Safety and efficacy of sodium-glucose cotransporter-2 inhibitors in adults with type 2 diabetes fasting during Ramadan: A meta-analysis

Abstract

BACKGROUND

Sodium-glucose cotransporter-2 inhibitors (SGLT2i) are widely used in managing type 2 diabetes (T2D). A significant number of these patients choose to observe religious fasting during Ramadan. Although existing guidelines recommend caution when administering SGLT2i during Ramadan due to potential adverse effects, there is limited data on their use in this patient population.

AIM

To assess the safety and effectiveness of SGLT2i in patients with T2D who fast during Ramadan.

METHODS

Relevant studies involving adults with T2D who received an SGLT2i in the intervention arm and other glucose-lowering drugs in the control arm were systematically searched through electronic databases. The primary outcome was the occurrence of adverse events in the two groups; additional outcomes included changes in glycemic and anthropometric parameters during the peri-Ramadan period. RevMan Web was used to conduct meta-analysis using random-effects models. Outcomes were presented as mean differences (MDs) or risk ratios (RRs) with 95%CI.

RESULTS

Twelve studies involving 3625 subjects were included. The risks of postural dizziness (RR = 6.39, 95%CI: 1.58-25.80, P = 0.009, I² = 44%), hypotension/postural hypotension (RR = 4.43, 95%CI: 1.35-14.55, P = 0.01, I² = 31%), and sodium loss (MD = -1.00 mmol/L, 95%CI: -1.34 to -0.67, P < 0.00001, I² = 0%) were higher in the SGLT2i group compared to the non-SGLT2i group. The SGLT-2i group achieved larger reductions in systolic (MD = -2.41 mmHg, 95%CI: -4.52 to -0.30, P = 0.02, I² = 46%) and diastolic blood pressure (MD = -1.71 mmHg, 95%CI: -2.70 to -0.72, P = 0.0007, I² = 20%), and experienced a lower risk of symptomatic hypoglycemia (RR = 0.53, 95%CI: 0.29-0.97, P = 0.04, I² = 69%). The two groups exhibited comparable changes in glycated hemoglobin, body weight, and renal function. The risks of other specific adverse events, including dehydration, dizziness, volume depletion, symptomatic hyperglycemia, severe hypoglycemia, and genitourinary infections, were identical in the two groups.

CONCLUSION

SGLT2i may be generally safe and effectively manage T2D during Ramadan; however, the results are less robust and should be interpreted with caution. Large multicenter randomized trials are necessary to confirm their safety, especially for at-risk groups, and to improve clinical decision-making.

Key Words: Type 2 diabetes; Sodium-glucose cotransporter-2 inhibitors; Ramadan; Religious fasting; Safety; Meta-analysis

Core Tip: This meta-analysis assessed the safety and effectiveness of sodium-glucose cotransporter-2 inhibitors (SGLT2i) in adults with type 2 diabetes fasting during Ramadan. Among 12 studies with 3625 participants, SGLT2i increased risks of postural dizziness, hypotension, and sodium loss, but achieved greater reductions in blood pressure and a lower risk of symptomatic hypoglycemia than other glucose-lowering drugs. SGLT2i did not impact glycemic control, body weight, or renal function. Other adverse events, like dehydration and severe hypoglycemia, showed no significant difference. SGLT2i appear safe and effective for diabetes management during Ramadan fasting, but larger studies are needed to confirm safety for vulnerable groups.

INTRODUCTION

Sodium-glucose cotransporter-2 inhibitors (SGLT2i) are a class of glucose-lowering drugs (GLDs) that increase urinary glucose excretion in individuals with type 2 diabetes (T2D), thereby reducing plasma glucose concentrations. SGLT2i are key in diabetes management, lowering blood glucose and offering cardiovascular, hepatic, and renal protection. They are often first-line or second-line treatments, sometimes replacing metformin, due to benefits such as improved glycemic control, weight loss, reduced risk of hypoglycemia, and reduced hospitalization for heart failure and kidney disease. Their role extends beyond glucose-lowering to organ protection, enhancing long-term health outcomes[1]. However, these drugs may cause several adverse effects, including genitourinary mycotic infections, dehydration due to osmotic diuresis, and diabetic ketoacidosis in some individuals[1,2]. Fasting during Ramadan, the ninth month of the Islamic lunar calendar, is one of the five pillars of the Islamic faith. During Ramadan, all adult Muslims are required to abstain from consuming food and drink (including medication) from sunrise to sunset[3]. It is estimated that approximately 148 million Muslims worldwide live with diabetes. Fasting duration varies by geographic location and can reach 19 hours a day in some parts of the world, particularly during summer[3].

The lifestyle changes during Ramadan, particularly abstaining from food and drink for extended non-fasting hours and utilizing GLDs, may predispose these individuals to adverse events (AE), such as dehydration and hypoglycemia[4]. Insulin and sulfonylureas pose the highest risk of hypoglycemia, while dipeptidyl peptidase-4 inhibitors generally have a lower risk but still need monitoring. Glucagon-like peptide-1 (GLP-1) agonists are considered safe with modest glycemic benefits but may cause gastrointestinal side effects like nausea and vomiting, which can be worsened by fasting. Pioglitazone has risks of fluid retention and weight gain, potentially complicating fasting. Careful dose adjustment and monitoring are essential to prevent hypoglycemia and other complications during Ramadan fasting[5,6]. SGLT2i may worsen dehydration in these individuals, particularly in those fasting in hot, humid environments. Additionally, there are concerns about an increased risk of diabetic ketoacidosis when using SGLT2i during Ramadan[4]. Few small studies have assessed the safety and efficacy of SGLT2i during Ramadan, focusing on the risks of dehydration, hypoglycemia, ketonemia, and renal function. The American Diabetes Association/European Association for the Study of Diabetes consensus recommends the use of SGLT2i with caution during Ramadan for fear of adverse drug effects[5]. The Diabetes and Ramadan International Alliance, in collaboration with the International Diabetes Federation, recommends initiating SGLT2i at least 2-4 weeks before Ramadan if these medications are to be used. The guidelines recommend promoting adequate hydration for users by encouraging them to consume plenty of water and other fluids during non-fasting hours and by evaluating renal function, as there may be an initial drop in glomerular filtration rate[6].

There are only a few observational studies with small sample sizes, and no well-designed randomized controlled trials (RCTs) have been conducted to analyze the adverse effects and efficacy of SGLT2i in patients with T2D who fast during Ramadan. In 2022, Gad et al[7] published a systematic review and meta-analysis on the impact of SGLT2i use during Ramadan fasting in patients with T2D, but they only included five studies, whereas many new ones have been published since then, and Abdelgadir et al[8] recently published a systematic review on the same topic without conducting meta-analyses. With this background, we conducted this meta-analysis to explore whether individuals with T2D using SGLT2i during the fasting month of Ramadan gain glycemic and other metabolic benefits, with or without an increased risk of AE.

MATERIALS AND METHODS

Ethical compliance

This meta-analysis was conducted following the procedures described in the Cochrane Handbook for Systematic Reviews of Interventions and is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist[9,10]. This study has been duly registered with PROSPERO (No. CRD42024627056), and the protocol summary is accessible online. Since ethical approval was obtained for the individual studies in the meta-analysis, no separate approval was required.

Search strategy

A comprehensive search was executed across multiple databases and registers, including PubMed, Scopus, Web of Science, the Cochrane Central Register of Controlled Trials, and ClinicalTrials.gov. This search encompassed the entirety of each database from its inception until December 5, 2024. Using the Boolean operators “AND” and “OR”, the following terms were searched: “Sodium-glucose cotransporter-2 inhibitors”, “SGLT-2 inhibitors”, “SGLT2 inhibitors”, “Empagliflozin”, “Dapagliflozin”, “Canagliflozin”, “Ertugliflozin”, “Bexagliflozin”, “Sotagliflozin”, “Ramadan fasting”, “Religious fasting”, “Type 2 diabetes mellitus”, “Type 2 diabetes”, “Type II diabetes”, “T2DM”, and “T2D”. The search terms were applied to all fields. The goal was to find both recently published and unpublished studies in English. Furthermore, the search included reviewing references in retrieved articles and related journals. The complete search strategies are included in the Supplementary material.

Study selection

The population, intervention, comparison, outcomes, and study framework was used to establish eligibility criteria for studies in this meta-analysis. The patient population consisted of adults of either sex with T2D who fasted during Ramadan; the intervention was SGLT2i, either as monotherapy or combined with other GLDs, for managing T2D; the comparison/control group comprised individuals receiving GLDs other than an SGLT2i, either as monotherapy or in combination. The outcomes encompassed safety measures, such as volume depletion and hypoglycemia, as well as the effectiveness of SGLT2i during Ramadan, including its effect on glycemic control. Prospective observational/intervention studies, or RCTs, were considered the study type for inclusion. The studies must have at least two groups: One with an SGLT2i alone or in combination with other GLDs (SGLT2i group), and the other with GLDs other than SGLT2i, either as monotherapy or in combination therapy (non-SGLT2i group). Exclusion criteria included studies conducted in patients with type 1 or other types of diabetes, those involving patients with both type 1 diabetes and T2D, studies involving pregnant or lactating women, studies lacking a non-SGLT2i control arm, and studies that did not report the outcomes of interest. The study selection involved four independent authors who, after removing duplicates, reviewed titles and abstracts according to specific inclusion and exclusion criteria. This process filtered out irrelevant studies, leaving potentially eligible ones for further analysis. Following consensus among the authors, the full texts of the remaining articles were retrieved and examined in detail to confirm eligibility. For each excluded article, reasons were documented, as illustrated in the PRISMA flow diagram. Any disagreements were settled through discussion and agreement.

Data extraction

Four review authors independently extracted data using standardized forms. Upon identifying several publications from the same study group, the results were consolidated, and relevant data from each article were included in the analyses. The following data were extracted from all eligible studies and included in the review: (1) First author; (2) Year of publication; (3) Country where the study was conducted; (4) Study design; (5) Major inclusion criteria of the study subjects; (6) GLDs used in the SGLT2i and non-SGLT2i groups; (7) Sample size; (8) Number of male and female participants; (9) Mean age; (10) Duration of diabetes; (11) Study duration; (12) Body weight; (13) Body mass index (BMI); (14) Waist circumference (WC); (15) Glycated hemoglobin (HbA1c); (16) Fasting plasma glucose (FPG); (17) Systolic blood pressure (BP); (18) Diastolic BP; (19) Serum urea; (20) Serum creatinine; (21) Estimated glomerular filtration rate (eGFR); (22) Serum electrolytes; and (23) AE, including features of volume depletion and hypoglycemia. All disagreements were resolved through consensus.

Dealing with missing data

The Supplementary material for the articles was obtained from the publishers’ websites. If needed, additional information was gathered from the corresponding authors of the relevant articles via email. All pertinent information collected this way was carefully incorporated into the meta-analysis. Additionally, attrition rates – such as dropouts, losses to follow-up, and withdrawals – were thoroughly analyzed.

Outcomes analyzed

The primary outcomes of interest were comparisons of the proportions of study subjects experiencing AE during Ramadan and mean differences (MDs) in changes from baseline to the end of the study in serum urea, serum creatinine, eGFR, and serum electrolytes between the SGLT2i and non-SGLT2i groups. Additional outcomes included changes in HbA1c, FPG, body weight, BMI, WC, and BP from baseline to study end between the two groups, as measured by MD.

Statistical analysis

The outcomes were reported using MDs for continuous variables and risk ratios (RRs) for dichotomous variables, along with 95%CI. The Review Manager (RevMan) computer program, version 7.2.0, was used to generate forest plots showing the RR or MD for the observed outcomes; the left side of the forest plot favored the SGLT2i group, while the right favored the non-SGLT2i group[11]. Random-effects analysis models were selected to address the expected heterogeneity arising from variations in population characteristics and study duration. The inverse-variance statistical method was used in all instances. The meta-analysis included forest plots combining data from at least two trials. A significance level of P < 0.05 was applied.

Assessment of the risk of bias

Two authors independently assessed the risk of bias (RoB) in the included studies using the Risk of Bias in Non-Randomized Studies of Interventions, Version 2 assessment tool[12]. The tool addresses seven distinct domains through which bias may be introduced. Based on responses to the signaling questions designed to elicit information relevant to the RoB judgment for each domain, the options for a domain-level RoB judgment include low, moderate, serious, or critical risk of bias, with an additional choice of no information. The evaluations in each area contribute to the overall RoB judgment for the assessed outcome. In cases of discrepancies, the sixth and seventh authors served as arbitrators to reach a consensus. The Risk-of-bias VISualization web application was used to create risk-of-bias plots[13]. When applicable (a minimum of ten studies in a forest plot), publication bias was assessed using funnel plots generated by RevMan web software[11,14].

Assessment of heterogeneity

Heterogeneity assessment began with an analysis of forest plots. Following this, a χ² test was conducted using N-1 degrees of freedom and a significance level of 0.05 to evaluate statistical significance. Additionally, the I² test was used in the subsequent analysis. The thresholds for I² values were set at 25% for low heterogeneity, 50% for moderate heterogeneity, and 75% for high heterogeneity[15].

RESULTS

Search results

Figure 1 illustrates the steps in selecting the studies using the PRISMA flow diagram. The initial search identified 252 articles, which were narrowed to 15 after screening titles and abstracts and subsequent full-text reviews. Ultimately, 12 studies involving a total of 3625 subjects who met all the predetermined criteria were included in this meta-analysis[16-27]. Three studies were excluded because they lacked a non-SGLT2i comparator group[28-30]. The key characteristics and findings of the excluded studies have been summarized in Supplementary Table 1[28-30].

Figure 1 Flowchart on study retrieval and inclusion in the meta-analysis. SGLT2i: Sodium-glucose cotransporter-2 inhibitors.

Characteristics of the studies included

Table 1 presents a summary of the studies included[16-27]. All included studies, except for Wan Seman et al[26], which was a randomized, open-label study, were prospective observational. All but one study had two arms: One with any SGLT2i alone or in combination with other GLDs (SGLT2i group) and the other with GLDs other than SGLT2i, either as monotherapy or in combination therapy (non-SGLT2i group). Five studies used empagliflozin[17,19,22,24,27], one used canagliflozin[20], one used ertugliflozin[23], and one used dapagliflozin[26] in the SGLT2i arm. Three studies did not specify which SGLT2i was used[16,21,25]. The study by Gameil et al[18] had two SGLT2i arms: One with dapagliflozin and the other with empagliflozin. These two arms were separately analyzed and identified as Gameil et al[18] (Dapa) and Gameil et al[18] (Empa), respectively. The mean age of the study participants ranged from 44.7 years to 58.4 years. Their baseline HbA1c ranged from 7.2% to 9.3%. The follow-up duration for the studies ranged from 6 weeks to 5 months.

Table 1 Baseline characteristics of individual studies and study participants included in the meta-analysis, mean (SD)/median (interquartile range).

Ref.	Study design	Major inclusion criteria	Groups	Glucose-lowering drug used	Number	Sex	Age (years)	Duration of diabetes mellitus (years)	Baseline body weight (kg)	Baseline HbA1c (%)	Study duration
Abdelgadir et al[16], 2019, United Arab Emirates	Prospective interventional	Age 18-75 years, all using insulin	SGLT2i	SGLT2i + insulin ± other OADs	49	Male: 19, female: 30	57.5 (9.1)	NA	82 (13.9)	7.95 (1.1)	14 weeks
			Non-SGLT2i	Insulin+ OADs other than SGLT2i	46	Male: 16, female: 36	55.4 (9.4)	NA	85 (11.1)	7.7 (1.0)
Ahmed et al[17], 2022, Pakistan	Prospective, observational	HbA1c 6.5%-8.5%, eGFR > 60 mL/minute/1.73 m2	SGLT2i	Metformin + empagliflozin	44	Male: 24, female: 20	44.7 (10.7)	NA	85.3 (7.8)	7.2 (0.8)	9 weeks
			Non-SGLT2i	Metformin + sitagliptin	88	Male: 43, female: 45	50.6 (10.5)	NA	77.6 (13.7)	7.8 (1.5)
Gameil et al[18], 2022, Egypt	Case-control observational	Age 45-60 years, HBA1c 7.5%-8.8%, body mass index > 25 kg/m2, eGFR > 60 mL/minute/1.73 m2	SGLT2i	Empagliflozin ± OAD/glargine	87	Male: 58, female: 29	52.4 (6)	5 (2-9)	89.6 (14.6)	7.9 (0.5)	NA
			SGLT2i	Dapagliflozin ± OAD/glargine	85	Male: 54, female: 31	52.5 (6.3)	6 (2-8)	92.5 (13.2)	8.1 (0.4)
			Non-SGLT2i	OAD other than SGLT2i ± basal insulin	73	Male: 48, female: 25	52.3 (6)	5 (2-9)	93.4 (12.1)	8.3 (0.5)
Goh et al[19], 2023, Malaysia	Prospective cohort	Age 18-75 years, eGFR > 30 mL/minute/1.73 m2	SGLT2i	Empagliflozin ± OAD/insulin	48	Male: 12, female: 36	48.0 (43.5-56.5)	10.5 (6-15)	77.7 (66.7-99.2)	8.4 (7.25-9.85)	6 weeks
			Non-SGLT2i	OAD except SGLT2i ± insulin	50	Male: 16, female: 34	51.5 (42-58)	4.5 (2-10)	75.3 (69.4-84.1)	8.1 (6.60-10.30)
Hassanein et al[20], 2017, Lebanon, Kuwait, and the United Arab Emirates	Non-randomised, parallel-cohort, prospective, comparative	Age 18-65 years, HbA1c ≤ 8.5%	SGLT2i	Canagliflozin + metformin ± DPP4i	162	Male: 100, female: 62	52.3 (7.7)	6.5 (5.9)	87.1 (14.8)	7.3 (0.8)	8 weeks
			Non-SGLT2i	Sulfonylurea + metformin ± DPP4i	159	Male: 87, female: 72	54.3 (7.4)	7.6 (5.5)	82.1 (14.1)	7.2 (0.8)
Hassanein et al[21], 2024, nine countries	Prospective, real-world, observational	Age ≥ 18 years	SGLT2i	iGlarLixi + SGLT2i use, with or without other OADs	174	Male: 98, female: 76	55.2 (9.6)	11.4 (6.3)	87.9 (16.2)	8.2 (1.3)	3-5 months
			Non-SGLT2i	iGlarLixi with or without other OADs	246	Male: 134, female: 112	58.4 (9.8)	12.6 (6.4)	85.5 (14.0)	8.2 (1.0)
Pathan et al[22], 2022, Bangladesh	Multi-centre, open-label, two-arm parallel-group study	Age >18 years, HbA1c 7%-10%, eGFR > 45 mL/minute/1.73 m2	SGLT2i	Empagliflozin + metformin± DPP4i	274	Male: 134, female: 140	49.0 (11.8)	7.8 (5.9)	70.7 (11.3)	8.4 (1.45)	24 weeks
			Non-SGLT2i	Metformin ± DPP4i	219	Male: 95, female: 124	49.7 (10.5)	6.3 (4.8)	66.4 (9.5)	8.1 (1.46)
Pathan et al[23], 2025, Bangladesh	Real-life experience study	Age ≥ 18 years, type 2 diabetes mellitus for at least 1 year, HbA1c < 10%	SGLT2i	Ertugliflozin ± OAD	703	Male: 276, female: 419	50.21 (11.11)	6.74 (5.39)	69.93 (11.54)	8.16 (0.91)	8 weeks
			Non-SGLT2i	OAD except ertugliflozin	670	Male: 281, female: 387	50.53 (11.17)	5.75 (4.46)	68.74 (10.53)	7.79 (0.87)
Samkari et al[24], 2023, Saudi Arabia	Prospective cohort study	Age 18-70 years, eGFR > 60 mL/minute/1.73 m2	SGLT2i	Empagliflozin + OAD	77	Male: 33, female: 44	56 (51-62)	9.00 (5.00-15.00)	NA	7.80 (7.10-8.77)	12 weeks
			Non-SGLT2i	Sulfonylurea and metformin ± DPP4i	77	Male: 35, female: 42	52 (47-64)	7.00 (5.00-15.00)	NA	7.50 (6.70-8.33)
Shao et al[25], 2018, Singapore	Prospective observational controlled cohort	Age 21-75 years, eGFR > 45 mL/minute/1.73 m2	SGLT2i	SGLT2i ± OAD/insulin	35	Male: 19, female: 16	49.9 (10.6)	15.7 (8.3)	88.2 (24.1)	9.3 (1.9)	NA
			Non-SGLT2i	OAD/insulin except SGLT2i	33	Male: 18, female: 15	54.7 (11.2)	12.3 (6.0)	73.9 (19.1)	8.7 (1.6)
Wan Seman et al[26], 2016, Malaysia	Randomized, open-label	Age 18-65 years, HbA1c 7%-10.5%	SGLT2i	Dapagliflozin + metformin	58	Male: 35, female: 23	53 (9.1)	5.0 (3.0- 9.0)	77.5 (13.9)	7.7 (7.08- 8.43)	12 weeks
			Non-SGLT2i	Sulfonylurea + metformin	52	Male: 31, female: 21	56 (9.1)	6.0 (3.0- 10.3)	75.6 (15.35)	7.6 (6.9-8.1)
Yousuf et al[27], 2022, Pakistan	Prospective case-control observational	Age 18-65 years, HbA1c < 8.5%	SGLT2i	Empagliflozin + metformin ± DPP4i	58	Male: 30, female: 28	52.1 (10.63)	6.98 (5.08)	NA	7.74 (0.89)	14 weeks
			Non-SGLT2i	Sulfonylurea + metformin ± DPP4i	58	Male: 31, female: 27	50.43 (9.23)	5.84 (5.82)	NA	7.9 (0.83)

DPP4i: Dipeptidyl peptidase-4 inhibitor; eGFR: Estimated glomerular filtration rate; HbA1c: Glycated hemoglobin; NA: Not available; OAD: Oral antidiabetic drug; SGLT2i: Sodium-glucose cotransporter-2 inhibitors.

Risk of bias of the included studies

Supplementary Figure 1 illustrates the RoB for the 12 studies included in the meta-analysis[16-27]. In most (75%) studies, the overall RoB was moderate, while one study (Gameil et al[18]) had a serious RoB and two studies (Abdelgadir et al[16], Hassanein et al[20]) were rated as critical. In all studies, confounding bias was present. Three studies had bias due to missing data[18,22,23]. The funnel plot for the studies in the meta-analysis comparing the proportions of subjects with symptomatic hypoglycemia in the SGLT2i group vs the non-SGLT2i group shows an asymmetrical distribution around the central line; therefore, significant publication bias is likely present (Supplementary Figure 2). Publication bias for other outcomes was not assessed because the forest plots included fewer than 10 studies.

Safety of SGLT2i

An identical proportion of study subjects broke fast during Ramadan in the SGLT2i and non-SGLT2i groups [RR = 0.38, 95%CI: 0.03-5.26, I² = 70% (moderate heterogeneity), P = 0.47]. The total number of days fasted was also comparable [MD = -0.26 days, 95%CI: -1.72 to 1.21, I² = 88% (high heterogeneity), P = 0.73]. Identical proportions of study subjects in the two groups experienced any AE, serious AE, hospitalization during the study period, abdominal pain, dehydration, dizziness, volume depletion, headache, hunger, sweating, generalized weakness, symptomatic hyperglycemia, increased urination, severe hypoglycemia, genital infections, and urinary tract infection. The risk of symptomatic hypoglycemia was lower in the SGLT2i group [RR = 0.53, 95%CI: 0.29-0.97, I² = 69% (moderate heterogeneity), P = 0.04]. The risks of postural dizziness [RR = 6.39, 95%CI: 1.58-25.80, I² = 44% (low heterogeneity), P = 0.009] and hypotension/postural hypotension [RR = 4.43, 95%CI: 1.35-14.55, I² = 31% (low heterogeneity), P = 0.01] were higher in the SGLT2i group. The two groups experienced comparable changes in serum urea, creatinine, potassium, and eGFR. A greater reduction in serum sodium was seen in the SGLT2i group [MD = -1.00 mmol/L, 95%CI: -1.34 to -0.67, I² = 0% (not important heterogeneity), P < 0.00001] (Table 2).

Table 2 Comparison of the safety outcomes in the sodium-glucose cotransporter-2 inhibitor vs non- sodium-glucose cotransporter-2 inhibitor arms.

Outcome variables (categorical)	Number of included studies	Number of participants with outcome/participants analyzed (%)		Pooled effect size	I2 (%)	P value
		SGLT2i arm	Non-SGLT2i arm
Any AE, RR (95%CI)	3	71/397	53/462	1.40 (0.83-2.39)	59	0.21
Serious AE, RR (95%CI)	2	1/220	0/211	2.74 (0.11-68.72)	Not available	0.54
Hospitalization, RR (95%CI)	2	2/135	0/129	2.84 (0.30-26.98)	0	0.36
Abdominal pain, RR (95%CI)	3	9/805	9/816	1.55 (0.62-3.88)	0	0.35
Dehydration, RR (95%CI)	5	64/1042	30/1050	1.79 (0.83-3.83)	58	0.14
Dizziness, RR (95%CI)	4	21/971	32/937	0.66 (0.10-4.21)	81	0.66
Postural dizziness, RR (95%CI)	5	65/427	4/396	6.39 (1.58-25.80)	44	0.009
Hypotension/postural hypotension, RR (95%CI)	7	61/1265	4/1195	4.43 (1.35-14.55)	31	0.01
Volume depletion, RR (95%CI)	4	40/571	33/513	1.34 (0.62-2.90)	54	0.46
Headache, RR (95%CI)	4	10/445	14/518	0.90 (0.38-2.14)	5	0.81
Hunger, RR (95%CI)	3	10/268	20/267	0.33 (0.04-2.78)	74	0.31
Sweating, RR (95%CI)	2	8/210	6/209	1.38 (0.51-3.78)	0	0.53
Generalized weakness, RR (95%CI)	4	22/326	27/319	0.58 (0.14-2.45)	54	0.46
Symptomatic hyperglycemia, RR (95%CI)	4	29/356	39/474	0.81 (0.54-1.20)	0	0.29
Increased urination, RR (95%CI)	4	23/322	14/357	2.15 (0.81-5.70)	16	0.12
Symptomatic hypoglycemia, RR (95%CI)	10	56/1640	135/1646	0.53 (0.29-0.97)	69	0.04
Severe hypoglycemia, RR (95%CI)	3	0/375	8/435	0.14 (0.02-1.11)	0	0.06
Genital infection, RR (95%CI)	4	7/326	3/319	2.03 (0.23-17.59)	47	0.52
Urinary tract infection, RR (95%CI)	5	15/1241	15/1188	1.18 (0.53-2.60)	0	0.69
Broke fast, RR (95%CI)	2	8/126	14/123	0.38 (0.03-5.26)	70	0.47
Number of days fasted, MD (95%CI)	3	252	296	-0.26 (-1.72 to 1.21)	88	0.73
Serum urea (mmol/L), MD (95%CI)	2	83	83	0.67 (-0.52 to 1.87)	86	0.27
Serum creatinine (mg/dL), MD (95%CI)	5	485	423	0.03 (-0.05 to 0.11)	64	0.43
Estimated glomerular filtration rate (mL/minute/1.73 m2), MD (95%CI)	7	1311	1189	-0.81 (-1.83 to 0.20)	0	0.12
Serum sodium (mmol/L), MD (95%CI)	3	1012	922	-1.00 (-1.34 to -0.67)	0	< 0.00001
Serum potassium (mmol/L), MD (95%CI)	3	1012	922	0.01 (-0.17 to 0.19)	51	0.89

AE: Adverse events; MD: Mean difference; RR: Risk ratio; SGLT2i: Sodium-glucose cotransporter-2 inhibitors.

Effect on glycemic control

After the studies’ end, the SGLT2i and non-SGLT2i groups had comparable changes in HbA1c [MD = -0.15%, 95%CI: -0.36 to 0.05, I² = 82% (high heterogeneity), P = 0.15] (Figure 2A)[16,17,20-23,26,27], and FPG [MD = 0.06 mmol/L, 95%CI: -0.73 to 0.86, I² = 91% (high heterogeneity), P = 0.87] (Figure 2B) from the baseline[21,23,25-27].

Figure 2 Forest plot demonstrating the mean differences in glycated hemoglobin and fasting plasma glucose in the sodium-glucose cotransporter-2 inhibitors group vs the non-sodium-glucose cotransporter-2 inhibitors group. A: Glycated hemoglobin; B: Fasting plasma glucose. SGLT2i: Sodium-glucose cotransporter-2 inhibitors.

Effects on anthropometric parameters

Study subjects in the SGLT2i and non-SGLT2i groups achieved comparable body weight [MD = -0.69 kg, 95%CI: -1.57 to 0.18, I² = 84% (high heterogeneity), P = 0.12] (Figure 3A)[16-23,25], BMI [MD = -0.15 kg/m², 95%CI: -0.43 to 0.14, I² = 51% (moderate heterogeneity), P = 0.31] (Figure 3B)[17-19,22,23,25,27], and WC [MD = -0.03 cm, 95%CI: -2.69 to 2.63, I² = 41% (mild heterogeneity), P = 0.98] (Figure 3C)[18,19] changes from baseline to the end of the trials.

Figure 3 Forest plot demonstrating the mean differences in body weight, body mass index, and waist circumference in the sodium-glucose cotransporter-2 inhibitors group vs the non-sodium-glucose cotransporter-2 inhibitors group. A: Body weight; B: Body mass index; C: Waist circumference. SGLT2i: Sodium-glucose cotransporter-2 inhibitors.

Effect on BP

Study subjects in the SGLT2i group achieved greater reductions in systolic BP [MD = -2.41 mmHg, 95%CI: -4.52 to -0.30, I² = 46% (low heterogeneity), P = 0.02] (Figure 4A)[16,17,19,20,22,23,25,27] and diastolic BP [MD = -1.71 mmHg, 95%CI: -2.70 to -0.72, I² = 20% (not important heterogeneity), P = 0.0007] (Figure 4B)[16,17,19,20,22,23,25,27] than the non-SGLT2i group.

Figure 4 Forest plot demonstrating the mean differences in systolic blood pressure and diastolic blood pressure in the sodium-glucose cotransporter-2 inhibitors group vs the non-sodium-glucose cotransporter-2 inhibitors group. A: Systolic blood pressure; B: Diastolic blood pressure. SGLT2i: Sodium-glucose cotransporter-2 inhibitors.

Sensitivity analysis

Leave-one-out sensitivity analyses were conducted for primary and secondary outcomes to detect changes in statistical significance and significant heterogeneity (at least a 2-step change) (Supplementary Table 2)[16-27]. For dehydration, heterogeneity among studies decreased after removing Hassanein et al[20], and the RR became significant after removing Pathan et al[23]. For dizziness, the levels of statistical significance or heterogeneity among studies did not change significantly after removing any included studies. For volume depletion, heterogeneity among studies decreased after removing either the study by Hassanein et al[20] or that by Samkari et al[24], and the RR became significant after removing Samkari et al[24]. The heterogeneity among studies did not change significantly after excluding any studies included for HbA1c. However, the SGLT2i group achieved a significantly greater reduction in HbA1c than the non-SGLT2i group in the study, Wan Seman et al[26], which was not considered. The heterogeneity among the studies decreased, and the difference in body weight reduction became statistically significant, favoring SGLT2i when the Hassanein et al[21] study was excluded. In summary, the parameters that became statistically significant after removing certain studies were dehydration (after removing Pathan et al[23]), volume depletion (after removing Samkari et al[24]), HbA1c reduction (when excluding Wan Seman et al[26]), and body weight reduction (after excluding Hassanein et al[21]). Sensitivity analysis reveals fragility: Excluding individual studies markedly alters key meta-analysis metrics, such as heterogeneity and the significance of the pooled effect. Such findings indicate that the pooled estimates are not very robust, as they depend heavily on certain studies, possibly due to outliers, biases, or small sample sizes, rather than consistent evidence. Therefore, they should be interpreted with caution.

DISCUSSION

This meta-analysis included 12 studies with a control population (non-SGLT2i) that fulfilled the inclusion criteria. The RoB was moderate in most included studies (75%), while the remainder were serious or critical, and there was a significant likelihood of publication bias. Only a few participants experienced AEs, including genitourinary infections, which required them to break their fast in both groups, demonstrating the safety of continued SGLT2i use during Ramadan. The risk of symptomatic hypoglycemia was low with the use of SGLT2i. While the risks of symptomatic hyperglycemia, increased urination, and dehydration were not heightened, the risks of hypotension/postural hypotension (RR = 4.43) and postural dizziness (RR = 6.39) were significantly greater in the SGLT2i group. The SGLT2i group achieved greater reductions in systolic and diastolic BP than the non-SGLT2i group. Laboratory parameters showed a statistically significant decrease in serum sodium in the SGLT2i group. No differences were observed in the renal parameters studied, including eGFR. Additionally, there were no differences in the changes in anthropometric parameters between the two groups. The HbA1c levels in both groups were comparable, suggesting the ongoing efficacy of SGLT2i during Ramadan without an increased risk of hypoglycemia. However, the observational nature of most included studies, which is expected to yield very low to low-grade evidence, the high heterogeneity observed across most outcome estimates, and the fragility of the evidence revealed in the sensitivity analysis all suggest that the results are less robust and should be interpreted with caution.

SGLT2i are recommended as the preferred medication for managing patients with T2D who have an established or high risk of atherosclerotic cardiovascular disease, heart failure (both preserved and reduced ejection fraction), and chronic kidney disease. Their global use is on the rise[31,32]. The American Diabetes Association and the International Diabetes Federation-Diabetes and Ramadan do not consider SGLT2i use during Ramadan a risk factor for fasting complications and recommend continuing their use during Ramadan[6,32]. The recommendations reiterate continued use given the lower risk of hypoglycemia and cardiorenal protection/safety. Still, caution should be taken regarding the risk of volume depletion and the resultant postural dizziness, especially in warm climates during prolonged fasting hours[6,32]. The use of SGLT2i is known to increase the risk of hypovolemia, dehydration, and postural hypotension, which can potentially worsen during prolonged periods of fasting. This is particularly important for frail individuals, the elderly, those on medications that can worsen volume depletion (e.g., diuretics), and patients receiving vasodilatory agents (e.g., antihypertensives)[33,34]. These individuals should be warned about the risk of falls due to postural instability, particularly in the later hours of fasting.

The increased risk of hypoglycemia in the control group could be attributed to the use of multiple GLDs, similar to the data obtained from the PROFAST Ramadan study, which demonstrated an increased risk of hypoglycemia with multiple GLDs[35]. Gad et al[7] reported in their meta-analysis that patients with T2D on SGLT-2i had fewer and less severe hypoglycemic episodes than those treated with sulfonylureas. While SGLT2i use can be associated with an increased risk of hypoglycemia in those using insulin or oral insulin secretagogues (such as sulfonylureas), these combinations were rarely used in the studies reviewed in the meta-analysis. Therefore, patients should exercise caution when using such combination therapies during fasting hours. However, glycemic control did not worsen with SGLT2i use throughout the study periods, as indicated by comparable changes in HbA1c and FPG between the SGLT2i and non-SGLT2i groups. Although the previous meta-analysis by Gad et al[7] demonstrated a significant reduction in HbA1c in the SGLT2i group compared with the non-SGLT2i group, we did not observe this effect. Incorporating additional studies with more effective GLDs, such as GLP-1 receptor agonists, in the non-SGLT2i comparator group may clarify the differences between our meta-analysis and the previous one.

Treatment with SGLT2i is associated with weight loss, attributed to negative energy balance induced by glycosuria[36]. We observed an average net weight loss of 0.69 kg in the SGLT2i group compared to the group not taking SGLT2i, but this difference was not statistically significant. Similarly, Gad et al[7] reported a non-significant trend toward weight reduction with SGLT2i. In all studies included in this meta-analysis, SGLT2i was initiated well before Ramadan, and the drug’s effect on weight is self-adaptive[36]. In most studies, the follow-up period was too short to show measurable changes in body weight. Furthermore, GLP-1 agonists, which have a more significant impact on body weight reduction, were used in the comparator arm of some included studies. As Ramadan fasting often lasts more than 12 hours in many regions, this fasting pattern resembles time-restricted eating (TRE), which is associated with metabolic and weight-loss benefits[37,38]. The liver stores glucose as glycogen during the fed state, which is depleted after about 12 hours of fasting, leading the body to utilize fat stores and improve adiposity in those who practice TRE[38,39]. This metabolic switch enables the body to use ketone bodies produced by fat breakdown during TRE. Enhancing ketogenesis with SGLT2i in this state may confer cardiometabolic benefits in patients with T2D during Ramadan fasting; however, this concept requires validation in future studies[40]. Similarly, the potential cardiovascular benefits of improvements in systolic and diastolic BP resulting from SGLT2i use during the fasting season require further investigation.

A previous study found that 73.9% of patients with T2D in the high-risk group (recommended not to fast) and 96.2% in the moderate-risk group (recommended against fasting) intended to fast during Ramadan[41]. This emphasizes the importance of individualized pre-Ramadan education on adequate fluid replacement, hypoglycemia, when to break the fast, and proper medical evaluation of risk factors and concomitant medicines to prevent complications that can compromise fasting quality and lead to breaking the fast. A recent systematic review on the same topic concluded that SGLT2i could be initiated at any point in the management of a person living with diabetes, including before and during Ramadan, with limited evidence from small studies on volume contraction and reductions in eGFR[8]. However, as mentioned above, taking special precautions with vulnerable patient populations is anticipated to improve clinical outcomes for these individuals. Given the increased risk of volume depletion, they should receive counseling on adequate fluid intake.

Strengths and limitations

Our meta-analysis's main strength lies in its relevance, as the increased use of SGLT2i in daily clinical practice makes this updated study, with a moderate sample size, valuable for providing reasonable evidence on their safety and effectiveness in the fasting population. However, the meta-analysis has many limitations. The included studies, except for one, were prospective observational, which limits their ability to generate high-quality evidence. The baseline characteristics of the participants in the two groups were not homogeneous across most studies. The patients in the SGLT2i group across all included studies received other GLDs, including insulin in some studies, in addition to an SGLT2i. Additionally, the comparator arm included various GLDs, making the comparator group heterogeneous. Therefore, it is difficult to determine whether the outcomes were solely attributable to SGLT2i. None of the studies included a control group that used SGLT2i without fasting during Ramadan. Therefore, it is challenging to determine whether AE and changes in other outcomes are due to the medication or to fasting, as fasting for such extended periods can have its own health impacts. All the studies extended beyond the Radman month, making it more difficult to determine the actual effects of the drugs during fasting. Furthermore, the duration of some studies was insufficient to assess changes in HbA1c accurately. Moreover, the RoB was moderate in 75% of the studies, with serious or critical risks in the remaining studies. We were unable to evaluate quantitative measures of agreement, such as the κ statistic or percentage agreement, between independent reviewers during the study selection or risk-of-bias assessment because the statistical software (RevMan web) we used did not support these analyses. For the same reason, we could not perform statistical tests (Egger or Begg tests) or formal adjustments (trim-and-fill) for publication bias. We also did not conduct a GRADE assessment to provide levels of certainty for the outcomes, as almost all included studies were observational. Additionally, for most outcomes, we observed considerable heterogeneity among studies. Although we performed sensitivity analyses for these outcomes, subgroup analyses were not feasible due to insufficient studies across subgroups, including study type. Despite these limitations, the current evidence from this meta-analysis reassures us that adults with T2D taking SGLT2i during Ramadan can continue using these medications, given their reasonable safety and efficacy.

CONCLUSION

SGLT2i may be generally safe and effective for glycemic control in adults with T2D who fast during Ramadan. Due to the observational design, high heterogeneity, and fragile evidence, the results are less robust and should be interpreted with caution. Patients should be aware of the higher risks of volume depletion, hypotension, and postural instability, particularly among frail, elderly individuals or those on high-risk medications. Large multicenter RCTs involving diverse ethnic groups are crucial for gaining a clearer understanding of the safety of SGLT2i during Ramadan.

ACKNOWLEDGEMENTS

We thank Dr. Marina G Kudiyirickal, BDS, MJDF-RCS, MSc, PhD, for providing the audio clip for the core tip of this paper.