Hydroxychloroquine Compared with Placebo in Rheumatoid Arthritis

Methods

Our 24-week, randomized, double-blind, parallel trial compared hydroxychloroquine, 400 mg/d, with placebo. Hydroxychloroquine was supplied as 400-mg tablets; placebo tablets were identical in shape, taste, and color to the test drug. A total of 126 consecutive patients with rheumatoid arthritis (American Rheumatism Association revised criteria, 1987) [13], attending the outpatient consultation clinic of the Rheumatology Service, Hospital General de Mexico, from June 1989 to August 1991, were enrolled in the study.

Inclusion criteria included 5 or fewer years since diagnosis of disease; age of 18 years or older; onset of rheumatoid arthritis after age 16 years; five or more actively inflamed joints; and unsuccessful treatment with at least two nonsteroidal anti-inflammatory drugs or salicylates. Exclusion criteria included current or previous treatment with second-line drugs or cytotoxic agents; current use of steroids; rheumatoid arthritis functional class IV; and refusal to give informed consent.

Disease activity was clinically assessed at study entry and every 4 weeks for a period of 24 weeks (7 visits) [14]. Patients were examined by the same observer throughout the trial, who was blinded to the knowledge of whether the study drug or placebo was being administered.

Assessments included a joint score (the sum of joint swelling score [58 joints graded on a scale of 0 to 3]) and a pain and tenderness score (60 joints graded on a scale of 0 to 3) [15]. The joint score [16] was designated as the primary outcome before the start of the study. Patients assessed their joint pain using a visual analog scale consisting of a 10-cm horizontal line labeled no pain at the left side and maximum imaginable pain at the right side. Grip strength was assessed using a mercury-column sphygmomanometer with a standard grip bag. Patients' global assessment of their disease activity was graded on a scale of 1 = asymptomatic, 2 = mild, 3 = moderate, 4 = severe, and 5 = very severe. Physicians' assessment of global efficacy was graded on a scale of 1 = marked improvement, 2 = moderate improvement, 3 = the same, 4 = moderate worsening, and 5 = marked worsening.

The hematocrit, leukocyte count, platelet count, and rheumatoid factor for each patient were assessed at entry. The Wintrobe erythrocyte sedimentation rate was tested every 4 weeks.

Every patient was questioned at each visit to determine compliance with prescribed nonsteroidal anti-inflammatory drugs and test medication. In addition, pill counts of the test drug were done.

An ophthalmic examination, including slit-lamp analysis [17], was done by an ophthalmologist at entry and at the end of the study. The Amsler grid [18] was provided to every patient with instructions for daily use.

Statistical Analysis

The sample size was estimated at 52 per group based on an error rate of 0.05; a error rate of 0.2; a delta on the joint score of 7; and a standard deviation of 12. Statistical analysis compared results of treatment on several variables of interest. For joint score, grip strength, morning stiffness, pain assessment, and the individual differences between the last visit and baseline for each patient were computed. These treatment differences were then used to compare grouped drug effects after having adjusted for individual imbalances at baseline. Unpaired t-tests and multiple regression analysis were used to compare the mean values of the treatment differences for each group. The assumption of normality was tested with the Anderson-Darling statistic A² [19], which has been reported as having good power properties. When the goodness-of-fit test rejected the hypothesis of normally distributed data, the Mann-Whitney nonparametric test was used. Chi-square tests were used to compare differences in proportions and frequencies of the clinical variables (physician and patient assessment, functional class, anatomic stage, and physician and patient efficacy). The natural logarithm of rheumatoid factor was used to transform the titers to a normal distribution. The statistical analysis was done using BMDP Statistical Software version PC-1987 (BMDP Statistical Software, Inc.; Los Angeles, California).

Results

A total of 126 patients fulfilled the entry criteria and were admitted to the study, and 101 (80.2%) completed the trial. Twenty-five patients dropped out (19.8%), 14 of the 61 (23.0%) on placebo and 11 of the 65 (16.9%) on hydroxychloroquine. The reasons for withdrawal from the study were as follows: Two patients moved, 1 patient became pregnant during the study, 1 patient developed severe anemia, 1 patient had severe depression, 5 patients did not return because of economic problems, and 15 did not return for unknown reasons. Twenty of these patients were included in the intent-to-treat analysis. Five patients did not return for the second visit and thus had no variables available to analyze, even for the intent-to-treat analysis. The final analysis was based on 121 patients, using the last assessment available. Compliance with taking the study tablets was excellent: Only 95 out of 20 160 tablets were not taken based on pill counts and patient questionnaires.

The two groups did not differ significantly in demographic characteristics, duration of disease, or laboratory test results (Table 1). Table 2 shows the mean changes (with CIs) in the continuous clinical variables (joint score, pain [visual analog scale], and grip strength) between the beginning and end of the trial. All three variables were statistically significant. The primary end pointjoint scoreshowed a clinical and statistical significant improvement favoring the hydroxychloroquine group. The hydroxychloroquine group had a mean reduction of 18.5 active joints compared with 11.9 in the placebo group at the end of the study, giving a difference between groups of 6.6 active joints favoring the hydroxychloroquine group (P = 0.03). Patient and physician global assessment also offered the following information: patient (hydroxychloroquine: improved, 64%; no change, 28%; worse, 8%; placebo: improved, 45%; no change, 34%; worse, 22%); and physician (hydroxychloroquine: improved, 89%; no change, 5%; worse, 6%; placebo: improved, 73%; no change, 19%; worse, 7%). Global assessment scores were also statistically different at the 0.05 level, favoring the hydroxychloroquine group. Figure 1 shows the monthly data for the 101 patients who continued to be followed at these time points. No changes were noted in the number of nonsteroidal anti-inflammatory agents and aspirin analgesics consumed. The erythrocyte sedimentation rate and the natural logarithm of the rheumatoid factor showed no differences between the two groups.

View larger version:

• In this window
• In a new window

Figure 1. VAS = visual analog scale. Raw mean scores for joint scores, patient assessment of pain, and grip strength variables for each visit and each treatment group.

Side effects occurred in 56 patients (46%), 28 in the active group and 28 in the placebo group. All side effects were minor except in 1 patient who had severe anemia and had to be withdrawn from the study; this patient was on placebo. Of these 56 patients, 32 had one side effect; 19 patients, two different side effects; and only 5 patients, three or more side effects. A total of 91 reactions were recorded (Table 3). No ocular toxicity was detected.

Discussion

This study was designed to test as homogeneous a group as possible with mild disease unresponsive to nonsteroidal therapy, the group for which the textbooks recommend it [11]. This was satisfactorily accomplished as evidenced by 80% of patients being functional class 1; 40%, being rheumatoid factor positive; and 93%, having five or fewer radiographic erosions. A clinically and statistically significant benefit was observed in validated clinical outcomes known to be responsive to change in clinical trials of rheumatoid arthritis [20].

Many previous controlled [3-10, 12] and uncontrolled [21-24] trials of hydroxychloroquine have been conducted, with differing results. Controversy exists concerning the benefits of hydroxychloroquine compared with other second-line agents. Some authors have suggested [25-27] that the degree of response seen with currently recommended doses of hydroxychloroquine is similar to that seen with gold (both intramuscular and oral), D-penicillamine, and azathioprine. On the other hand, a 2-year, double-blind study conducted at the Mayo Clinic [28], comparing hydroxychloroquine and D-penicillamine, showed that D-penicillamine was superior to hydroxychloroquine at 6 months. The controlled trials have had methodologic flaws such as the use of a heterogeneous rheumatoid study sample in terms of disease severity and activity and inclusion criteria [3-5, 7, 8, 12]. Some trials used hydroxychloroquine [7-9] and others used chloroquine [3-6, 10]; different doses were tested or doses were not standardized [7-9]; or baseline differences existed between active and control groups [3, 4, 6, 7, 12]. Some studies were not blinded [6, 7, 9]; the sample size in some was small [4, 5]; and the end points were not standardized.

These problems have led many clinicians not to use hydroxychloroquine. A survey in the United States in 1987 showed that only 16 of 91 rheumatologists (18%) and 4 of 63 internists (6%) selected hydroxychloroquine as their first choice of a second-line agent in patients who failed to respond to nonsteroidal anti-inflammatory agents [29]. Pullar [30], in reviewing the choice of second-line therapy in the United Kingdom, stated: Most British Rheumatologists would probably agree that sodium aurothiomalate is the most effective of the standard agents, with sulfasalazine and penicillamine similar to each other and next in ranking order and with antimalarials the least effective. Brooks and colleagues [31] surveyed Australian rheumatologists in 1984 and 1989 and found that antimalarial agents were used often by 54% in 1984 and by 60% in 1989 in comparison with gold being used often by 94% in 1984 and by 89% in 1989. The recent meta-analyses by Felson and colleagues [25] based on both comparative and placebo-controlled studies of both hydroxychloroquine and chloroquine suggest that antimalarial drugs are more effective than are azathioprine and auranofin. Our study increases the precision of the Felson meta-analysis estimate and may provide useful evidence on the efficacy of currently recommended doses of interest to rheumatologists and internists considering whether to prescribe it.

The side effects caused by hydroxychloroquine were minor and transient, occurring in 28 patients in the hydroxychloroquine group and in 28 in the placebo group; there were no cases in the active group in which the therapy had to be discontinued (Table 3). Nausea and epigastric pain were the most frequent and were found in 30.2% in the active group and 17.5% on the placebo group (P = 0.11); these gastrointestinal complaints have been reported so far to be the most frequent reactions and mimic those of nonsteroidal anti-inflammatory drugs [32]. Headache was also present in 12.7% in the active group compared with no headache in the control group (P = 0.004). The toxicity profile in our study did not differ from those of other studies [10, 12]. As expected, no cases of ocular abnormalities occurred. This type of toxicity, as is well documented, is low and has not, to our knowledge, been reported with the doses we used during this length of observation.

The place of hydroxychloroquine in the sequence of slow-acting agents is in early mild disease. The study by Davis and colleagues [12] showed that patients with very mild disease (synovitis limited to the hands and feet, and erythrocyte sedimentation rate <30 mm/h) who would normally receive only a nonsteroidal anti-inflammatory agent, showed benefit from hydroxychloroquine. In contrast, our study shows that patients who, despite a trial of nonsteroidal anti-inflammatory drugs, have widespread synovitis of longer duration, higher sedimentation rates, and a greater degree of disability, show clinical benefits.

Hydroxychloroquine is less toxic than any of the other second-line agents [33]. Lower toxicity [27] has been confirmed in a study using life-table analysis that compared hydroxychloroquine to intramuscular gold and D-penicillamine.

The treatment effect of hydroxychloroquine on joint score takes some weeks to separate out from the improvement seen in the placebo group; some authors have reported that the maximal benefit may not be observed for a year (Figure 1) [27]. Because at 6 months in our study, the lines (shown in Figure 1) between placebo and hydroxychloroquine were continuing to diverge, it is possible that this benefit will continue to increase. In contrast, the benefit in pain relief was substantially different between groups beginning with the second visit (Figure 1).

A related finding is the clinical and statistical improvement of both the placebo as well as the hydroxychloroquine group during the course of the study. For the joint score, our primary variable, the improvement in the placebo group of 11.0 joints (P = 0.001) was much higher than in most other reported studies of slow-acting agents. This finding has been noted in other studies of patients with early disease [25]; our patients were in an early stage of rheumatoid arthritis, with a mean duration of disease of 30 months for the hydroxychloroquine group and 27.9 months for the placebo group. Because improvement is associated with the natural history of the disease, it is not surprising in view of the relatively short duration of disease in this trial compared with patients studied in many of the reported trials of slow-acting agents. Even with this large placebo effect, however, our results are statistically significant in favor of the hydroxychloroquine group.

Our study showed that hydroxychloroquine has modest but useful benefit in patients with early rheumatoid arthritis who would benefit from its relatively low toxicity. It was superior to placebo during a period of 6 months and offered benefit as the first choice of a second-line agent in patients with early rheumatoid arthritis who did not respond to therapy with appropriate nonsteroidal anti-inflammatory drugs.