The Effect of Hydroxychloroquine on Insulin Sensitivity and Lipid Parameters in Non-Diabetic Patients with Rheumatoid Arthritis: A Randomized Blinded Cross-Over Trial

Abstract

Objectives

Observational studies suggest that hydroxychloroquine (HCQ) may reduce the risk of developing diabetes mellitus in patients with RA. We examined the effect of HCQ on insulin resistance in non-diabetic subjects with stable RA.

Methods

Twenty-three RA subjects not currently using HCQ completed a 16 week double blind cross-over study. Subjects were randomly allocated to receive HCQ (6.5 mg/kg per day) or placebo for the first 8 weeks followed by cross-over to the other arm for the final 8 weeks. Subjects underwent oral glucose tolerance testing and fasting lipid measurements at baseline, 8 and 16 weeks. The change (standard deviation) from baseline in insulin sensitivity index (ISI), insulin resistance (HOMA-IR), and lipid parameters were compared between placebo and HCQ using linear regression.

Results

The mean age was 56 years, with 96% females and the median body mass index was 26.0 kg/m². After 8 weeks of HCQ, the ISI increased 0.4 (SD 2.9) compared with a small increase during placebo of 0.14 (SD 3.1) (adjusted p = 0.785), and HOMA-IR decreased 0.3 (SD 1.5) during HCQ versus a decrease of 0.42 (SD 1.4) during placebo (adjusted p = 0.308). Small decreases in total cholesterol (12.7mg/dl) and LDL cholesterol (12.4mg/dl) were observed during the HCQ treatment periods (both adjusted p < 0.05 compared to placebo).

Conclusion

HCQ use for 8 weeks in non-diabetic patients with stable RA produced no significant change in insulin resistance. We observed small and statistically significant improvements in total and LDL cholesterol during HCQ treatment.

Introduction

Hydroxychloroquine (HCQ) was originally developed as an anti-malarial drug that is currently widely used in systemic inflammatory conditions, such as rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE). The mechanism of action is thought to be related to interference with lysosomal activity inhibiting antigen presentation as well as inhibition of toll-like receptor signaling.(1, 2) The effects on lysosomes seem to inhibit TGF beta production;(3) HCQ may also reduce the lysosomal degradation of insulin.(4)

This effect on insulin metabolism appears to have pharmacologic significance based on several lines of evidence. Case reports document patients developing hypoglycemia while using HCQ.(5) Several randomized controlled trials among patients with poorly controlled diabetes have demonstrated that HCQ significantly reduces levels of glycated hemoglobin, on the order of what is observed with sulfonylureas.(6, 7) Finally, large epidemiologic studies among patients with RA have shown that HCQ use is associated with significant reductions in the risk of diabetes.(8, 9)

These potential beneficial effects of HCQ could have a role in treatment decision-making for patients with systemic inflammatory conditions. It is plausible that such patients with known diabetes might be preferentially treated with HCQ to both manage the underlying inflammation and help to improve diabetes control. As well, since patients with RA and SLE are at an increased risk of cardiovascular disease, if HCQ reduces the risk of diabetes and/or insulin resistance, this might make it an important component of long-term treatment regimens.

In a prior pilot cross-over study, we found that six weeks of HCQ improved insulin sensitivity among obese, non-diabetic patients with insulin resistance.(10) Based upon the effects observed at six weeks of therapy, we undertook the current 16-week blinded randomized cross-over study to assess HCQ's effects on insulin and glucose metabolism after eight weeks of HCQ treatment in non-diabetic patients with RA. Because prior epidemiologic data suggested an improvement in lipid parameters among patients using HCQ, we also measured lipid profiles.

Methods

Study design and population

We undertook a cross-over trial to allow each patient to serve as her own control. This design is advantageous in small mechanistic studies when outcomes change rapidly, such as variables like insulin sensitivity and resistance. This design was useful in our pilot study where we observed changes in insulin and glucose metabolism over six weeks.(10)

After receiving approval from the appropriate Institutional Review Board, adult subjects ≥ 18 with RA were recruited if there was no: history of diabetes, prednisone use (or its equivalent) > 5mg per day in the prior 3 months, changes in disease modifying anti-rheumatic drugs (DMARDs) in the prior 3 months , or any contraindication to HCQ. The requirement for stable DMARDs was meant to reduce the probability of treatment changes during the study period.

Subjects meeting all criteria and giving written informed consent were then randomized to receive 8 weeks of HCQ (period 1) followed by 8 weeks of placebo (period 2) or 8 weeks of placebo (period 1) followed by 8 weeks of HCQ (period 2). This cross-over design allows patients on active treatment to be compared with themselves on placebo, with subjects serving as their own “controls”. Based on our pilot study in obese, non diabetic subjects without a diagnosis of RA, we anticipated that HCQ's effect would be apparent within 6 weeks of initiation. Most importantly, we observed no carry-over effect during our previous 12 week cross-over study (6 weeks on HCQ and 6 weeks off). Thus, we assumed no carry-over effect during this 16 week study (8 weeks on HCQ and 8 weeks off). Subjects, their treating rheumatologists, and the study assessors were all blinded to treatment sequence. Randomization sequence was generated through a random number generator and kept only by the study pharmacist.

Study procedures

Randomized subjects underwent baseline assessment, consisting of a physical examination, history and laboratory testing. The physical examination included 28-joint tender and swollen joint counts. These were used to calculate the Disease Activity Score (DAS-CRP) according to established equations.(11) Weight and height were measured using a Tanita Model WB300 scale and a Seca stadiometer, and manual blood pressure assessments were conducted with subjects seated. The laboratory assessment was performed after an overnight fast and consisted of an oral glucose tolerance test (OGTT) with blood draws at 5 time points – 0, 30, 60, 90, and 120 minutes after a 75 gram glucose load. Plasma glucose and insulin levels were measured in all five samples. A full lipid profile was also assessed in the fasting blood samples. These procedures were repeated at weeks 8 and 16.

HCQ was administered at a dosage of 6.5mg per kg/d, as used in clinical practice, and not to exceed 600 mg per day. Identical placebo pills were used during the placebo period.

Study outcomes

The primary outcome measure was the insulin sensitivity index (ISI) following HCQ treatment. The ISI was calculated based on the equation of Matsuda and Defronzo.(12) In addition, the HOMA-IR and HOMA-B were calculated.(13) The formulas for each of these calculations are as follows:

Matsuda Insulin Sensitivity Index:

$${\text{ISI}}_{\left(\text{Matsuda}\right)}=\frac{10000}{\sqrt{{\mathrm{G}}_0\times {\mathrm{I}}_0\times {\mathrm{G}}_{\text{mean}}\times {\mathrm{I}}_{\text{mean}}}}$$

ISI – insulin sensitivity index

G₀ – fasting plasma glucose (mg/dL)

I₀ – fasting plasma insulin (mIU/L)

G_mean – mean plasma glucose during OGTT (mg/dL)

I_mean – mean plasma insulin during OGTT (mIU/L)

HOMA-IR:

$$\text{HOMA}-\text{IR}=\frac{\text{Glucose}\times \text{Insulin}}{405}$$

HOMA-B:

$$\text{HOMA}-\beta =\frac{360\times \text{Insulin}}{\text{Glucose}-63}\%$$

Secondary study endpoints included enzymatic tests for total cholesterol, high density lipoprotein (HDL), calculated low density lipoprotein (CLDL), and triglycerides.

Statistical Analyses

Descriptive statistics such as minimum, maximum, range, median and inter-quartile range (IQR) were used to describe the primary outcome (ISI) and secondary outcomes. The primary outcome measure was the change in ISI levels during the HCQ period compared with the change during the placebo period. Assuming no carry-over effect, the treatment effect was evaluated by comparing the difference between the mean change during HCQ treatment versus placebo, using the time 0 measures as baseline for both periods. The differences were then compared using Wilcoxon signed-rank tests. Small changes in body weight were observed during the trial, so linear regression models were also fit adjusting for changes in body weight. Similar analyses were conducted for HOMA-IR and HOMA-B, the secondary insulin and glucose metabolism outcomes, and lipid parameters (total cholesterol, LDL, HDL and Triglycerides). Subgroup analyses were conducted based on baseline values of ISI, HOMA-IR, BMI and oral glucocorticoid use. We also evaluated the HCQ effect separately by sequence (period 1 versus period 2), and tested a possible sequence effect in a separate linear model. All p-values were calculated with two-sided significance level of 0.05.

Based on our prior work,(10) we assumed the following regarding the required sample size: expected improvement in ISI 4.4; standard deviation in ISI 6.2; within patient correlation 0.50; alpha 0.05; and power 90%. Assuming no carry-over effect and no period effect, we required 23 patients to observe a true difference between arms. Data analyses were performed using SAS 9.1.2 (SAS Institute, Inc, Cary, North Carolina). All outcomes were pre-specified and thus no correction for multiple testing was performed.

Results

The recruitment algorithm describing the study cohort. We screened 107 potentially eligible subjects with RA during 2010-2011; 7 subjects did not meet inclusion/exclusion criteria at screening and 63 declined participation (mainly because of their prior use of HCQ or of the requirements for fasting OGTTs). Of the remaining 37 subjects initially consented, 7 withdrew consent before randomization primarily because of subject related concerns (time commitment, not wishing to undergo OGTT). Thirty patients were randomized and received study drug. Two withdrew because of adverse events (1 with a rash and 1 with laboratory abnormalities likely unrelated to study drug) and 5 subjects chose to discontinue the protocol after the first OGTT. These seven patients did not have data at all three time points and were not included in the analyses. The final cohort was comprised of 23 subjects and their baseline characteristics are shown in Table 1.

Table 1. Baseline characteristics of study subjects.

	Mean (± standard deviation) or Median (interquartile range) or n (%)
Age, years, mean (SD)	56 (± 11.4)
Rheumatoid arthritis disease duration, n (%)
< 2 years	7 (30.4%)
2-5 years	2 (8.7%)
>5-10 years	2 (8.7%)
> 10 years	12 (52.2%)
Female gender, n (%)	22 (95.7%)
Body mass index, kg/m2 , median (IQR)	25.6 (22.9, 29.8)
DMARD use, n (%)	23 (100.0%)
Methotrexate	16
TNF blocker	13
Other DMARD	7
Oral glucocorticoid use, %	3 (8.7%)
Median dosage among users, milligrams	5.0
Swollen joint count, median (IQR)	7.0 (4, 11)
Tender joint count, median (IQR)	3.5 (1, 5)
CRP, mg/L, median (IQR)	2.3 (0.8, 4.7)
DAS-28 (CRP), median (IQR)	3.5 (2.9, 4.0)
Serum glucose, mg/dl, mean (SD)	115.4 (± 34.2)
Serum insulin, μIU/mL, mean (SD)	37.1 (± 31.6)
Total cholesterol, mg/dL, mean (SD)	188.3 (± 39.2)
Low density lipoprotein, mg/dL, mean (SD)	108.1 (± 32.5)
High density lipoprotein, mg/dL, mean (SD)	61.3 (± 19.4)
Triglycerides, mg/dL, mean (SD)	93.7 (± 43.0)

Laboratory measurements were taken at screening and not after a fast.

We compared the change in ISI, HOMA-IR and HOMA-B during the HCQ periods versus the placebo periods; change was calculated from the baseline values to the end of the respective periods (see Table 2 and Supplementary Figure). The ISI mean value at baseline was 7.7 with values > 8.0 considered normal. No statistically significant change in ISI was observed during HCQ compared with placebo: HCQ produced an increase in ISI of 0.4 (SD 2.9) and placebo an increase of 0.14 (SD 3.1) (adjusted p = 0.785). The HOMA-IR mean value at baseline was 2.0 with values < 2.5 considered normal. No statistically significant change in HOMA-IR was observed during HCQ compared with placebo: HOMA-IR decreased 0.30 (SD 1.5) during HCQ versus a decrease of 0.42 (SD 1.4) during placebo (adjusted p = 0.80).

Table 2. Mean (standard deviations) for measures of insulin, glucose metabolism and lipid parameters.

	Baseline	Placebo period value	Change during placebo*	Hydroxy-chloroquine period value	Change during HCQ*	Unadjusted P-value†	Adjusted P value‡
Insulin sensitivity index	7.7 (4.8)	7.8 (4.0)	0.14 (3.1)	8.1 (4.5)	0.4 (2.9)	0.930	0.785
HOMA-IR	2.0 (1.7)	1.6 (0.79)	-0.42 (1.4)	1.7 (1.0)	-0.3 (1.5)	0.575	0.308
HOMA-B	116.5 (100.4)	109.7 (82.6)	-6.8 (78.0)	110.8 (81.6)	-5.8 (72.9)	0.468	0.902
Total cholesterol (mg/dl)	192.4(37.5)	189.4(37.9)	-3.0 (22.4)	179.7(44.3)	- 12.7 (20.0)	0.004	0.004
LDL(mg/dl)	114.1(29.8)	109.9(33.1)	-4.2 (17.7)	101.7(36.7)	- 12.4 (20.1)	0.009	0.011
HDL (mg/dl)	58.1(17.9)	60.3(17.8)	2.2(9.8)	59.4(18.2)	1.3(6.7)	0.730	0.208
TRI (mg/dl)	100.6(44.0)	95.6(44.9)	-5.0(20.1)	92.4(41.7)	-8.2(45.0)	0.487	0.884

^*Change calculated from baseline value under the assumption of no carry-over effect.

^†P-value for the difference between the change during hydroxychloroquine versus placebo using Wilcoxon signed-rank tests.

^‡P value for the difference between the change during hydroxychloroquine versus placebo from linear regression model adjusting for weight change.

A number of pre-specified subgroups were examined, understanding that none were adequately powered (see Table 3). These analyses stratified the full study cohort by baseline ISI, HOMA-IR, BMI, and oral glucocorticoid use. None of the subgroups analyses demonstrated differences between treatment periods in any of the insulin and glucose metabolism measures. We also examined the results in patients based on their sequence of treatments, i.e., HCQ first or second, and this made no difference with respect to the primary outcome (data not shown).

Table 3. Mean (standard deviations) for measures of insulin and glucose metabolism in subgroups of interest.

	Baseline	Placebo	Change during placebo*	Hydroxy-chloroquine	Change during Hydroxychloroquine*	P-value†
HOMA-IR > 2.5 at baseline (n=5)
Insulin sensitivity Index	2.5 (0.44)	4.2 (1.7)	1.6 (1.7)	4.5 (1.8)	1.9 (1.9)	0.81
HOMA-IR	4.5 (2.1)	2.5 (1.2)	-1.9 (2.6)	2.7 (1.3)	-1.8 (2.8)	>0.99
HOMA-B	212.6 (137.0)	184.8 (154.2)	-27.8 (114.3)	168.5 (145.9)	-44.1 (96.1)	0.63
ISI > 8.0 at baseline (n=9)
Insulin sensitivity Index	12.2 (4.2)	10.3 (4.8)	-1.8 (3.9)	11.2 (5.5)	-0.96 (4.0)	0.73
HOMA-IR	1.1 (0.18)	1.2 (0.21)	0.06 (0.19)	1.1 (0.15)	-0.00 (0.13)	>0.99
HOMA-B	106.6 (96.6)	95.7 (39.1)	-10.9 (93.3)	87.6 (34.3)	-18.9 (80.5)	0.36
No oral glucocorticoid use at baseline (n=20)
Insulin sensitivity Index	7.7 (5.1)	7.8 (4.3)	0.08 (3.3)	7.8 (4.7)	0.08 (3.0)	0.76
HOMA-IR	2.1 (1.7)	1.6 (0.83)	-0.49 (1.5)	1.8 (1.1)	-0.31 (1.6)	0.36
HOMA-B	123.9 (105.9)	109.4 (87.3)	-14.5 (80.3)	113.0 (86.7)	10.8 (76.4)	0.58
BMI ≥ 30 at baseline (n=5)
Insulin sensitivity Index	5.0 (2.9)	5.1 (2.6)	0.07 (2.2)	6.1 (4.8)	1.1 (2.6)	>0.99
HOMA-IR	2.6 (1.5)	2.5 (1.1)	-0.07 (1.4)	2.9 (1.3)	0.30 (0.86)	0.44
HOMA-B	177.6 (132.9)	185.8 (146.1)	8.2 (80.8)	203.8 (118.8)	26.2 (49.9)	0.63
DAS28-CRP>3.2 at baseline (n=12)
Insulin sensitivity Index	5.9 (2.9)	6.5 (2.5)	0.5 (1.8)	6.1 (2.7)	0.2 (1.8)	0.38
HOMA-IR	2.0 (1.3)	1.8 (1.0)	-0.26 (0.9)	2.0 (1.2)	-0.03 (0.8)	0.89
HOMA-B	118.9 (98.0)	118.2 (109.7)	-0.7 (51.4)	134.5 (104.2)	15.6 (39.4)	0.62

^*Change calculated from baseline value under the assumption of no carry-over effect.

^†P-value for the difference between the values during hydroxychloroquine versus placebo using Wilcoxon signed-rank tests.

The baseline lipid profile results were in the normal range (see Table 1). During the HCQ treatment period, the levels of total cholesterol fell by 12.7mg/dl and LDL cholesterol by 12.4 mg/dl (see Table 2). When compared to the placebo treatment period, these reductions were statistically significant after adjusting for changes in body weight. Neither HDL nor triglyceride levels changed during HCQ treatment.

Discussion

While patients with RA have improved treatment options for the pain and functional limitations associated with this systemic inflammatory disease, they continue to suffer from important extra-articular manifestations such as cardiovascular disease. There is strong interest in determining the role of immunosuppressive treatment in managing the risk of cardiovascular disease and diabetes in patients with systemic inflammatory disease as well as the general population. Strong clinical and basic science data suggest that HCQ, an anti-malarial approved for use in RA and SLE, reduces glycated hemoglobin levels in diabetics and reduces the risk of incident diabetes.(6, 7) Based on these prior studies, we examined the effect of HCQ on insulin and glucose metabolism as well as lipid parameters and found no improvement in any parameters of insulin or glucose metabolism. Total cholesterol and LDL were slightly improved.

The results of this trial seem to disagree with the prior clinical studies of diabetes prevention as well as insulin and glucose metabolism involving HCQ and patients with RA. There are many plausible explanations for these apparent differences. First, our study population included patients with stable and mildly active RA at baseline as demonstrated by a low CRP. It is possible that HCQ may improve insulin and glucose metabolism in patients with more active RA than the subjects enrolled in this study. However, we found no differences in the subgroup with more active RA at baseline. As well, worse baseline HOMA-IR did not identify a subgroup of patients for which HCQ improved ISI, HOMA-IR or HOMA-B. The 14 patients with low insulin sensitivity at baseline (ISI ≤ 8) had no improvement in any of the indicators either. Second, it is possible that the duration of HCQ treatment did not allow us to observe an effect of HCQ on insulin and glucose metabolism. Our trial was adequately powered to observe a significant difference in ISI based on results observed in the pilot study (10) (see Statistical Analysis section). It is possible that a longer trial may have allowed an effect of HCQ to become apparent. Third, it is possible that the epidemiologic studies are confounded by unmeasured factors, such as RA disease activity or glucocorticoid differences. These variables may change the risk of diabetes among patients using HCQ but this may not be causal. Finally, we measured indices of insulin sensitivity, insulin resistance, and insulin secretion in this study. If insulin metabolism rather than insulin sensitivity or secretion is impacted by HCQ, these indices may not be reliable.

We included measurement of the lipid profile as a secondary outcome because of several prior epidemiologic studies that suggest improvements associated with HCQ…(14, 15) Since lipid measurements were not the primary outcome, the current study was not adequately powered or of long enough duration to draw strong conclusions.

There are many strengths of the current study. First, the study design was based on pilot data of shorter duration that demonstrated an improvement in ISI. Second, the cross-over design allows for excellent control of confounders because subjects serve as their own controls. Third, patients, providers, study staff and laboratory personnel performing blood tests were all blinded to which treatment period involved HCQ. Fourth, the study population was comprised of subjects with stable RA, allowing for the unique effects of HCQ upon insulin metabolism to be examined. Finally, the ISI is highly correlated with the gold standard euglycemic insulin clamp.(12)

Several potential limitations warrant discussion. It is possible that the subject group studied did not have enough baseline insulin resistance to observe improvement with HCQ. We tested this in a pre-planned, but underpowered, subgroup analysis and found no improvement even in subjects with HOMA-IRs > 2.5. Another possible limitation is the relatively short timeframe of this study. Finally, it is possible that the cross-over design could have obscured differences in the period on and off HCQ. While allowing for a washout period may have helped rule out such a possibility, the pilot study suggested no such washout period was required.

In conclusion, we conducted a 16-week randomized blinded cross-over trial testing the effect of HCQ on insulin and glucose metabolism and lipid parameters. No improvement in insulin or glucose metabolism was observed during 8-weeks of HCQ treatment. It is possible that HCQ may have a role in preventing diabetes, however this study's findings are not in line with prior epidemiologic studies that found a reduced risk of diabetes among patients with RA taking this medication. Longer term studies focused on patients with insulin resistance may document a role for HCQ in diabetes prevention among patients with RA.

Significance and Innovation.

• HCQ for 8 weeks in non-diabetic patients with RA produced no significant change in insulin sensitivity or resistance.

• Small and statistically significant improvements in total and LDL cholesterol were observed during HCQ treatment.