Methods

Search strategy and selection criteria

A literature search for papers in English from 1982 to 2007 inclusive was performed using the MEDLINE and EMBASE databases and by manual search.

The GRADE system was used to analyse the quality of the evidence7 (see supplementary table 1 for definitions). The final score for the evidence available on each item was agreed on by all the authors.

See also supplementary methods.

Results

The selection process is shown in fig 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Selection process.

Effects of AMs on SLE activity

A total of 11 studies were found that included data on the effect of AMs on lupus activity: 4 randomised controlled trials (RCTs),8 9 10 11 4 prospective12 13 14 15 and 2 retrospective16 17 cohort studies, and 1 retrospective analysis of data from the extension of a RCT18 (table 1). Three of these studies, including a RCT,11 were carried out in pregnant women with SLE.11 13 15 Two studies were limited to patients with lupus nephritis.16 17 Lupus activity was defined using either accepted activity scales, predefined clinical manifestations or criteria for remission of lupus nephritis.

Whatever the definition of lupus activity was used, all the studies consistently found it to be significantly reduced, over 50% in most studies, in patients treated with AMs. Severe flares were also reduced, however with borderline statistical significance.8 18 In the two retrospective studies of patients with lupus nephritis, HCQ was identified as an adjuvant treatment for achieving remission.16 17 AMs allowed a significant reduction of corticosteroid treatment in three studies.9 11 14 One study found lower blood levels of HCQ among patients experiencing SLE flares.13

In summary, no major limitations were found regarding the composition of the study groups, fulfilling American College of Rheumatology (ACR) classification criteria and diverse in ethnic origin and disease severity. Groups of treated and untreated patients were well balanced, without major differences in potentially confounding variables. Two RCTs (one in patients who were pregnant and one in patients who were non-pregnant) were double blind.8 11

AMs reduce SLE activity in patients who were pregnant and patients who were non-pregnant with a high quality of evidence. The effect on severe flares and on lupus nephritis was either of borderline statistical significance or supported by retrospective observational studies only, thus with a low quality of evidence.

Effects of AMs on lipid profile

Nine studies have analysed the effect of AMs on the lipid profile19 20 21 22 23 24 25 26 27 (see supplementary table 2). All the studies were carried out in prospective cohorts of patients with SLE using a cross-sectional design and one study25 also analysed the effect of AMs longitudinally.

All but two studies23 27 found significant differences in serum levels of lipid parameters in patients who were receiving AMs in comparison with those who were not (supplementary table 2). The study by Petri et al26 analysed the largest cohort (264 patients with SLE) and was the only study to confirm the results by multivariate analysis.

The three studies21 22 25 that specifically analysed the effect of AMs on the lipid profile in patients taking corticosteroids found a significant reduction in total cholesterol (TC),22 25 very low density lipoprotein cholesterol (VLDLc),21 22 low density lipoprotein cholesterol (LDLc)22 and triglyceride (TG)21 levels and a significant increase in high density lipoprotein cholesterol (HDLc) levels21 compared with patients taking corticosteroids alone.

In addition, one cross-sectional study focused on the prevalence and risk factors of metabolic syndrome in SLE,28 defined by internationally accepted criteria. A total of 102 patients with SLE were studied, with a prevalence of metabolic syndrome around 30%, higher than in controls. Current use of HCQ was similar among patients with or without metabolic syndrome.

In summary, seven out of nine observational studies found that the use of AMs had a beneficial effect on serum lipid levels, including patients taking corticosteroid therapy. The two studies in which a beneficial effect was not found were performed in patients from China23 and from Iran.27 Although it is possible that a racial influence modulates the effect of AMs, the negative results of two studies adds inconsistency. The magnitude of the effect was variable, but generally modest. The quality of evidence supporting a clinically meaningful beneficial effect of AMs in the lipid profile in patients with SLE is rated as low.

There are no data supporting any protective effect of AMs on the development of metabolic syndrome.

Effect of AMs on bone metabolism

Two cross-sectional studies have investigated the effect of HCQ on bone mineral density (BMD).29 30 Both were performed in small cohorts of patients with SLE (see supplementary table 3). The multivariate analyses showed that current/past use of HCQ was associated with a higher mean BMD of the spine29 and the hip,30 the length of HCQ therapy correlated positively with the mean BMD of the hip and negatively with osteopenia/osteoporosis of the spine30 and the cumulative HCQ dose correlated positively with the BMD of the spine.29

A case-control study by Calvo-Alen et al31 found a lower exposure time to HCQ in patients with osteonecrosis, although the difference was not significant in the multivariate model. A similar study by Prasad et al32 including 130 patients (65 cases and 65 controls) found no differences in the frequency of treatment with HCQ at any time between cases and controls.

A cross-sectional study performed in 50 patients (25 with SLE and 25 with fibromyalgia) found lower levels of 1–25 (OH)₂ vitamin D in HCQ users, although the difference was statistically significant only when patients with SLE and patients with fibromyalgia were combined.33

In summary, a protective effect of HCQ on the BMD, especially of the spine, was found with a moderate quality of evidence.

The evidence supporting a protective effect of HCQ on osteonecrosis was rated as very low.

Likewise, the influence of AMs on 1–25 (OH)₂ vitamin D was supported by very low quality of evidence.

Effects of AMs on atherosclerosis

Eight studies analysed the effect of AMs on atherosclerosis,34 35 36 37 38 39 40 41 defined as the presence of carotid plaque and/or abnormal intima/media index by carotid ultrasound (n = 5), of coronary calcifications by CT scan (n = 1) or by studies of vascular elasticity (n = 2) (see supplementary table 4). All the studies had a cross-sectional design within prospective cohorts of patients with SLE. Only one study was specifically designed to analyse the effect of treatment with HCQ on atherosclerosis.41 Exposure to AMs was defined as “ever prior to the study time” in five studies,36 37 38 39 40 one of them using the continuous variable “time on CQ”.37 The remaining three studies analysed current treatment with HCQ.34 35 41 Patients with clinically apparent cardiovascular disease were not excluded from any study.

Five studies did not find any effect of current (n = 1) or past (n = 4) treatment with AMs on the presence of atherosclerosis.34 37 38 39 40 One study36 found a borderline protective effect of previous HCQ and another study showed significantly lower vascular stiffness in premenopausal SLE women receiving HCQ.35 Only one small study found increased large artery elasticity and reduced systemic vascular resistance among patients treated with HCQ, as compared with untreated patients or those receiving corticosteroids only, respectively.41

Data on the effect of AMs on atherosclerosis, usually subclinical, are limited by the low consistency, lack of specific design and lack of quality of the studies. The effect was not quantified in most cases and the exposure to AMs has been heterogeneously defined, without taking into account the time of exposure or a possible dose effect. The quality of evidence is rated as very low.

Effects of AMs on thrombosis

Eight studies have investigated the effects of AM treatment on thrombosis (table 2). Of them, 5 were prospective observational cohort studies,4 6 42 43 44 1 was a cross-sectional study including 133 patients with anti-phospholipid antibodies, of whom 58 had SLE,45 1 was a cross-sectional study in which the potential predictors of symptomatic cardiovascular disease were analysed in retrospect46 and 1 was a retrospective study focused on the risk and predictors of thrombosis, comparing 162 patients with SLE and nephritis with 181 patients with primary glomerulonephritis.47 Three studies were designed to specifically analyse the role of AMs on thrombosis42 44 45; the other five studies focused on the search for variables potentially related with thrombosis in lupus.4 6 43 46 47

Exposure to AMs was defined as “ever treated” in five studies4 6 42 46 47 and “treated prior to the thrombotic event” in the remaining three.43 44 45 Only one study analysed the relationship between AM treatment and thrombosis in a time-dependent manner.44

Studies using the exposure variable “HCQ ever” showed a diversity of results: one found an inverse association between HCQ treatment and a history of thrombosis42; one showed no effect on the endpoint “arterial thrombosis”6; one found a borderline protective effect on venous but not arterial thrombosis4; and one found large differences between patients with and without a history of cardiovascular disease for the number of months and the cumulative dose of HCQ (33 vs 56 months and 244 vs 631 g, respectively), without reaching the level of statistical significance.46 The study by Mok et al did not find HCQ to protect from thrombosis.47 However, patients with SLE and non-SLE glomerulonephritis were mixed up in the analysis, HCQ being almost exclusively taken by patients with lupus, itself a predictor of thrombosis, which results in a major “confounding by indication” bias.

The three studies in which exposure to AMs was considered prior to the thrombotic event found a significant protective effect on thrombosis, arterial and venous taken as a whole.43 44 45 This effect was lost after multivariant analysis in the study by Ho et al.43 HCQ retained its effect after a bivariate analysis in the study by Erkan et al.45 The only study analysing the effect of AMs with a Cox multiple failure time survival analysis showed that they were protective against thrombosis (hazard ratio (HR) 0.28, 95% CI 0.08 to 0.90), after controlling for other variables such as history of thrombosis, presence of anti-phospholipid antibodies and treatment with aspirin.44

In summary, the effect of AMs in preventing thrombotic events was consistently found in studies taking into account the exposure previous to the event. A dose effect was suggested by one study with a small sample size, which prevented the statistical significance of the results. The magnitude of the effect is high according to the only study with time-dependent analysis, however, confidence intervals were wide. Accordingly, the quality of evidence available on this issue is rated as moderate. However, there are insufficient data to address whether the antithrombotic effect is present for arterial and venous events.

Effects of AMs on irreversible organ damage

Two prospective observational cohort studies have investigated the effect of HCQ on the accrual of irreversible organ damage, as measured using the Systemic Lupus International Collaborating Clinics–American College of Rheumatology Damage Index (SDI)48 (see supplementary table 5).

The first study was performed in a cohort of 151 patients with lupus from Israel, 93.4% of Jewish origin.49 After a mean follow-up of 45 months, treatment with HCQ had an inverse and independent relation with SDI values (p = 0.02). Additionally, patients treated with HCQ showed a higher damage-free survival than those untreated (p<0.001). No estimate of the effect was given by the authors.

The second study was performed in the multiethnic LUMINA (for “Lupus in Minorities: Nature vs Nurture”) cohort.50 A total of 518 patients were included in the analysis. The effect of HCQ on SDI was adjusted using the propensity score, which results in a form of pseudorandomisation of observational cohort studies.51 HCQ treatment taken at enrolling the cohort was protective against damage accrual after adjusting for the propensity score (HR 0.73, 95% CI 0.52 to 1.00). In those patients with no damage at the time of entering the study, the protective effect was more marked (propensity score-adjusted HR 0.55, 95% CI 0.34 to 0.87).

In summary, the two studies focused on the relation of AMs are consistent in finding a protective effect of HCQ. The protective effect was quantified in only one of the studies, HR being above 0.5, with wide 95% CIs. Therefore, the quality of evidence on this issue is rated as moderate.

Effects of AMs on survival

Three studies have analysed the long-term effects of AM use in the survival of patients with SLE44 52 53 (table 3): one case-control study (case = deceased patients with SLE with autopsy, one control per case, matched by age and disease duration),53 one prospective observational cohort study44 and one nested case-control study (case = deceased patients, three controls per case matched for time to event) within a prospective cohort.52 Exposure to AMs was always defined as “ever treated”. The first study used a logistic regression analysis including indexes for severity and lupus activity, as well as all the variables related to mortality.53 The propensity score analysis (see above) was used in the other two studies.44 52

The first study was performed in patients with SLE from Mexico City, Mexico, selected from the autopsy registers and hospital databases.53 A total of 76 cases and 76 controls were included. Data were obtained from chart review. The dose of CQ was lower in deceased than in living patients (3.6 mg/day vs 39.4 mg/day, p<0.001). However, it was not a significant predictor in the multiple regression model. This study is severely limited by the retrospective acquisition of data and the likely “confounding by indication” bias.

The second study was performed in a cohort of 232 patients from the Basque Country, Spain,44 99% of whom were white. In all, 83% of deaths took place among patients never treated with AMs. No patients taking either CQ or HCQ died of cardiovascular causes. Unadjusted and propensity score-adjusted HR were similar: 0.13 (95% CI 0.04 to 0.39) and 0.14 (95% CI 0.04 to 0.48), respectively.

The third study was performed within the prospective LUMINA cohort,52 selecting 244 patients from the 608 participants in the cohort. In this case, patients were almost equally divided among Hispanics, African–Americans and Caucasians. All patients treated with AMs took HCQ. Within the whole cohort, 17% of patients not taking HCQ died during the follow-up, vs 5% of those treated with HCQ (p<0.001). In the case-control study, unadjusted and propensity score-adjusted odds ratios (ORs) were 0.28 (95% CI 0.054 to 0.301) and 0.319 (95% CI 0.118 to 0.864), respectively.

Given its limitations, the Mexican study can only suggest a beneficial effect of CQ on survival. The two prospective studies, both using propensity score analysis, have shown a consistent protective effect of AMs, with a reduction in mortality higher than 50% in both cases. Major biases are unlikely to fully explain these large differences. Thus, we upgraded the score and consider the quality of evidence on this issue as high.

Adverse effects of AMs

We found four studies analysing the prevalence of all types of adverse events (AE) in patients with SLE treated with AMs54 55 56 57 (table 4). Two were retrospective studies,56 57 one was a prospective cohort study55 and one was a small RCT comparing clofazimine and CQ.54 One study included diseases other than SLE.57 Patients received HCQ only in one study,56 CQ only in one study54 and one did not make a difference between HCQ and CQ users.55 Only 1 study57 has compared the frequency of AE among HCQ and CQ users in a large series of 940 patients (including 178 with SLE).

The frequency of AE reported in all the studies has been low, mainly gastrointestinal and cutaneous, usually mild.54 55 56 57 The only study comparing the toxicity of HCQ and CQ found a higher frequency of AE in patients receiving CQ in comparison with those receiving HCQ (28.4% vs 14.7%, p<0.001). Overall, 15% of patients discontinued AMs due to toxicity, patients receiving HCQ being less likely to discontinue the drug due to side effects than those taking CQ (adjusted HR 0.62, 95% CI 0.40 to 0.96).

Seven studies have evaluated the frequency of retinal toxicity in patients treated with AMs57 58 59 60 61 62 63 (table 4). Three studies included diseases other than SLE.57 60 62 63 Four studies included HCQ users,59 60 61 63 one CQ users58 and two compared HCQ and CQ users.57 62

In all, 4 studies,57 58 62 64 including 647 patients treated with CQ for a mean of >10 years, found that 16 (2.5%) patients were diagnosed as having definite retinal toxicity, in comparison with only 2 (0.1%) of 2043 patients taking HCQ for a similar period included in 6 studies57 59 60 62 63 65 (OR 25.88, 95% CI 6.05 to 232.28, p<0.001). When patients classified as having probable retinal toxicity were added, there were 17/647 (2.6%) CQ users and 6/2043 (0.3%) HCQ users with toxicity (OR 9.16, 95% CI 3.42 to 28.47, p<0.001).

Johnson et al61 analysed seven patients taking HCQ for a mean of 14.2 years with an accumulated dose >1 g. Exhaustive ophthalmological assessment found fine granularity of the macular pigment epithelium in two patients with borderline abnormalities of colour vision in one, classified as a possible case of early toxicity.

Cardiotoxicity of AMs in patients with SLE has been evaluated in 2 studies66 67 including 70 patients treated with HCQ and 28 patients treated with CQ, respectively. No cases of clinically relevant cardiotoxicity were reported.

In addition, we analysed the clinical characteristics of case reports of AM toxicity68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 (see supplementary table 6 and supplementary material on toxicity case reports).

In summary, toxicity associated to AM use was infrequent and generally mild. Large series are consistent in showing the lack of serious adverse events, even after prolonged use. Overall, the frequency of adverse events with CQ was higher to that with HCQ, although data come only from observational studies. Therefore, evidence supporting the global safety of AMs (HCQ and CQ) is rated as high. HCQ offers a safer profile than CQ, with a moderate grade of evidence.

Adverse effects of AMs on the fetus

In all, 10 studies reported on the effects of AMs on children born to lupus mothers.11 14 94 95 96 97 98 99 100 101 A total of 275 pregnant patients with SLE received HCQ and 36 received CQ. In addition, 1 study reported on 21 patients (14 with HCQ and 7 with CQ) who were diagnosed as having either SLE or RA.97 Five studies included a control group of pregnant women with SLE who had not been treated compared with patients who had been treated.14 94 96 99 Monitoring for toxicity on the baby included examination for malformations and also surveillance of ocular, auditory and neurological problems arising from birth until more than 4 years of age, depending on the study (see supplementary table 7).

All the studies are concordant in showing the absolute safety of AMs during pregnancy: congenital malformations were not more frequent than in unexposed children and no cases of ocular, auditory or neurological toxicity were reported.

Therefore, AMs are safe for the fetus when given to the pregnant mother, with no cases of fetal malformations (beyond those normally expected) or toxicity reported among more than 300 children exposed to these drugs. Due to the much higher number of patients with SLE using HCQ as compared with CQ, the quality of evidence is rated as high for HCQ and moderate for CQ.

Miscellanea

A retrospective nested cohort study of 130 patients with SLE analysed the effect of AMs in delaying the completion of ACR classification criteria in patients with lupus-like disease.102 Patients taking HCQ delayed the time until diagnosis of full-blown SLE (1.08 vs 0.29 years, p = 0.018), being less likely to present with proteinuria (p<0.05), leucopenia (p<0.05) or lymphopenia (p<0.001).

There was 1 study that investigated the effect of AMs (CQ or HCQ) on the risk of developing cancer in a prospective cohort of 235 patients with SLE.103 In all, 13 patients developed cancer; 2/156 (1.3%) in the AM group vs 11/79 (13%) in the non-AM group (p<0.001). The adjusted HR of developing cancer in patients treated with AMs vs patients not treated with AMs was 0.15 (95% CI 0.2 to 0.99).

Both studies need further confirmation of their results. Thus, current evidence in both cases is rated as low.

See also supplementary discussion.

Discussion

Taking into account the current evidence analysed and discussed in this paper (table 5), we recommend the treatment with AMs, preferably HCQ, of most patients with SLE, starting as soon as the diagnosis has been made. This treatment could be maintained long term if toxicity does not ensue, whatever the subsequent course of lupus (pregnancy included) and the additional medications needed. Routine ophthalmological controls in patients with SLE could be performed according to the recommendations of the American College of Ophthalmology,74 although less or more frequent controls could be agreed with the attending ophthalmologist. These conclusions, as well as the ideal long-term maintenance dose of HCQ (400 mg/day, 200 mg/day or even less) should be confirmed and defined in future studies and meta-analyses.