Efficacy and safety of hydroxychloroquine for managing glycemia in type-2 diabetes: A systematic review and meta-analysis.

Aims: No Cochrane meta-analysis with grading of evidence is available on use of hydroxychloroquine (HCQ) in type-2 diabetes (T2DM). This meta-analysis evaluated the efficacy and safety of HCQ in T2DM.
Methods: Electronic databases were searched using a Boolean search strategy: ((hydroxychloroquine) OR (chloroquine*)) AND ((diabetes) OR ("diabetes mellitus") OR (glycemia) OR (glucose) OR (insulin)) for studies evaluating hydroxychloroquine for glycemic control in T2DM. The primary outcome was a change in glycated haemoglobin (HbA1c). The secondary outcomes were changes in other glycemic/lipid parameters and adverse effects.
Results: Data from 11 randomized controlled trials (RCTs) (3 having placebo as controls [passive controls] and 8 having anti-diabetes medications as controls [active controls]) involving 2,723 patients having a median follow-up of 24 weeks were analyzed. About 54.54% of the RCTs were of poor quality as evaluated by the Jadad scale. The performance bias and detection bias were at high risk in 63.64% of the RCTs. The HbA1c reduction with HCQ was marginally better compared to the active (mean differences [MD]-0.17% [95%, CI:-0.30--0.04;P=0.009;I2=89%; very low certainty of evidence, VLCE]), and passive (MD-1.35% [95%CI:-2.10--0.59;P=0.005;I2=74%]) controls. A reduction in fasting glucose (MD-16.63mg/dL[95%, CI: -25.99 - -7.28mg/dL;P<0.001;I2=97%;VLCE]) and post-prandial glucose [MD -8.41mg/dL (95%CI: -14.71 - -2.12mg/dL;P=0.009;I2=87%;VLCE]), appeared better with HCQ compared to active controls. The total adverse events (risk ratio [RR]0.93 [95% CI:0.68-1.28]; P=0.65;I2=66%) were not different with HCQ compared to the controls.
Conclusion: The routine use of HCQ in T2DM cannot be recommended based on the current evidence.

Introduction

Hydroxychloroquine (HCQ) has been extensively used in different rheumatologic disorders for more than half a century now.[1] The studies in patients with lupus, Sjogren syndrome, and rheumatoid arthritis have highlighted the mild anti-hyperglycemic properties of HCQ.[12] People with autoimmune disorders on HCQ receiving glucocorticoids had a lower incidence of diabetes.[12] A reduction in systemic inflammation, a favorable impact on adipocytokines, and a reduction in insulin degradation are some of the possible mechanisms attributed to the glycemic benefits of HCQ.[3] HCQ has also been used as an immunomodulator for the treatment and prophylaxis in the recent coronavirus 2019 (COVID-19) pandemic.[4] HCQ retinopathy is a dreaded, but, fortunately, a rare complication of the HCQ therapy, typically seen in people on HCQ for decades at doses more than 5 mg/kg/day.[3] HCQ has been recently approved as a third-line agent for managing glycemia in people with type-2 diabetes (T2DM) in India.[3] However, it must be realized that no other country in this world has included HCQ in their guidelines for managing T2DM or has made a recommendation for the use of HCQ in T2DM. T2DM is a chronic disorder needing decades of treatment, especially in a country like India, where the peak incidence of T2DM is nearly two decades earlier than in the western world.[5] With the peak incidence of T2DM in India being in the thirties, and the average life span of the Indians being 75 years, many of the people living with T2DM may end up receiving HCQ for up to 40 years.[67] A few randomized controlled trials (RCTs) and reviews have been published evaluating the role of HCQ for managing glycemia in T2DM.[8] However, a literature search reveals that to date, no Cochrane meta-analysis with the grading of evidence is available which holistically evaluated the efficacy and safety of HCQ in managing T2DM. We undertook this meta-analysis to address this knowledge gap.

Methods

The meta-analysis was carried out according to the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions.[9] The predefined protocol has been registered with the international prospective register of systematic reviews (PROSPERO) having registration number: CRD42021227109. All the RCTs satisfying the inclusion criteria, published till December 2020 were considered This meta-analysis has been reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).[10] Since ethical approval already exists for the individual studies included in the meta-analysis, no separate approval was required for this study.

The Population, Intervention, Comparison, Outcomes and Study type (PICOS) criteria were used to screen and select the studies for this meta-analysis with patients (P) being individuals with T2DM; intervention (I) being use of HCQ over the background of standard care for T2DM; control (C) being the patients with T2DM on standard care for managing glycemia but not receiving HCQ; outcomes (O) being evaluated, and study type (S) being RCTs for this meta-analysis. Only those RCTs which had at least two arms were included, with the intervention arm receiving HCQ on the background of standard care for T2DM and the non-intervention or control arm receiving placebo or any other approved medication for T2DM were considered.

The primary outcome of the meta-analysis was to evaluate the changes in HbA1c. The secondary outcomes of this study were to evaluate the alterations in fasting plasma glucose (FPG), 2 h post-prandial glucose (PPG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), very low-density lipoprotein cholesterol (VLDL-C), bodyweight change, discontinuation of medication due to adverse events, and any other adverse events as described by the authors and death (all-cause). The RCTs whose outcomes evaluated the primary end-point or at least two secondary end-points were included in the meta-analysis. A sub-group analysis was done based on whether the control group received an active comparator (another anti-diabetes medication)—labeled here as the active control group (ACG) or a placebo—labeled as the passive control group (PCG).

Search method for identification of studies

A detailed search of the electronic databases of Medline (via PubMed), Embase (via Ovid SP), Cochrane central register of controlled trials (CENTRAL) (for trials only), ctri.nic.in, clinicaltrials.gov, global health, and Google Scholar was done using a Boolean search strategy: ((hydroxychloroquine) OR (chloroquine*)) AND ((diabetes) OR (“diabetes mellitus”) OR (glycemia) OR (glucose) OR (insulin)).

Data extraction and study selection

Data extraction was carried out independently by two authors using the standard data extraction forms. The details have been elaborated in another meta-analysis carried out by our group and published elsewhere.[11]

Assessment of risk of bias in the included studies

Three authors independently assessed the risk of bias using the risk of the bias assessment tool in the Review Manager (RevMan) version 5.3 (The Cochrane Collaboration, Oxford, UK, 2014) software. We specifically looked for selection bias, performance bias, detection bias, attrition bias, reporting bias and any other bias like publication bias. The details of how risk of bias assessment was done has already been elaborated elsewhere.[11] Quality assessment of the included studies was also conducted using the Jadad scale which consisted of three domains: randomization (0–2 points), blinding (0–2 points), and an account of all the patients (0–1 point). We classified the quality of RCTs as good (4–5 points), fair (3 points), or poor (0–2 points).[12]

Measures of treatment effect

For continuous variables, the outcomes were expressed as MD. The final results were presented both in conventional units as well as SI units. For dichotomous outcomes the results were expressed as risk ratio (RR) with 95% CI.

Assessment of heterogeneity

Heterogeneity was initially assessed by studying the forest plot generated. Subsequently, heterogeneity was analyzed using a Chi-square test on N-1 degrees of freedom, with an alpha of 0.05 used for statistical significance and with the I test.[7] The detail of assessment and interpretation of heterogeneity has already been elaborated elsewhere.[11]

Grading of the results

An overall grading of the evidence related to each of the primary and secondary outcomes of the meta-analysis was done using the GRADE (Grades of Recommendation, Assessment, Development, and Evaluation) approach.[13] The details of how grading of the study results was done and how the summary of findings (SoF) table was developed [Table1], has been elaborated elsewhere.[11] We specifically looked for publication bias by plotting the Funnel Plot. Presence of one or more study outside the inverted funnel plot was proof of significant publication bias [Supplementary Figure1].

Table 1

Summary of findings

HYDROXYCHLOROQUINE compared to CONTROL for managing glycaemia in type-2 diabetes: A meta-analysis
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
	Risk with CONTROL	Risk with Hydroxychloroquine
HbA1c ACG	The mean hbA1c ACG was 8.41%	MD 0.17% lower (0.3 lower-0.04 lower)	-	2334 (8 RCTs)	⨁◯◯◯ VERY LOW†,‡,§
Fasting Glucose ACG	The mean fasting Glucose ACG was 153.04 mg/dl	MD 16.63 mg/dl lower (25.99 lower-7.28 lower)	-	2334 (8 RCTs)	⨁◯◯◯ VERY LOW†,‡,§
Post-prandial Glucose ACG	The mean post-prandial Glucose ACG was 267.12 mg/dl	MD 8.41 mg/dl lower (14.71 lower-2.12 lower)	-	2312 (7 RCTs)	⨁◯◯◯ VERY LOW†,‡,§
Total Cholesterol ACG	The mean total Cholesterol ACG was 169.73	MD 5.78 lower (9.52 lower-2.04 lower)	-	1154 (4 RCTs)	⨁⨁⨁◯ MODERATE†
Total Adverse Events	212 per 1,000	197 per 1,000 (144-271)	RR 0.93 (0.68-1.28)	2723 (11 RCTs)	⨁⨁◯◯ LOW †,¦
Hypoglycaemia	168 per 1,000	136 per 1,000 (73-243)	OR 0.78 (0.39-1.59)	1974 (10 RCTs)	⨁⨁◯◯ LOW†,¦
Weight	The mean weight was 70.23 kg	MD 0.54 kg lower (1.11 lower-0.03 higher)	-	2046 (7 RCTs)	⨁⨁◯◯ LOW†,‡

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; MD: Mean difference; RR: Risk ratio; OR: Odds ratio; ACG: Active control sub-group (people receiving any anti-diabetes medication as the active control in the control group. GRADE Working Group grades of evidence. High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect. † Performance bias (blinding of participants and investigators) and detection bias (blinding of outcome assessors) was judged to be at high risk of bias in 6 out of 10 studies (60%). ‡ Due to a large variation in the effect, the confidence intervals do not overlap, theP-value for heterogeneity is <0.01, and I2 is <90%. § High publication bias suspected as evidenced by the Funnel Plot. ¦ Due to large variation in effect, the confidence intervals do not overlap, theP-value for heterogeneity is <0.05, and I2 is <60%

Data synthesis

Data were pooled as a random effect model for the analysis of outcomes which were expressed as 95%CI. Forrest plots were plotted with the left side of the graph favoring HCQ and the right side of the graph favoring control using the RevMan 5.3 software.

Results

A total of 471 articles were found after the initial search [Figure 1]. Following the screening of the titles, abstracts, the search was reduced down to 21 studies whose full texts were evaluated in detail for inclusion in this meta-analysis [Figure 1]. Eleven RCTs in people with T2DM which fulfilled all criteria were analyzed in this meta-analysis (eight from India, one each from the USA, Canada, and Italy).[1516171819202122232425] The remaining 10 studies were evaluated in detail but were excluded as they did not fulfill the inclusion criteria of this meta-analysis.[26272829303132333435] One RCT in prediabetes and another RCT focusing on dyslipidemia instead of glycemia were excluded.[2627] The studies by Gupta A,[28] Baidya A et al. (2018),[29] Baidya A et al. (2018a),[32] and Chandra AK et al.[33] evaluating HCQ in T2DM were excluded as they did not have valid control groups. The study by Jagnani VK et al.[31] was excluded as it was a retrospective study without an appropriate control group.

Figure 1

Flowchart elaborating on study retrieval and inclusion in the meta-analysis. RCT: randomized controlled trial; HCQ: hydroxychloroquine; COVID: novel coronavirus 2019

Of the 11 studies included, in 3 studies (Chakravarti et al.[15] Gerstein et al.[16] and Quatraro et al.[19]), the control groups were receiving placebo, and hence, have been sub-grouped under the passive control sub-group (PCS). The duration of the follow-up of 12, 24, and 24 weeks in the studies by Chakravarti et al.[15], Gerstein et al.[16] and Quatraro et al.[19], respectively, was used for analysis. In the remaining eight studies, the control groups were receiving anti-diabetes medication as the comparator agent, and hence, have been sub-grouped under the active control sub-group (ACS). The active comparator was pioglitazone 15 and 45mg/d in the studies by Pareek et al. [17] and Hsia et al.[18] respectively. The duration of the follow-up was 24 and 16 weeks in the studies by Pareek et al.[17] and Hsia et al.[18] respectively. Vildagliptin 100mg/d was the active comparator in the study by Baidya et al.[20] Teneligliptin 20mg/d was the active comparator in the studies by Ranjan et al.[22] and Kumar (2019) et al.[23] Sitagliptin 100mg/d was the active comparator in the studies by Kumar (2018) et al.[21] Singh (2018) et al.[24] and Singh (2018a) et al.[25] The duration of the follow-up was 24 weeks in the studies by Baidya et al. (2018c),[20] Kumar et al. (2018),[21] Kumar et al. (2019),[23] Ranjan et al.[22] Singh et al.(2018),[24] and Singh et al. (2018a).[25] The dose of HCQ used in the studies by Chakravarti et al.[15], Pareek et al.[17], Hsia et al.[18], Baidya et al. (2018c)[20], Ranjan et al.[22], Kumar (2018) et al.[21], Kumar (2019) et al.[23], Singh (2018) et al.[24] and Singh (2018a) et al.[25] was 400mg/d. In the studies by Gerstein et al.[16] and Quatraro et al.[19], the dose of HCQ used was 300–600 and 600mg/d, respectively. The details of all the RCTs included have been elaborated in Supplementary Table 1. The characteristics of the studies excluded are given in Supplementary Table 2.

Supplementary Table 1

Patients characteristics of randomized controlled trials on use of hydroxychloroquine in type-2 diabetes in this meta-analysis

Author, year, (Reference)	Study design, Study duration, Study population	Add on Intervention (I), Comparator (C)	Number of study subjects Recruited/analysed	Age (years)	Duration of diabetes (years)	Body weight (kg)	Glycemic parameters: Hba1c (%), FPG (mg/dl), PPG (mg/dl)	Lipid parameters in mg/dl: T-chol, LDL-c, HDL-c, VLDL-c, TG	Other points
Chakravarti et al., 2020, [14] India	Double blind RCT, 12 wk & 12 wk follow up, T2DM (HbA1c ≥7-8.5%), aged 18-75 yrs, on glimepride 4mg + metformin 500 mg	I(1): HCQ 200 mg OD I(2): HCQ 300 mg OD I(3): HCQ 400 mg OD C: Placebo	I(1): 66/61 I(2): 66/61 I(3): 128/121 C: 66/61	I(1): 53.36±8.5 I(2): 51.76±8.0 I(3): 52.96±9.6 C: 53.36±7.8	I(1): 5.66±5.0 I(2): 6.16±4.7 I(3): 6.46±5.7 C: 6.46±5.0	I(1): 59.6±6.7 I(2): 61.6±5.4 I(3): 70.6±8.2 C: 62.6±6.1	HbA1c: I(1): 7.83±0.5 I(2): 7.89±0.5 I(3): 7.93±0.5 C: 7.72±0.5 FPG: I(1): 164±38 I(2): 168±42 I(3): 170±45 C: 157±32 PPG: I(1): 275±53 I(2): 278±56 I(3): 282±51 C: 269±58	NR
Hsia et al., 2019, [17] USA	Open label RCT, 16 weeks, T2DM (HbA1c ≥7.5% to <11.0%), age 18-75 years, taking maximally tolerated metformin and sulfonylurea for ≥3 months	I: HCQ 400 mg OD C: Pioglitazone 45 mg OD	I: 15/15 C: 7/7	I: 53±10 C: 57±13	I: 5.1±3.1 C: 7.8±4.3	I: 85.7±14.3 C: 83.7±18.2	HbA1c: I: 8.6±1.0 C: 9.1±0.7 FPG: I: 179±53 C: 170±50 PPG: NR	T-chol: I: 170.2±32.6; C: 149.3±30.1; LDL-c: I: 94.8±29.7; C: 72.9±27.7; HDL-c: I: 40.8±7.7; C: 40.6±6.3; VLDL-c: NR; TG: I: 188±132; C: 178±80	Also measured baseline insulin, hsCRP, WBC, ALT, AST, eGFR
Kumar et al., 2019, [22] India	Prospective Open label RCT, 24 weeks, T2DM (HbA1c ≥7.5%), body weight ≥60 kg, on metformin 1000 mg/day and glimepiride 2mg/day for at least 2 weeks	I: HCQ 400 mg OD C: Teneligliptin 20 mg OD	I: 90/90 C: 90/90	I: 66±9 C: 66±7	I: 6±2 C: 6±3	I: 70±8 C: 72±13	HbA1c: I: 8.1±0.3 C: 8.2±0.2 FPG: I: 169±18 C: 171±16 PPG: I: 232±18 C: 239±21	NR	Also measured creatinine
Baidya et al., 2018, [19] India	Prospective Open label RCT, 24 weeks, T2DM (HbA1c ≥7.0% and ≤9%), body weight ≥60 kg, on stable dose of metformin 1000 mg/day and glimepiride 2mg/day for at least 12 weeks	I: HCQ 400 mg OD C: Vildagliptin 100 mg/d	I: 50/50 C: 50/50	I: 58.36±8.59 C: 56.24±7.40	I: 10.23±6.27 C: 10.19±6.18	I: 83.16±11.46 C: 82.61±10.87	HbA1c: I: 7.8±0.9 C: 7.9±0.8 FPG: I: 149.6±28.4 C: 150.43±19.8 PPG: I: 240.4±44.4 C: 241.3±38.5	NR	Also measured creatinine, ALT
Kumar et al., 2018, [20] India	Open label randomized observational study, 24 weeks, T2DM (HbA1c ≥7.5% and ≤10%), aged 18-65 years, body weight	I: HCQ 400 mg OD C: Sitagliptin 100 mg OD	I: 338/300 C: 343/300	I: 58.3±9.1 C: 57.2±9.3	I: 12.5±4.2 C: 12.8±3.9	I: 67.8±4.2 C: 67.6±4.1	HbA1c: I: 8.6±0.5 C: 8.7±0.5 FPG: I: 179.1±59.6 C: 175.8±51.6	NR	Also measured creatinine
	≥60kg, serum Cr <1.5mg/dL, on at least 2000mg/d of metformin and 2mg/d glimepiride with stable insulin dose and following routine diet and exercise regime for at least 12 weeks						PPG: I: 292.1±66.4 C: 290.9±68.0
Ranjan et al., 2018, [21] India	Prospective RCT, 24 weeks, T2DM (HbA1c 7.5-10%), aged 18-80 years, BMI 22-45 kg/m2, on metformin and glimepride, and insulin (>30 units/d for atleast 4 weeks)	I: HCQ 400 mg OD C: Teneligliptin 20 mg OD	I: 160/148 C: 160/152	I: 55±8 C: 56±9	I: 9±3 C: 10±2	NR	HbA1c: I: 8.3±0.5 C: 8.2±0.5 FPG: I: 134.03±16.44 C: 135.07±16.51 PPG: I: 239.89±21.5 C: 234.72±23.1	T-chol: I: 159.71±25.43 C: 163.08±26.94 LDL-c: I: 104.54±18.57 C: 105.21±16.61 HDL-c: I: 46.64±10.48 C: 49.69±13.34 VLDL-c: NR TG: I: 132.81±35.80 C: 131.87±32.91
Singh et al., 2018, [23] India	Observational RCT, 24 weeks, T2DM (HbA1c 7-8%), adults, body weight ≥60 kg, on fixed dose of 500 mg/d metformin and 80 mg/d gliclazide for atleast 12 weeks	I: HCQ 400 mg OD C: Sitagliptin 100 mg OD	I: 155/150 C: 155/150	I: 60±11 C: 60±9	I: 7.4±5.5 C: 7.4±5.3	I: 64.05±9.68 C: 64.72±9.65	HbA1c: I: 7.76±0.4 C: 7.77±0.4 FPG: I: 142.01±19.27 C: 142.55±20.88 PPG: I: 261.21±29.82 C: 262.38±31.26	NR
Singh et al., 2018a, [24] India	Open label randomized real world observational study, 24 weeks, T2DM (HbA1c ≥7%), age 38-65 years, on stable doses of sulfonylurea with ≥1000 mg/d metformin for atleast 3 months	I: HCQ 400 mg OD C: Sitagliptin 100 mg OD	I: 320/300 C: 320/300	I: 49.2±5.1 C: 53.1±7.9	Atleast 2 years, exact values NR	I: 76.3±9.6 C: 72.8±8.5	HbA1c: I: 8.3±1 C: 8.3±1 FPG: I: 147.7±24 C: 149.8±26.3 PPG: I: 278.3±36 C: 278.8±36	T-chol: I: 176.9±25.2 C: 172.9±25.4 LDL-c: I: 109.6±20.2 C: 107.4±20.2 HDL-c: I: 44.5±8.5 C: 44.5±8.5 VLDL-c: NR TG: I: 162.3±58.4 C: 159.3±58.4	Also measured SBP, DBP, creatinine, eGFR, hsCRP, insulin, HOMA-IR
Pareek et al., 2014, [16]India	Double blind multicentric RCT, 24 weeks, T2DM (HbA1c ≥7.5% to <11.5%), age 18-65 yrs,	I: HCQ 400 mg OD C: Pioglitazone 15 mg OD	I: 135/115 C: 132/117	I: 52.60±8.55 C: 52.23±8.35	I: 4.23±3.80 C: 4.02±3.64	I: 69.86±9.13 C: 69.30±8.53	HbA1c: I: 9.2±1.2 C: 9.1±1.1 FPG: I: 180.9±51.3	T-chol: I: 175.17±41.76 C: 181.75±42.54 LDL-c: I: 99.38±38.28	Also measured baseline waist circumference, SBP, DBP
	post 3 months treatment with ≥1000 mg/d metformin, and ≥4 mg/d glimepiride or ≥160 mg/d gliclazide						C: 174.2±57.4 PPG: I: 261.5±66.2 C: 259.9±67.9	C: 103.63±39.06 HDL-c: I: 44.08±10.44 C: 43.7±13.15 VLDL-c: NR TG: I: 171.83±99.2 C: 177.15±108.06
Gerstein et al., 2002, [15] Canada	Double blind RCT, 18 months (6month’s data used for analysis), T2DM (HbA1c ≥11%), age 35-80 years, BMI >25, on at least 2 months therapy on maximal doses of sulfonylurea either alone or in combination with metformin	I: HCQ 300 mg OD, max 300 mg BD C: Placebo	I: 69/69 C: 66/66	I: 58±9.6 C: 57±10.1	I: 8.5±5.2 C: 8.9±5.8	NR	HbA1c: I: 13.6±2.3 C: 13.3±2.0 FPG: NR PPG: NR	Baseline values NR Only reduction in T-chol and LDL-c reported
Quatraro et al., 1990, [18] Italy	Double blind RCT, 24 weeks, T2DM, adults, overweight, on maximal doses of OHA and/or insulin	II: HCQ 200 mg TDS CI: Placebo IG: HCQ 200 mg TDS CG: Placebo	II: 11/11 CI: 11/11 IG: 8/8 CG: 8/8	II: 57±5.3 CI: 58±5.6 IG: 58±2.9 CG: 57±3.4	II: 10.9±3.3 CI: 11.2±3.6 IG: 11.3±2.8 CG: 11.1±3	NR	Hba1c: II: 12.5±1.98 CI: 12.1±3.36 IG: 12.0±1.4 CG: 12.5±2.83 FPG & PPG: NR	NR	Diurnal plasma glucose profile was measured, Also measured c-peptide

T2DM: type 2 diabetes mellitus, HbA1c: glycated haemoglobin, TG: triglyceride, HDL-c: high density lipoprotein-cholesterol, LDL-c: low density lipoprotein cholesterol, T-chol: total cholesterol, VLDL-c: very low density lipoprotein cholesterol, FPG: fasting plasma glucose, PPG: post-prandial plasma glucose, SBP: systolic blood pressure, DBP: diastolic blood pressure, WBC: white blood cells, ALT: alanine aminotransferase, AST: aspartate aminotransferase, eGFR: estimated glomerular filtration rate, hsCRP: high sensitivity c-reactive protein, HOMA-IR: Homeostatic Model Assessment of Insulin Resistance, RCT: randomised control trial, NR: Not reported

Supplementary Table 2

Characteristics of excluded studies

Author, year, reference	Study design, Study duration, Study population	Study details
Sheikhbahaie et al., 2016, [25]	Double blind randomized controlled trial, 12 weeks, Prediabetes	39 patients with prediabetes randomly assigned to receive HCQ (6.5 mg/kg/day) (n=20) or placebo (n=19)
Pareek et al., 2015, [26]	Double blind randomized controlled trial, 24 weeks, Primary dyslipidemia	328 patients with primary dyslipidemia randomized to receive either atorvastatin 10 mg (n=167) or atorvastatin 10 mg + hydroxychloroquine 200 mg (n=161)
Gupta A, 2019, [27]	Prospective observational study, 48 weeks, T2DM, 18-65 years, HbA1c between ≥7% and <10.5%, on a combination of multiple OHA	HCQ 400 mg OD added to pre-existing therapy (n=250), No control group
Baidya et al., 2018, [28]	Open labelled randomized observational study, 24 weeks, T2DM, on stable doses of insulin with glimepride 2mg/d and metformin 1000mg/d	240 patients were randomly allocated to either HCQ 200 mg OD (n=120) or HCQ 400 mg OD (n=120)
Jagnani et al., 2017, [29]	Retrospective comparative observational study, 24 weeks, T2DM, adults, body weight ≥60 kg, on metformin 2g/d and glimepride 2mg/d	Compared HCQ 400 mg OD (n=100) with Teneligliptin 20 mg OD (n=100) as add on therapy
Singh et al., 2018, [30]	Retrospective observational study, 24 weeks, T2DM, HbA1c 7.5% to 9.5%, on teneligliptin 20 mg in addition to metformin and glimepiride with or without other antidiabetic therapy	Teneligliptin 20 mg OD was replaced by HCQ 400 mg OD (n=500), No control group
Baidya et al., 2018, [31]	Randomised observational study, 24 weeks, T2DM, HbA1c ≥7.0-≤9.0%, on metformin 1000mg/day and glimepiride 2mg/day for at least 12 weeks	165 patients randomized into 3 groups: Group A (n=55): uptitration of metformin to 2g/d with glimepride 2 mg/d; Group B (n=55): HCQ 400 mg OD added to metformin 1g/d and glimepride 2mg/d; Group C (n=55): uptitration of glimepride to 4 mg/d with metformin 1g/d
Chandra et al., 2019, [32]	Open label observational study, 24 weeks, T2DM, aged 40-70 years, BMI 22-35 kg/m2, HbA1c ≥8%, on gliclazide 80 mg/day, with metformin 1000 mg/day, along with insulin glargine (≥30 units/day) for over three months	HCQ 400 mg OD given in addition to insulin glargine, gliclazide and metformin (n=105)
Wasko et al., 2015, [33]	Double blind RCT, 13 weeks, Non-diabetic volunteers, age >18, overweight or obese, with one or more markers of insulin resistance	Participants were randomized to receive HCQ 400 mg OD (n=17) or placebo (n=15)
Powrie et al., 1991,[34]	Double blind RCT, 3 days, T2DM, controlled on diet only	Patients randomized into 2 groups (n=10 in each group) to receive chloroquine phosphate 250 mg QID or placebo; hyperinsulinemic euglycemic clamp studies combined with a primed continuous infusion of stable isotope to allow calculation of hepatic glucose production and glucose utilization were done

T2DM: type-2 diabetes; RCT: randomized controlled trial; HCQ: hydroxychloroquine; OD: once daily; OHA: oral hypoglycaemic agents; BMI: body mass index

Risk of bias in the included studies

The summaries of the risk of bias of the 11 studies included in the meta-analysis have been elaborated in Figures 2a, and b. The random sequence generation bias was judged to be low risk in 5 out of the 11 RCTs (45.45%). The allocation concealment and reporting bias were judged to be at low risk of bias in all the 11 RCTs (100%). The performance bias (blinding of participants and investigators) and detection bias (blinding of outcome assessors) were judged to be at low risk of bias in 4 out of 11 studies (36.36%). Attrition bias was low risk in 8 out of 11 studies (72.72%). The source of funding, especially pharmaceutical, authors from the pharmaceutical organizations, and conflict of interests were looked into the “other bias” section. The other bias was judged to be at a low risk in 10 out of 11 studies (90.90%) [Figure 2a, b]. Among the 11 trials evaluated in this study, four were of good quality, one of fair quality, and six trials (54.54%) were of poor quality as evaluated by the Jadad scale [Supplementary Table 3].

Figure 2

(a) Risk of bias graph: review authors’ judgments about each risk of bias item presented as percentages across all the included studies; (b) Risk of bias summary: Review authors’ judgments about each risk of bias item for each included study

Supplementary Table 3

Risk of bias assessment of RCTs in this meta-analysis using Jadad scale

Study	Randomization (0-2)	Blinding (0-2)	An account of all patients (0-1)	Total scoring (quality)*
Baidya (2018)	1	0	1	2
Chakravarti (2020)	2	2	1	5
Gerstein (2002)	2	2	1	5
Hsia (2020)	2	0	1	3
Kumar (2018)	1	0	1	2
Kumar (2019)	1	0	1	2
Pareek (2014)	2	2	0	4
Quatraro (1990)	2	2	1	5
Ranjan (2018)	1	0	1	2
Singh (2018)	1	0	1	2
Singh (2018a)	1	0	1	2

*The thresholds for assessing quality are as follows: (1) good (4-5 points); (2) fair (3 points); and (3) poor (0-2 points)

Effect of HCQ on the primary outcomes: Efficacy

HbA1c

Data from eight studies involving 2,334 patients were analyzed to find out the impact of HCQ on HbA1c in the ACG. The reduction in HbA1c after 16-24 weeks was significantly greater in the HCQ group as compared to the active controls (MD -0.17% [95%CI: -0.30 – -0.04; P =0.009; I2=89%][considerable heterogeneity]); Figure 3a; very low certainty of evidence (VLCE); Supplementary Table 1]. Data from three studies involving 284 patients with T2DM were analyzed to find out the impact of HCQ on HbA1c in the PCG. The reduction in HbA1c after 12-24 weeks was significantly greater in people receiving HCQ as compared to those receiving placebo in the PCG (MD -1.35% [95% CI: -2.10 – -0.59; P = 0.005; I2 = 74%] [moderate heterogeneity]); Figure 3b].

Figure 3

Forest plot evaluating the impact of hydroxychloroquine on (a) HbA1c in the active control group (ACG); (b) Hba1c in the passive control group (PCG); (c) Fasting glucose in the ACG; (d) Fasting glucose in PCG

Effect of HCQ on the secondary outcomes: Efficacy

Fasting glucose

Data from eight studies involving 2,334 patients were analyzed to find out the impact of HCQ on fasting glucose in the ACG. The reduction in the fasting glucose after 16-24 weeks was significantly greater in the HCQ as compared to the active controls (MD -16.63 mg/dL [0.9 mmol/L] [95% CI: -25.99 – -7.28 mg/dL] [-1.4 - -0.4 mmol/L]; P < 0.001; I2 = 97% [considerable heterogeneity]; Figure 3c; VLCE; Table 1). Data from only one study involving 112 patients with T2DM were analyzed to find out the impact of HCQ on fasting glucose in the PCG, which was significantly higher in the HCQ as compared to those receiving placebo in the PCG (MD -22.00 mg/dL [-1.2 mmol/L] [95% CI: -32.47 – 11.53 mg/dL] [-1.8 - -0.6 mmol/L]; P < 0.001; Figure 3d).

Post-prandial Glucose

Data from seven studies involving 2,312 patients were analyzed to find out the impact of HCQ on 2-hour PPG in the ACG. The reduction in the 2-h PPG was significantly greater in the people receiving HCQ as compared to the ACG (MD -8.41 mg/dL [-0.5 mmol/L] [95% CI: -14.71 – -2.12 mg/dL] [-0.8 - -0.1 mmol/L]; P = 0.009; I2 = 87% [considerable heterogeneity]; Figure 4a; VLCE; Table 1). Data from one study involving 112 patients (Chakravarti et al[14]) with T2DM was analyzed to find out the impact of HCQ on 2-h PPG in the PCG. The reduction in 2-h PPBG was significantly higher in the people receiving HCQ as compared to those receiving placebo in the PCG (MD -41.00 mg/dL [-2.3 mmol/L] [95% CI: -55.64 – 26.36 mg/dL] [-3.1 - -1.5 mmol/L]; P < 0.001).

Figure 4

Forest plot evaluating the impact of hydroxychloroquine as compared to active controls on (a) 2-h post-prandial blood glucose (PPBG); (b) Total cholesterol (TC); (c) LDL cholesterol; (d) HDL cholesterol; (e) Triglycerides; (f) Percentage of people achieving HbA1c <7%

Effect of HCQ on secondary outcomes: Lipid parameters

Data from four studies involving 1,154 patients with T2DM was analyzed to find out the impact of HCQ on TC and LDL cholesterol in the ACG. The reduction in the TC was significantly greater in the HCQ as compared to those receiving any other anti-diabetes medication in the ACG (MD -5.78 mg/dL [-0.15 mmol/L] [95% CI: -9.52 – -2.04 mg/dL] [-0.25 - -0.05 mmol/L]; P = 0.002; I2 = 35% [low heterogeneity]; Figure 4b; moderate certainty of evidence (MCE); Table 1). The reduction in LDL cholesterol was significantly greater in HCQ as compared to the ACG (MD - 4.38 mg/dL [-0.11 mmol/L] [95% CI: -6.41 – - 2.34 mg/dL] [-0.17 - - 0.06 mmol/L]; P < 0.001; I2 = 26% [low heterogeneity]; Figure 4c; MCE).

Data from three studies involving 854 patients were analyzed to find out the impact of HCQ on HDL cholesterol in the ACG which was not significantly different (MD -0.36 mg/dL [-0.01 mmol/L] [95% CI: -2.99 – 2.47] [-0.08 – 0.06 mmol/L]; P = 0.79; I2 = 60% [moderate heterogeneity]; Figure 4d). Data from three studies involving 854 patients were analyzed to find out the impact of HCQ on serum triglycerides in the ACG. Serum triglycerides was significantly lower in people receiving HCQ as compared to those receiving any other anti-diabetes medication in the ACG (MD -5.99 mg/dL [-0.07 mmol/L] [95% CI: -6.91 – - 5.06 mg/dL] [-0.08 - - 0.06 mmol/L]; P < 0.001; I2 = 0% [low heterogeneity]; Figure 4e; MCE).

The effect of HCQ on secondary outcomes: HbA1c <7% (53 mmol/mol)

Data from three studies involving 854 patients were analyzed to find out the impact of HCQ on the percentage of people achieving HbA1c <7% (53 mmol/mol) in the ACG. The percentage of people achieving HbA1c <7% (53 mmol/mol) was not different in the patients receiving HCQ as compared to the ACG (odds ratio [OR] 0.78 [95% CI: 0.06 – 10.58]; P = 0.85; I2 = 69% [moderate heterogeneity]; Figure 4f).

Effect of HCQ on secondary outcomes: Safety

Data from 11 studies involving 2,723 patients were analyzed to evaluate the impact of HCQ on the occurrence of adverse events. The adverse events reported were primarily minor with gastrointestinal adverse events being the most common (gastritis, flatulence, constipation). Two patients noticed hyperpigmentation. In a study by Pareek et al.,[16] two deaths were reported from the HCQ group (one due to acute myocardial infarction and the other due to acute pulmonary edema), and were believed to be unrelated to the drug as per the investigators.[14] Three reports of non-proliferative diabetic retinopathy (NPDR) were noted from the same study (two in the pioglitazone ACG and one in the HCQ group).[14] In the study by Gerstein et al.,[15] one patient in the HCQ group developed proliferative retinopathy (PDR) requiring laser therapy after 18 months of treatment.[13] Another patient in the HCQ group developed abnormal visual fields from drusen in both eyes and macular edema after 7 months of therapy. One patient in the placebo control group developed macular degeneration after 10 months of follow-up.[13] The occurrence of total adverse events (TAEs) was not statistically different in patients receiving HCQ as compared to the controls (RR 0.93 [95% CI: 0.68 – 1.28]; P = 0.65; I2 = 66% [considerable heterogeneity]; Figure 5a; low certainty of evidence [LCE]; Table 1).

Figure 5

Forest plot evaluating the impact of hydroxychloroquine on (a) Total adverse events; (b) Hypoglycemia; (c) Bodyweight

Data from 10 studies involving 1,974 patients were analyzed to evaluate the impact of HCQ on the occurrence of hypoglycemic events. The occurrence of hypoglycemia was not statistically different in patients receiving HCQ as compared to the controls (RR 0.78 [95% CI: 0.39 – 1.59]; P = 0.50; I2 = 66% [considerable heterogeneity]; Figure 5b; LCE; Table 1). A reduction in the insulin dose was noted in the patients receiving HCQ in the studies by Gernstein et al[15]. Quatraro et al[18] and Kumar et al.[20] Once, the patient in the HCQ group in the study by Gernstein et al[15]. had severe hypoglycemia, necessitating a significant reduction in the total daily dose of insulin.[13] There were no other reports of severe hypoglycemia.

Data from seven studies involving 2,046 patients were analyzed to evaluate the impact of HCQ on the bodyweight, which was not significantly different (MD -0.54 kg [95% CI: -1.11 – 0.03]; P = 0.06; I2 = 91% [considerable heterogeneity]; Figure 5c; LCE; Table 1).

Discussion

A meta-analysis has been published on December 31, 2020, evaluating the glycemic efficacy and safety of HCQ in T2DM.[36] That study included data from eight RCTs (1,763 patients).[36] The authors noted that HCQ use over 6 months was associated with HbA1c reduction of -0.88% (95% CI: -1.01 to - 0.75) compared to placebo and by -0.32% (95% CI: -0.37 to -0.26) compared to other oral diabetes agents.[36] Our meta-analysis is a more updated version having included data from three more RCTs which were missed in the prior analysis. We evaluated data from 11 RCTs (three having placebo as controls and eight having anti-diabetes medications as controls involving 2,723 patients). Ours is the first Cochrane meta-analysis to holistically analyze the efficacy and safety of HCQ in managing glycemia in people with T2DM with rigorous grading of the results. In our meta-analysis, HCQ use over 6 months was associated with HbA1c reduction of -1.35% (95% CI:-2.10 to -0.59) compared to placebo and by -0.17% (95% CI:-0.30 to -0.04) compared to other oral anti-diabetes agents. When compared to active controls, the additional benefit with HCQ with regards to HbA1c reduction was much more tempered in our analysis as compared to the previous observations.[36] The results of a meta-analysis are as good as the quality of the studies evaluated in it. These observations are clouded by the high data heterogeneity and the very low quality of evidence generated, which raises questions, if the same can be replicated in the real-world scenario. Hence, this meta-analysis highlights the poor quality of data evaluating the short-term glycemic efficacy of HCQ as the performance and detection bias was high in 60% of the studies, with >50% of the RCTs being of poor quality as per the Jadad scale. The issues of poor data quality and high heterogeneity were also noted previously.[36] A marginally better weight reduction seen with HCQ in this meta-analysis may be an indirect consequence of the use of pioglitazone in some of the studies as the active comparator agent, which is well known to cause weight gain.

This meta-analysis highlights that the current safety data with regard to the use of HCQ for managing diabetes is primarily limited to 6 months. There is a glaring lack of long-term safety data with regard to the use of HCQ in T2DM. Hence, it is surprising that the research society for the study of diabetes in India (RSSDI) has recommended the use of HCQ as a third-line agent in the management of T2DM.[37] Till definitive further long-term data are available, it may be said that the good clinical practices with the long-term use of HCQ include ensuring that the total daily dose should not cross 400 mg, preferable to keep a daily dose <5 mg/kg/day, keep the total lifetime cumulative dose to less than 1,000 mg, restrict the total duration of therapy to less than 6 months in accordance with the current available safety and efficacy data in the use of HCQ in T2DM, and to reduce the risks of retinal toxicity.[838] It must be remembered that diabetes per se is a risk factor for retinal injury. A recent study from India involving more than 51,000 patients reported a 19% (95% CI: 18.9-19.5) prevalence of diabetic retinopathy.[39]

The potential issues regarding long-term cardiovascular safety and ocular safety can only be established with long-term large multi-centric RCTs.

To conclude, it may be said that the current evidence with regards to the use of HCQ in managing T2DM remains sketchy in view of the predominantly short-term studies of poor quality and high heterogeneity. The routine use of HCQ for managing T2DM cannot be recommended based on the current available data.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

Funnel plot for (a) HbA1c in the ACG; (b) Fasting glucose in the ACG; (c) Post-prandial glucose in the ACG; (d) Total cholesterol in the ACG; (e): Total adverse events; (f): hypoglycaemia; (g) body weight