Changes in Glycated Hemoglobin after Initiation of Hydroxychloroquine or Methotrexate in Diabetic Patients with Rheumatologic Diseases

Abstract

Objective

Prior research demonstrates that hydroxychloroquine (HCQ) lowers glycated hemoglobin (HbA1c) in non-rheumatic diabetic patients. We examined medical records of patients with diabetes mellitus (DM) and concomitant rheumatic illness to measure changes in HbA1c after starting HCQ or methotrexate (MTX).

Methods

We utilized electronic medical records to identify patients initiating either HCQ or MTX, with a diagnosis of DM or HbA1c ≥ 7%, and with at least one HbA1c measurement before and within 12 months after initiation. A structured medical record abstraction examined rheumatologic diagnosis, use of oral glucocorticoids, body mass index, age, gender. Adjusted linear regression models determined changes in HbA1c from pre-drug values to the lowest post-drug values within twelve months.

Results

We identified 45 HCQ users and 37 MTX users who met inclusion criteria. Half in each group carried a rheumatoid arthritis diagnosis. Age, gender, and mean pre-drug HbA1c levels were similar across groups. Mean BMI for HCQ users (35.4 kg/m²) was slightly higher than MTX users (32.2 kg/m²) (p = 0.13). Glucocorticoid use appeared more common in MTX (46%) than HCQ (29%) users (p = 0.17). The mean reduction in HbA1c between pre- and post-HCQ was 0.66% (95% CI 0.26–1.05) versus 0.11% (95% CI −0.18 – 0.40) for MTX. In fully adjusted analyses, the reduction in HbA1c among HCQ users was 0.54% greater than the drop among MTX users (p = 0.041).

Conclusion

HCQ initiation was associated with a significantly greater reduction in HbA1c as compared to MTX initiation among diabetic patients with rheumatic disease.

People with rheumatic diseases, such as rheumatoid arthritis (RA) or systemic lupus erythematosus (SLE), are at an increased risk of developing cardiovascular disease (CVD) [1–4]. Cardiovascular disease is a leading cause of death in individuals with rheumatic diseases; however the risk factors for CVD in this population are not clear. Both traditional risk factors and inflammation likely contribute to CVD [5–7]. Current management guidelines for CVD in rheumatic disease suggest measures applicable to the general population, with the addition of slightly more aggressive risk stratification [8–10]. It would be ideal if management strategies for CVD and its risk factors could be tailored for rheumatic disease patients.

Several risk factors – both traditional and those specific to rheumatic conditions – may indicate an elevated risk of CVD in rheumatic disease patients [5–7]. Although there is no consensus regarding the risk of diabetes mellitus (DM) in RA [11, 12], it is clear that patients with rheumatic disease experience a substantially elevated risk of insulin resistance [13, 14]. Insulin resistance refers to a state of impaired insulin sensitivity and glucose metabolism, which commonly precedes development of DM [15]. It is strongly associated with the metabolic syndrome, defining one of the WHO criteria for this constellation of CVD risk factors, and predicts future CVD events [16].

It has been shown that inflammation directly influences insulin and glucose metabolism through cytokines such as TNFα and IL-6 [13, 17–19]. Considering this, several disease modifying anti-rheumatic drugs (DMARDS) – agents prescribed to primarily treat rheumatic diseases – have also been tested as interventions to improve insulin and glucose metabolism. Anakinra – an IL-1 receptor antagonist FDA approved for the treatment of RA – was examined as a therapy for DM, and associated with significantly lower glycated hemoglobin (HbA1c) [20]. Several TNFα antagonists have been shown to improve insulin metabolism in patients with RA or ankylosing spondylitis [17–19]. In addition, hydroxchloroquine (HCQ), an FDA approved DMARD for the treatment of RA and lupus, has been found in at least two randomized controlled trials to improve diabetes control in non-rheumatic subjects [21, 22]. One large epidemiologic study also showed that HCQ reduces the risk of DM among patients with RA [23].

It is unclear whether the effects of HCQ seen among diabetics would generalize to rheumatic disease patients. As well, it is not clear whether other, more potent DMARDs, such as methotrexate (MTX), would affect HbA1c. Considering these questions, we compared the effect of HCQ with that of MTX on HbA1c in diabetic adults under treatment for a rheumatologic condition

PATIENTS AND METHODS

Study Cohort

Adult subjects were identified from a large academic medical center’s clinical data repository. This clinical data repository includes information on patients’ charges, diagnoses, procedures, laboratory values, prescribed medications, medical history, and sociodemographics.

One author (LRR) abstracted the medical records of potentially eligible subjects using a structured review. Items of interest included age, gender, body mass index (BMI), rheumatologic diagnosis, DM diagnosis, history of insulin use or non-insulin DM medication, and history of oral corticosteroid use.

We searched the repository for patients aged 18 years and older, with a diagnosis of DM (or a pre-treatment HbA1c of 7% or greater) [24] with two or more HbA1c values available; and who had initiated HCQ but not MTX, OR had initiated MTX but not HCQ. From this group of potentially eligible subjects (n = 537) we reviewed full medical records, and excluded: those without one pre-treatment HbA1c and without one post treatment HbA1c (n = 391); those taking both HCQ and MTX (n = 26); and those who lacked a pre-treatment HbA1c of 7% or higher or a clinical diagnosis of DM (n = 38).

The Partners HealthCare System Institutional Review Board approved all aspects of this study.

Exposure

The exposures of interest were initiation of either HCQ or MTX; concomitant users were excluded. Information about HCQ or MTX initiation was derived from physician notes as well as the electronic medication list. We made no distinctions based on the dosage of these medications. While simultaneous use of HCQ and MTX was not permitted, subjects were commonly prescribed other anti-rheumatic therapies during the same period; these cases were included. In addition, glucocorticoid use was not excluded.

Outcome of Interest

The primary outcome of interest was change in HbA1c, from the measurement most proximal prior to HCQ or MTX initiation to at least 12 weeks after initiation. Two different post-exposure HbA1c values were examined: the most proximal to HCQ or MTX initiation and the lowest within 12 months following initiation. The HbA1c values were derived from routine clinical laboratory practice, which uses a Tosoh HLC-723G8 Analyzer instrument to measure HbA1c (coefficient of variation < 0.5%) [25].

Statistical Analysis

Descriptive statistics such as mean, median, standard deviation for each continuous variable, and frequencies for each categorical variable were used to summarize the data. The normality of the distribution of primary outcomes of interest was examined with a normal probability plot. Two sample t-tests, Chi-squared tests, Fisher’s exact tests or non-parametric tests were used for baseline comparisons between the two exposure groups when applicable. The change in HbA1c was calculated separately for HCQ and MTX groups, as was HbA1c_PRE – HbA1c_POST. First, the change in HbA1c within each group was assessed using paired t-tests or Wilcoxon signed rank tests. We then compared changes in HbA1c across drugs using general linear regression adjusted for rheumatologic diagnosis, cumulative steroid use, duration (months) between drug initiation and lowest HbA1c, a change in DM medication, body mass index, age, and gender. Data analyses were performed using SAS 9.2 (SAS Institute, Inc. Cary, North Carolina).

RESULTS

The baseline characteristics of the 82 subjects (45 HCQ users, and 37 MTX users) who are included in the analyses are shown in Table 1. The mean age for both groups was 61 years and most patients were female: 82% for HCQ and 73% for MTX. Approximately half of the subjects in each treatment group carried a diagnosis of RA (53% for HCQ and 57% for MTX); 7% of the HCQ treated subjects were diagnosed with SLE, compared with 0% for MTX. There was a trend toward more frequent oral glucocorticoid use in the twelve months following drug initiation in MTX (46%) than in HCQ (29%) (p = 0.17), as well as a slightly higher cumulative dose in MTX (612±982 grams) than in HCQ (464±809 grams) (p = 0.46). Mean BMI for HCQ users (35.4 kg/m²) was slightly higher than MTX users (32.2 kg/ m²) (p = 0.13). A physician’s diagnosis of DM was recorded in 89% of HCQ users and 97% of MTX users, with roughly 30% in both groups prescribed insulin therapy and close to 60% in both groups prescribed non-insulin DM medication.

Table 1.

Baseline Characteristics of Diabetic Adults with Rheumatic Disease at Time of New Initiation of Hydroxychloroquine or Methotrexate

	Hydroxychloroquine n = 45	Methotrexate n = 37	P value
	Mean ± SD or %
Age, mean ± SD, years	61 ± 13	61 ± 13	0.79
Female gender, %	82%	73%	0.20
Body mass index	35.4±8.5	32.2±6.1	0.13
Rheumatoid arthritis, %	53.3%	56.8%	0.95
Systemic lupus erythematosus, %	6.7%	0	0.50
Diabetes mellitus, by physician diagnosis %	88.9%	97.3%	0.22
Prescribed insulin, n, %	33.3%	29.7%	0.75
Prescribed non-insulin diabetes medication, n, %	62.2%	59.5%	0.82
Cumulative* steroid use, %	28.9%	46.0%	0.17
Cumulative* steroid use, grams	464±809	612±982	0.46

^*Cumulative refers to the 12 months after initiation of HCQ or MTX.

Mean pre-drug initiation HbA1c values were similar in both groups (HCQ 7.71% and MTX 7.38%, p = 0.35) (Table 2). The most proximally recorded HbA1c after initiation of drug were similar (HCQ 7.28% and MTX 7.47%, p = 0.58), as were the absolute values for the lowest HbA1c values within twelve months after treatment initiation (HCQ 7.05% and MTX 7.27%, p = 0.49). However, a comparison of the change in HbA1c from pre treatment to lowest values post treatment revealed a significant reduction with HCQ (0.66%; 95% CI 0.26 – 1.05), as compared to the significantly smaller reduction observed with MTX (0.11%; 95% CI −0.18 – 0.40) (p = 0.04) (Figure 1). In adjusted analyses this drop in HbA1c among HCQ users was 0.54% greater than that in MTX users (p = 0.041).

Table 2.

Change in HbA1c with Initiation of Hydroxychloroquine or Methotrexate, in Diabetic Adults with Rheumatic Disease

	Hydroxychloroquine n = 45	Methotrexate n = 37	P value*
	Mean ± SD
Pre-drug HbA1c	7.71±1.46	7.38±1.66	0.35
Most proximal post-drug HbA1c	7.28±1.63	7.47±1.40	0.58
Lowest post-drug HbA1c within 12 mos.	7.05±1.54	7.27±1.34	0.49
ΔHbA1c (Pre-drug minus lowest post-drug)	0.66±1.31	0.11±0.87	0.04

^*P-values from general linear regression adjusted for rheumatologic diagnosis, cumulative steroid use, duration (months) between drug initiation and lowest HbA1c, a change in DM medication, body mass index, age, and gender.

Figure 1.

Effect of HCQ versus MTX on HbA1c within 12 months post drug initiation in diabetic adults with rheumatologic condition. P-value compares pre-treatment HbA1c to the lowest HbA1c within 12 months post-treatment; p-value calculated from a paired t-test. Horizontal bar indicates sample median.

DISCUSSION

The literature suggests that HCQ may possess unique hypoglycemic effects, in addition to its remittive actions for rheumatic diseases. To better understand whether the hypoglycemic effects of HCQ are in part related to its anti inflammatory properties or another mechanism, we examined the potential hypoglycemic benefits of MTX, another DMARD commonly used to treat RA and other rheumatic diseases. In this study, we examined the change in HbA1c among patients with rheumatic diseases initiating either drug; all patients had DM with a pre-drug HbA1c of 7% or higher. Comparing pre-drug levels with those within one year following drug initiation, HCQ produced a significant reduction in HbA1c that was larger than the change associated with MTX. Patients starting MTX did not experience a significant reduction in HbA1c, but relatively few subjects starting MTX were studied.

While inflammation is known to be associated with impaired glucose control, the mechanism by which HCQ exerts a hypoglycemic effect is not entirely clear. Chloroquine has been shown to increase C-peptide response, which Gerstein et al. cite as a potential effect of improved beta cell functioning with decreased blood glucose [22, 26]. Additionally, HCQ’s inhibitory effect on insulin metabolism has been demonstrated in animal models, with effects including reductions in intracellular insulin degradation, and increases in insulin accumulation [27, 28]. Whether due to a novel mechanism or the established anti-inflammatory effect of this therapy, the results from this study align with the clinically observed responses detailed in epidemiologic studies and case reports.

Data from our study represent novel findings, highlighting HCQ’s potential ability to decrease HbA1c in diabetic persons with systemic inflammatory disease. Several previous studies in non-rheumatic patients with DM have demonstrated the hypoglycemic effect of HCQ as well. HCQ therapy contributed to significant improvements in type 2 DM management, as compared with a placebo, in a small group of treatment refractory individuals in a randomized trial [21]. Subsequently, HCQ was found in another randomized trial to improve glucose control and HDL cholesterol in a larger cohort of type 2 diabetics [22]. In addition, a large epidemiologic study among patients with RA found that HCQ was associated with a reduced risk of incident DM, even after controlling for disease activity and glucocorticoid use [23]. Considering these prior data and the significant effect we observed with HCQ, our findings generate interesting questions about the potential long-term role of this therapy, apart from its effect on arthritis symptoms, in treating systemic rheumatic disease.

Our study has several limitations. As a non-randomized retrospective chart review, there is the possibility for confounding by unmeasured factors, including changes in diabetes medications and BMI. The data was collected on a limited number of patients at one academic medical center, potentially limiting its generalizability. In addition, information on rheumatic disease activity was not collected, and only extremely limited data on concomitant glucocorticoid use was available. Finally, the study was small and not specifically powered to detect a difference in MTX; the null effect may be an artifact of the small sample size.

In the last decade, there has been a growing awareness of the increased cardiovascular morbidity associated with RA and SLE. The reasons for this observation are in part related to inflammation as well as to traditional risk factors like diabetes mellitus and impaired insulin resistance. We wanted to better understand whether HCQ might improve glucose control in rheumatic disease patients and begin to explore whether HCQ’s effects were related to a reduction in inflammation, or to another, independent effect of HCQ. We examined the effects of HCQ and MTX, two commonly used DMARDs, on HbA1c among rheumatic disease patients. HCQ significantly lowered glycated hemoglobin; MTX did not, but the sample size was limiting. These findings add to an existing literature suggesting beneficial effects of HCQ on insulin and glucose metabolism, perhaps distinct from MTX. Comprehensive, larger, and prospective studies are clearly warranted to determine the potential role of HCQ in mitigating cardiovascular morbidities in patients with systemic rheumatic diseases.