Low dose chloroquine decreases insulin resistance in human metabolic syndrome but does not reduce carotid intima-media thickness

These single center studies were approved by the Washington University Human Research Protection Office. All subjects gave informed consent and were compensated for their participation. A data safety and monitoring board convened by the National Heart, Lung and Blood Institute (NHLBI) supervised the trials.

For both the dose escalation and double blind trials, inclusion criteria included men and women of any ethnic group between the ages of 18–60 (Trial 1), and 18–70 (Trial 2) (the difference in age between protocols was intended to make the randomized trial more pertinent to a broader age group) who had at least 3 of the components of the metabolic syndrome: elevated fasting triglycerides (≥ 1.69 mmol/L); low HDL cholesterol (< 1.29 mmol/L in women, < 1.03 mmol/L in men); hypertension (≥ 130/85 mm Hg ≤ 160/100 mmHg) untreated, or hypertension controlled (≤ 150/90 mm Hg) on a stable medication regimen; increased waist circumference (> 89 cm in women, > 102 cm in men); elevated fasting glucose (≥ 5.6 mmol/L and < 7.0 mmol/L). For a diagnosis of hypertension, measurements were repeated on two separate days. Fasting glucose was also repeated on two separate days. Fasting lipids were performed once before enrollment but repeated throughout the interventions.

Exclusion criteria were extensive and included: known prior treatment with chloroquine or hydroxychloroquine, BMI > 45 kg/m², coronary artery disease or other vascular disease, history of stroke, eGFR (estimated glomerular filtration rate) < 60 ml/min/1.73 m², diabetes, seizure disorder, history of psoriasis, hematological disorders (including anemia), current malignancy, asthma or COPD (chronic obstructive pulmonary disease), liver disease (including transaminases > 2× upper limit of normal), active infection (including HIV-human immunodeficiency virus), any serious illness requiring ongoing medical care, major psychiatric illness, lipid lowering medications (other than statins and < 1 g daily fish oils), severe hypertension at baseline or taking more than three antihypertensives, use of cimetidine or vitamin E, pregnancy or lactation or intention to become pregnant, inadequate use of contraception, glucose-6-phosphate dehydrogenase deficiency, auditory disease or hearing loss, retinal disease (especially presence of drusen or pigmentary changes at the macula), and any ocular disease interfering with the ability to rigorously assess the retina.

A CONSORT statement flow diagram for the first trial is shown in Fig. 1. The study design for the first trial is shown in Fig. 2a. One hundred forty-four subjects were screened for participation in Trial 1 and 35 individuals with the metabolic syndrome qualified. Twenty-five subjects completed the protocol. Ten discontinued the study for reasons that included: starting a new medication during the protocol [3], new diagnosis of diabetes during the protocol [1], difficult IV access [1], worsening joint symptoms [1], and miscellaneous issues unrelated to the protocol (such as relocation, 4). For this single blind, placebo-controlled study, each subject took a capsule that was identical in appearance for 21 days. For the first limb, capsule contents were inert. For the second limb, three of the 21 capsules (administered at weekly intervals) contained 80 mg of chloroquine and the remaining capsules were inert. For the third and fourth limbs, capsules contained 80 or 250 mg of chloroquine, respectively. Each limb was separated by a washout period to allow recovery from the blood drawing of the clamp procedure. It was not possible to randomize limb order due to the extremely long half-life of chloroquine; ingestion during an early limb would be associated with tissue levels of the drug and its metabolites during a later placebo limb.

A CONSORT statement flow diagram for the second trial is shown in Fig. 3. The study design for the second trial is shown in Fig. 4a. The blind was held by a research pharmacist, who provided the randomization sequence to the blinded research nurses or physicians who enrolled participants and assigned interventions that included study medications that were identical in appearance. Three hundred fifty-seven subjects were screened for participation in Trial 2 and 155 non-diabetic individuals with metabolic syndrome qualified at the initial screening visit. Of these, 21 withdrew consent before the second screening visit, 4 withdrew consent after the second screening visit and 14 did not qualify for randomization due to abnormal eye exams [4], abnormal carotid imaging [2], anemia [3], G6PD deficiency [2], abnormal liver function tests [2], and psoriasis [1]. The remaining 116 subjects were randomized to placebo or chloroquine. Of those randomized, 19 assigned to placebo and 20 assigned to chloroquine were being treated for hypertension, and were taking 1.3 ± 1.1 and 1.1 ± 1.0 antihypertensive medications respectively. 15 subjects in each group were taking a statin.

At the end of each limb depicted in Fig. 2a, subjects presented after an overnight fast. The antecubital vein was catheterized for infusion of insulin, glucose and tracer solutions, a hand vein was catheterized for blood sampling, and the hand was maintained in a hotbox at ~ 55 °C to arterialize venous sampling. Baseline samples were obtained prior to initiating infusions.

Insulin sensitivity was assayed using a two-step euglycemic clamp technique as described previously [11]. The stable isotope [6,6-²H₂] glucose (22 μmol/kg prime and 0.25 μmol/kg/min constant infusion) was infused during a 90 min basal period (to allow isotope equilibration) and then continued throughout each study. Plasma glucose was measured at 5–10 min intervals following the basal period. Insulin was initially infused in two stages consisting of rates of 181 and 486 pmol/m²/min, but after analyzing results from the first 7 patients, the protocol was modified to include two stages consisting of insulin infusion rates of 56 and 181 pmol/m²/min in order to avoid nearly complete suppression of hepatic glucose production at higher insulin infusion rates. Twenty % dextrose (containing 1.5% [6,6-²H₂] glucose) was infused to maintain blood glucose at 5.00 mmol/L for ~ 2 h at each stage. Plasma was assayed for 6,6-²H₂-glucose enrichment by mass spectrometry. Glucose kinetics were determined using samples obtained at 10 min intervals during the last 30 min of the basal period and during each stage of the clamp.

Oral glucose tolerance testing was performed by administering 75 g of glucose followed by sampling at 0, 30, 60, 90 and 120 min. Samples were assayed for glucose, insulin, C peptide, and glucagon. AUC calculations were performed using the trapezoid rule for approximating integrals. Lipids, lipoproteins, serum chemistries, and blood counts were assayed by the Washington University Core Laboratory for Clinical Studies, which is CLIA-certified. Since this project addressed the effects of chloroquine on insulin resistance, the cause of which remains obscure, the laboratory also performed exploratory measurements of putative metabolic mediators that have been implicated in insulin resistance and/or its vascular complications [12‐16]. These included adiponectin, leptin, iron, VEGF (vascular endothelial growth factor), and IL-17A. Samples for adiponectin and iron studies were assayed at the University of Utah [17].

Two techniques were employed: auscultation of seated subjects at rest was performed by a trained observer who recorded the first and fifth phases of the Korotkoff sounds; and, a portable oscillometric device (SpaceLabs Medical) recorded results every 20 min during the day and every h during the night then data were analyzed as mean values over 24 h, between 0800 and 2200 (daytime), and between 2200 and 0800 (nighttime).

Carotid artery intima-media thickness (CIMT) was measured from B-mode images by a single sonographer using standard approaches [18, 19]. The average of 3 measurements of the posterior wall of each common carotid ~ 1 cm proximal to the carotid bulb was determined; plaques (wall thickening > 50% of the surrounding wall) were excluded. To establish reproducibility, 10% of the images were randomly selected and re-measured in a blinded fashion by two independent observers. Intra-observer and inter-observer correlation coefficients were 0.900 and 0.992.

A contouring tool on the Philips ViewForum workstation was used to outline the inner and outer common carotid artery wall regions (right and left) in the first frame immediately below the bifurcation in the T2-weighted image, then areas and diameters of the vessel wall and lumen were calculated. The maximum thickness of the common carotid in the same frame was measured with the line tool. Similar measurements were made for the internal carotid artery in a slice just beyond the bifurcation. For common carotid artery wall contrast enhancement, which represents gadolinium leakage and late hyper enhancement, pre- and post-injection T1 weighted images were compared by dividing the artery into quadrants. The slice nearest the bifurcation manifesting the greatest signal in terms of numbers of quadrants enhanced was chosen by visual comparison with the same pre-injection slice. If the pre-injection signal in the carotid wall appeared bright in any quadrant due to signal spatial heterogeneity, and with reference to the other arterial wall segments in adjacent or contralateral arteries, then the post- injection signal was deemed nonenhancing. Concurrence of two readers was achieved in blinded interpretations.

Human monocytes were isolated using a negative selection technique employing a RosetteSep Human Monocyte Enrichment Cocktail (StemCell Technologies #15028), exactly as described by the manufacturer. Cells were lysed using a buffer containing 1% Tween 20 and 0.2% NP40 with protease and phosphatase inhibitors, and lysates were aliquoted and stored at − 80 °C.

Total and phospho (Thr183/Tyr185) JNK were assayed using chemiluminescent sandwich ELISA kits (Cell Signaling Technologies #7869S and #7849S). Extracts from subjects before and 1 year following treatment were analyzed in the same assay by an observer masked to randomization status. Signals detected from 1 μg lysate protein (assayed with BCA reagents from Pierce) were determined using a Synergy 4 plate reader and the ratio of phospho-JNK to total JNK was calculated. For assays of cholestane-3β,5α,6β-triol, plasma samples stored at − 80 °C were thawed and combined with BHT (50 μg/mL). Oxysterols were purified, derivatized, then quantified by GC–MS (m/z 456) in assays performed within the linear response range of appropriate standard curves [20].

Results are expressed as mean ± SD except in Fig. 2 where results are presented as mean ± SE to simplify data presentation. Statistical analyses were performed using GraphPad Prism software. For the dose escalation study, most data were analyzed using repeated measures ANOVA with post hoc testing using Tukey’s multiple comparison test for inter-group differences. For the yearlong treatment study, most data were analyzed using a mixed model approach for repeated measures to allow for dealing with missing values. P values for mixed model analyses represent an assessment of whether the dependent variable was affected by the randomization group (placebo vs. chloroquine) vs. time interaction. Some analyses in both trials involving only two groups were performed using a two-tailed Student’s t test. Spearman correlation coefficients were calculated for the relationships between baseline CIMT in the 116 randomized subjects and variables associated with atherosclerosis. Two-tailed paired t tests were used to analyze the baseline vs. 12-month results for MRI imaging and JNK activation. Binary baseline characteristics and adverse events were analyzed using contingency tables and Chi Square Test or Fisher’s Exact Test.

Characteristics for those completing the dose escalation study are shown in Table 1. The primary endpoint of the first trial was clamp-determined insulin sensitivity. Two-stage hyperinsulinemic euglycemic clamps were initially performed using insulin at 181 and 486 pmol/m²/min. However, data analysis after the first seven patients demonstrated complete suppression of glucose production at the higher dose and nearly complete suppression at the lower dose, decreasing the ability to detect differences in endogenous glucose production in this group of patients with metabolic syndrome but not diabetes. Subsequent two-stage clamps were performed using insulin infusions at 56 and 181 pmol/m²/min. The results from these first seven patients are included in Fig. 2b, c; the 486 pmol/m²/min condition represents data points from these first patients for each limb while the 181 pmol/m²/min condition includes data points from these patients in addition to those from those undergoing clamps at 56 and 181 pmol/m²/min. There were no differences in glucose disposal rates between any of the chloroquine doses as shown in Fig. 2b.

There was a dose-dependent effect of chloroquine on hepatic glucose production (clamp-determined rate of appearance) with significant suppression at doses of 80 mg/day and 250 mg/day as determined by repeated measures ANOVA (Fig. 2c). Figure 2d shows that chloroquine increased hepatic insulin sensitivity based on data from the 56 pmol/m²/min stage.

There was also an apparent dose-dependent effect of chloroquine on fasting levels of total cholesterol, non-HDL cholesterol, and LDL cholesterol with a significant decrease at 250/day as determined by repeated measures ANOVA (Fig. 2e). Serum triglycerides and HDL cholesterol were unchanged (Fig. 2f).

Chloroquine in this dose escalation study had no effect on blood pressure determined using an ambulatory monitor as shown in Table 2. SBP, DBP, MAP, and magnitude of BP decrease overnight (“dipping” status) were unaffected in all limbs. Consistent with an effect of chloroquine on hepatic insulin sensitivity, fasting glucose at the time of the clamp was significantly decreased (Table 2). Consistent with the lack an effect of chloroquine on glucose disposal, OGTT AUC results were unaffected in all limbs (Table 2). There was also no effect on body weight, glucagon, leptin, NEFA, C-peptide or insulin (Table 2).

There was a trend toward lower levels of TNFα (Table 2), but other inflammatory markers including CRP and adiponectin were unaffected (Table 2), consistent with other large studies of the relationship between this adipokine and insulin sensitivity [21].

Iron is linked to the development of diabetes [22] and ferritin is the major storage form of iron. Chloroquine treatment was associated with a significant decrease in ferritin levels in humans (Table 2), confirming an effect seen in rodents [23].

Baseline characteristics for the 107 subjects who completed the second trial, a double-blind placebo controlled study are shown in Table 3. The primary endpoint of the second trial was CIMT. To validate carotid measurements in this study cohort, baseline CIMT values for the 116 subjects randomized to chloroquine or placebo were correlated with common variables associated with atherosclerosis. CIMT was positively correlated with age (r_s = 0.3963, P < 0.0001), systolic blood pressure (r_s = 0.3379, P = 0.0002), triglycerides (r_s = 0.2936, P = 0.0014), cholesterol (r_s = 0.3593, P < 0.0001), and LDL-C (r_s = 0.2590, P = 0.0050). There were no correlations with HDL-C, diastolic blood pressure, waist circumference, BMI, glucose, TNFα, IL-6, VEGF, or IL-17A (not shown).

There was no effect of chloroquine on CIMT after 12 months (Fig. 4b). Because chloroquine has a long half-life, imaging was repeated at 24 months, following a 12-month washout period. These results also showed no difference between groups. MRI studies were performed only at baseline and at 12 months. Chloroquine did not affect contrast enhancement, a pre-specified secondary endpoint of the study (Fig. 4c, P = 0.0697, paired two-tailed t test). Representative images of vessels at baseline and 12 months are shown in Fig. 4d. Carotid dimensions were also assessed by MRI. Chloroquine was associated with a decrease in common carotid artery luminal area (Fig. 4e, P = 0.0033) and diameter (Fig. 4f, P = 0.0019).

Drug treatment over the first year was associated with a significantly lower diastolic blood pressure (P = 0.0252 at 12 months), results that were driven by a 4.9 mm Hg decrease at the 6 month time point (Fig. 5a). Ambulatory monitoring results (performed at baseline, 6 months, and 12 months but not at 24 months) were consistent with a beneficial effect of chloroquine on blood pressure. Drug treatment over the first year was associated with significantly lower daytime mean arterial pressure (Fig. 5b, P = 0.0484). Daytime (P = 0.0804) and nighttime (P = 0.0659) systolic pressures by ambulatory monitoring were not different (Fig. 5c, d). By mixed model testing for the interaction between group and time, there was no effect of chronic chloroquine on lipids. However, at 12 months, a pre-specified secondary endpoint, total cholesterol was 9% lower (Fig. 5e, P = 0.0089 by unpaired t test), non-HDL cholesterol was 12% lower (Fig. 5f, P = 0.0048 by unpaired t test), and LDL-C was 13% lower (Fig. 5g, P = 0.0038 by unpaired t test) in the chloroquine group. Glucose area under the curve (AUC) as well as HOMA-IR and the Matsuda index, measures of insulin sensitivity [24], were not different between groups (Table 4). There was no effect of chloroquine on insulin AUC, C peptide AUC, or the insulinogenic index (the difference between insulin at 30 min and 0 min divided by the difference between glucose at 30 min and 0 min) (Table 4), suggesting that chronic chloroquine treatment at this dose does not impact insulin levels in humans. Glucagon AUC, creatinine, ALT, and hematologic variables were not different between treatment groups (Table 4).

JNK is a stress kinase implicated in obesity-associated chronic inflammation [25]. JNK activation in monocytes, a pre-specified secondary endpoint, was significantly decreased in chloroquine-treated but not placebo-treated subjects (Fig. 6a). Reactive oxygen species activate JNK and have been implicated in the pathophysiology of obesity-associated conditions [26]. Plasma cholesterol oxidation products are elevated in diabetes and coronary artery disease [20]. Cholestane-3β, 5α, 6β-triol, one species that may reflect clinically relevant oxidative stress [20], was significantly decreased in the chloroquine group at 12 months (Fig. 6b).

For the dose escalation study, there were no serious adverse events (Table 5). Not unexpectedly given the rigors of a series of clamps, there were several study-related events. Of adverse events that were not directly study-related, infections were the most common.

For the double blind study, there were 7 serious events. In the placebo group, one subject presented with a new diagnosis of prostate cancer. One subject underwent a right knee replacement. In the chloroquine group, one patient developed transverse myelitis presenting with left leg weakness that subsequently improved but did not completely resolve. One subject underwent carotid endarterectomy for asymptomatic near-total carotid occlusion detected at 6 months. One subject was hospitalized for a cutaneous staphylococcus infection that resolved with appropriate therapy. One subject was treated for acute appendicitis, stopped the study medication for 1 week then resumed without incident. One subject was found to have a positive PPD without evidence of pulmonary tuberculosis.

Total adverse events were similar between groups (Table 6). There were more infectious events in those randomized to chloroquine but this was not significant (51 vs. 42, P = 0.0609). No subject in either study developed retinal pigmentary changes that may occur with chloroquine. For the double blind study, observers masked to the treatment assignment examined each subject at 0, 6, 12, and 24 months. Ocular adverse events were not significantly different between treatment groups (Table 6).