concentrations of all 3 metabolites (P < 0.002 for all cases; data not shown).
The mean blood concentration–time profile for HCQ during the 24-week study is shown in Figure 2A. Due to the large interindividual variation (coefficients of variation ≥46%) in the blood concentration of HCQ and its metabolites across dose groups, the mean blood concentration for each dose group is presented without error bars. There was extensive overlapping of blood HCQ concentrations across dose groups. Blood HCQ concentrations gradually increased to a plateau around week 5. After week 6, there was a sharp decline in blood HCQ concentrations for patients receiving 800 and 1,200 mg HCQ daily, when their doses were changed to 400 mg daily. Patients receiving 1,200 mg HCQ daily appeared to require more time to achieve a steady-state blood concentration than did patients receiving 400 and 800 mg HCQ. HCQ concentrations in the 1,200 mg group also took longer than those in the 800 mg group to decline to the same levels as those in the 400 mg group after all patients were changed to 400 mg after week 6. This may indicate a subtle dose-dependency in the pharmacokinetics of HCQ.
Figure 2B shows blood DHCQ concentrations in patients receiving 400, 800, and 1,200 mg HCQ daily. DHCQ concentrations rose to a plateau over the first 3 weeks of HCQ administration. After week 6, blood DHCQ concentrations in patients receiving 800 and 1,200 mg HCQ daily declined to levels similar to those in patients receiving 400 mg HCQ daily.
The data to calculate the ACR 20% improvement response were not available for this study, and therefore the Paulus criteria were those taken as the primary outcome measure for this study. The proportion of patients achieving the Paulus 20% improvement criteria during the 24-week study is shown in Figure 3. In general, for all dose groups, the proportion of patients achieving the Paulus 20% improvement criteria increased with time during weeks 1–5 and appeared to plateau after the fifth week, when 50–70% of patients achieved the Paulus 20% improvement criteria. After 6 weeks, the proportion of patients achieving the Paulus improvement criteria in the 400 mg HCQ dose group steadily rose to 70% by the twenty-fourth week. It should be noted that clinical data beyond week 6 were derived from the open-label extension treatment period, and therefore must be viewed in that context with great caution. Blood concentration–effect analysis was confined to weeks 1–6 (during double-blind drug administration), when all patients were analyzed for their response to treatment by the Paulus 20% improvement criteria.
We examined the relationship between the proportion of patients achieving the Paulus 20% improvement criteria and the blood HCQ, DHCQ, DCQ, and BDCQ concentrations at steady-state, i.e., during weeks 5 and 6. In separate univariate logistic regression analyses, blood DHCQ (P < 0.001) and DCQ (P = 0.030) concentrations were significantly correlated with the Paulus improvement criteria (Table 1), and blood HCQ (P = 0.027 and 0.021) concentrations approached statistical significance (a P value less than 0.01 was required for statistical significance, to account for multiple comparisons). Utilizing a forward conditional approach that assumes a hierarchy of predictors based on univariate analysis, the blood DHCQ was the only variable to enter the regression model (P < 0.001). Further addition of other drug metabolite concentrations did not improve the regression prediction (P values for variables not in the equation, i.e., HCQ, DCQ, and BDCQ, ranged from 0.086 to 0.955). Figure 4 shows the relationship between the observed proportion of patients achieving the Paulus 20% improvement criteria and blood DHCQ concentration at incremental intervals of 250 ng/ml.
|Regression model, week||No. of patients meeting criteria||Step 1||Step 2|
In the earlier clinical trial, it was found that although adverse events were frequent, there was little evidence of an HCQ dose–adverse event relationship. However, it was found that during weeks 0–6, there was a greater incidence of nausea, vomiting, and abdominal pain in the 800 mg and 1,200 mg dose groups as compared with the 400 mg dose group. A further analysis was performed in the present study to determine if a relationship exists between adverse events and the steady-state concentration, using pooled data from weeks 5 and 6 (Table 2). Adverse events other than GI complications (i.e., neurologic, ophthalmic, genitourinary, allergic, cardiovascular, musculoskeletal, infectious disease, psychological, and hematologic systems) were not related at the P < 0.01 significance level to the HCQ or metabolite blood concentrations (DHCQ, DCQ, or BDCQ) during weeks 5 and 6 (P = 0.10–0.89). Two of the 40 tested relationships approached statistical significance. During weeks 3–24, neurologic side effects were associated with higher blood HCQ concentrations (P for individual weeks = 0.012–0.060, P for trend = 0.03). Musculoskeletal adverse reactions were related to a lower blood DCQ concentration (P = 0.05) (data not shown). Since the appropriate statistical alpha level was < 0.01, and because 2 of the 40 repeated tests would have approached 0.05 by chance alone, these latter 2 findings are not considered further.
|Type of event, week||No. of patients with event||Step 1||Step 2|
Given the findings of dose-dependent GI side effects during weeks 0–6 in the previous study, we examined the relationship between GI adverse events (principally nausea, vomiting, dyspepsia, abdominal pain, or diarrhea; also, 3 cases of GI bleeding/ulcers) and blood concentrations of HCQ, DHCQ, DCQ, and BDCQ during visits 1–6 (screening, flare-up, baseline, and treatment weeks 1, 2, and 3). The proportion of patients reporting GI adverse events correlated with the blood HCQ concentration during weeks 1 (visit 4) (P = 0.005) and 2 (visit 5) (P = 0.001), and approached significance at week 3 (visit 6) (P = 0.021). In a pooled univariate analysis of data from weeks 1–3 (visits 4–6), GI adverse events were correlated with higher blood HCQ concentrations at those times (P < 0.001), and were not related to blood DHCQ, DCQ, or BDCQ concentrations (P values ranged from 0.11 to 0.30) (data not shown). Figure 5 shows the relationship between the proportion of patients reporting GI adverse events and the blood HCQ concentration during weeks 1, 2, and 3.
We examined the relationship between ophthalmic adverse events and blood concentrations of HCQ and its oxidative metabolites using pooled data from weeks 8–24. The relationship between eye problems and blood HCQ, DHCQ, DCQ, and BDCQ concentrations was assessed using forward conditional logistic regression. There was a trend toward higher blood BDCQ concentrations in patients reporting eye problems compared with those not reporting ophthalmic adverse events (P = 0.036). Although a compilation of various eye problems may bear no specific relationship to the major, and rarer, complication of retinopathy, it can still be worrisome to the patient. No clinical loss of vision was noted. The specific ophthalmic adverse events noted and ascribed to HCQ therapy during this study were as follows: visual field abnormality (n = 5), color vision abnormality (n = 5), Amsler grid abnormality (n = 2), foveal reflex abnormality (n = 3), increased pigment (n = 5), disc abnormality (n = 1), macular abnormality (n = 4), and decreased visual acuity (n = 1) (>1 event could occur in a given individual). Each of these potential toxic reactions has been documented for HCQ (16–20). Altogether, 26 ocular abnormalities occurred in 17 patients. Other events that were unlikely to be associated with HCQ toxicity, such as “red eye” or “conjunctivitis,” were not included in the correlation analysis. Correlations of blood HCQ, DHCQ, and DCQ concentrations to ophthalmic adverse events did not approach statistical significance (P values ranged from 0.19 to 0.97).
Several published studies have addressed the question of whether a concentration–effect relationship exists for HCQ, and the results of the studies have been equivocal. Tett et al examined the relationship between blood HCQ concentration and several measures of efficacy in 43 RA patients receiving HCQ for 6 months (7). They found that RA patients who had no or mild (<0.5 hours) morning stiffness or were negative for rheumatoid factor had higher HCQ concentrations than did those with more active disease (P < 0.05). Other measures of disease activity and an unweighted summed score of disease activity were unrelated to blood HCQ concentrations. Studies that measured serum or plasma HCQ concentrations were not able to detect a concentration–effect relationship between clinical outcomes and HCQ concentration (2, 6). In our study, the ability to discern a relationship between effect and drug concentration was enhanced by the larger number of patients studied, who were relatively homogeneous and had a similar history of background medications, and by the inclusion of metabolite measurements. In addition, the use of 3 different HCQ doses produced a wide range of blood HCQ and metabolite concentrations, which may have increased our ability to discern concentration–effect relationships.
The major finding of this study was that a relationship exists between efficacy and the blood DHCQ concentration. By limiting our correlation analysis to weeks 5 and 6, at a time when plateau drug concentrations were reached, when the study was double-blind, and where a dose–response relationship could be demonstrated, we found that the proportion of patients achieving the Paulus 20% improvement criteria was related to the blood concentrations of DHCQ, DCQ, and BDCQ when each was entered as a single predictor variable. When the logistic regression was performed allowing for multiple variables, blood DHCQ concentration alone was identified to be a significant predictor. The addition of other blood concentration variables did not improve the regression fit. Significant coefficients of variation (r2) were observed between the blood DHCQ (r2 = 0.370), DCQ (r2 = 0.557), and BDCQ (r2 = 0.123) concentrations and concentration of parent HCQ during weeks 5 and 6 (P < 0.002 in all cases). It should be noted that the deletion of other concentration variables does not imply that the blood concentrations of HCQ or other metabolites might not contribute, to some extent, to the therapeutic effects of HCQ. Given the colinearity between HCQ and metabolite concentrations, it may simply mean that the contribution of the other active species is overshadowed by the blood DHCQ concentrations.
Although the statistical correlation approaches significance, the degree of correlation between drug concentration and efficacy/toxicity for antirheumatic drugs is often not very strong. The serum concentration–to–response relationship for NSAIDs such as naproxen, carprofen, and ibuprofen is of the same order as that observed for DHCQ (21, 22). In fact, the relationship found in our study, although weak, represents only the second serum concentration–to–response relationship found for a non-NSAID antirheumatic treatment. The other relationship was for minocycline, a drug which has been considered controversial as an antirheumatic treatment (23). Although a dose–response relationship has been found for auranofin, gold sodium thiomalate, and methotrexate, no serum-level response has been found for these drugs (24–29). Many other variables, such as tissue concentrations, receptor concentrations, understanding of drug mechanisms, detailed understanding of environmental influences (e.g., activity level, smoking, alcohol intake), genomic differences, sex, and age, probably have effects that could confound our present results.
A higher incidence of GI side effects was associated with an HCQ dose of 800 mg and 1,200 mg in the dose-loading phase of the earlier trial (5). There was no relationship between the incidence of GI side effects and blood concentrations of HCQ or its oxidative metabolites at weeks 5 and 6 (as outlined in the Results). However, in a similar analysis of pooled data from weeks 1, 2, and 3, we found that blood HCQ concentrations were positively associated with an increased risk of adverse GI reactions, including nausea, vomiting, and GI irritation at weeks 5 and 6.
The high volume of distribution of HCQ and its metabolites indicates that the drug accumulates, to a large degree, in the tissues. The concentrations of HCQ in the blood at weeks 1, 2, and 3 combined with the large volume of distribution of HCQ implies that significant, although slow, tissue accumulation has occurred. It is possible that a certain accumulation of the drug in the tissues may be needed to result in clinical symptoms, including those in the GI tract. The increased amount of time for this accumulation to occur may account for the delay between increasing HCQ blood concentrations in the first 3 weeks and the appearance of clinical effects several weeks later. These correlations are relatively low and mean that HCQ alone probably accounts for only a small proportion of the GI toxic events in this RA population as a whole. However, since GI problems are known to occur in RA patients, even a small overall effect may denote a very significant difference in some patients, especially because the correlation in this case was determined in a relatively small number of patients (as noted in the original dose–response report). It may be that patients develop GI side effects early and either drop out or continue to be treated and develop tolerance to the GI side effects of HCQ after 3 weeks of daily treatment.
Because the retinopathy associated with HCQ treatment is slow to develop and infrequent, it is not surprising that we could not detect a relationship between adverse effects in the eye, potentially related to HCQ's mechanism of action, and the blood HCQ and metabolite concentrations during the loading phase (at weeks 5 and 6). Therefore, we expanded our analysis of ophthalmic adverse events to include pooled data from weeks 8–24. The relationship between eye problems and blood HCQ, DHCQ, DCQ, and BDCQ concentrations was assessed using logistic regression. Although this pooled analysis was confounded by statistical problems (i.e., the problem of repeated measures within individual patients and the low incidence of ophthalmic adverse events), we found a trend toward higher blood BDCQ concentrations in patients reporting the type of ocular adverse effects potentially related to HCQ therapy, compared with those not reporting such adverse events (P = 0.060). Blood HCQ, DHCQ, and DCQ concentration relationships to adverse events did not approach statistical significance (P values ranged from 0.19 to 0.97). The blood BDCQ concentration (mean ± SD) in 5 patients reporting eye problems was 485 ± 452 ng/ml, whereas 85 patients who did not report eye problems had a mean blood BDCQ concentration of 244 ± 219 ng/ml at week 24.
The eye symptoms reported included transient blurred vision or eye signs such as episodes of abnormal Amsler grid changes. For this analysis, we only examined eye signs and symptoms potentially related to HCQ's mechanism(s) of action, such as changes in visual fields, changes in color vision, Amsler grid abnormalities, foveal reflex, pigment, disc changes, macular changes, or changes in visual acuity (16–20). These events do not relate to retinopathy which, thankfully, did not occur. Since retinopathy is quite rare, the lack of such an occurrence is not unexpected. Nonetheless, eye symptoms may be bothersome to the patient, and their potential relationship to the more serious eye changes should lead to further examination of the possible toxic effects of BDCQ on the eye.
The findings that the blood DHCQ concentration is related to treatment efficacy, and that the blood HCQ concentration is associated with GI side effects have important implications. Even using standard treatment regimens (no loading dose), GI side effects can occur in 10–33% of patients at the usual doses of HCQ, and this has been shown to be the cause of drug discontinuation in 12% of patients (30–32). We speculate that individuals who exhibit rapid and extensive metabolism of HCQ to DHCQ may respond more favorably to HCQ. Conversely, an individual with diminished capacity to eliminate HCQ through the formation of DHCQ would be at greater risk of developing GI side effects and would have a less favorable clinical response. We note that the blood concentration ratio of DHCQ to HCQ varied by ∼10-fold, from 0.1 ng/ml to 1.0 ng/ml, during weeks 5 and 6, and this indicates that the disposition of DHCQ in relation to HCQ is highly variable in the population of patients included in this analysis. It is possible that measuring DHCQ:HCQ ratios could help predict those patients who might respond better to HCQ, with fewer toxic reactions, although this requires prospective testing.
This report has described a relationship between the concentration of DHCQ, which is a metabolite of HCQ, and clinical efficacy. At the same time, the data indicate at least an association between HCQ concentrations and GI adverse events. There is also the suggestion that BDCQ might be associated with ophthalmic adverse events. We hope that these results may lead to the testing of DHCQ as a drug with greater efficacy (because it can be used at higher doses) and less toxic effects on the GI and ophthalmic systems than has been shown for HCQ (because these toxicities appear to be related to the parent compound or another metabolite).
We gratefully acknowledge the statistical support provided by Dr. Qing Yao of the Department of Biostatistics, University of Washington, and David Kerr of Axio Research Corporation.
- 4for the Cooperating Clinics Committee. Hydroxychloroquine sulfate in RA: a six-month, double-blind trial. Bull Rheum Dis 1962; 13: 287–90., ,
- 11KlippelJH, WeyandCM, WortmannR, editors. Primer on the rheumatic diseases. 10th ed. Atlanta: Arthritis Foundation; 1993.
- 14High-performance liquid chromatographic assay for hydroxychloroquine and metabolites in blood and plasma, using a stationary phase of poly(styrene divinylbenzene) and a mobile phase at pH 11, with fluorimetric detection. J Chromatogr 1985; 344: 241–8., , .
- 29Gold compounds in rheumatoid arthritis: clinical-pharmacokinetic correlates. J Rheumatol 1979; 5: 51–5..
- 31Chloroquine and hydroxychloroquine in rheumatological therapy. Clin Rheum Dis 1980; 6: 545–66., .