Design and Methodological Characteristics of Studies
The key features of study design are presented in Table 1 for 10 RCTs and in Table 2 for 20 observational studies.
Sulfonylureas and drugs of the incretin class appear to be the most widely studied non-insulin treatments in comparative trials during Ramadan. However, individual RCTs sometimes did not differentiate among the different sulfonylureas used by patients, simply grouping different sulfonylureas (e.g., gliclazide, glibenclamide, glimepiride; glipizide) together as a class when reporting results (Table 1). This may not be appropriate, as these drugs have durations of actions of 12–18 h, 12–16 h, 12–24 h, and 6–10 h, respectively . In addition, pooled data from a number of agents in the sulfonylurea class mask the different hypoglycemia risks associated with the individual drugs, as newer generation sulfonylureas such as gliclazide are associated with a lower hypoglycemia risk compared with other sulfonylureas . In fact, a meta-analysis showed no difference in hypoglycemia incidence with gliclazide compared with DPP-4 inhibitors . By contrast, results for drugs of the incretin class have been reported individually for vildagliptin (two RCTs [38, 39] and five observational studies [40–44]) for sitagliptin (three RCTs [45–47]) and liraglutide (two RCTs [33, 34]). The use of metformin was common, but not universal (Table 1). Both RCTs using liraglutide included metformin [33, 34], as did the two studies using vildagliptin [38, 39] and one of three RCTs using sitagliptin . However, metformin was optional for two of the RCTs using sitagliptin [46, 47] and one trial using pioglitazone . Metformin was not used in the two trials examining repaglinide [49, 50].
Treatments used in the observational studies were quite heterogeneous, with some studies even combining patients using various combinations of OADs and/or insulin and/or diet alone into a single group [51–57] (Table 2).
Eligibility, Number, and Comparability of Patients
In the RCTs, the number of subjects varied from 41 to 1066. Inclusion and exclusion criteria were mentioned in some way for all RCTs, but the level of details presented varied greatly. All of the RCTs described prior diabetes treatments, and most of the trials enrolled subjects previously treated with metformin or sulfonylurea either as monotherapy or in combination. However, in a few trials, a minority of patients had been treated with thiazolidinediones, glinides, or acarbose (Table 1). A specific cutoff for glycemic control (HbA1c) for eligibility was only specified for four RCTs (7–10% ; 6.5–12.0% ; ≤8.5% ; ≤10% [46, 47]). One trial indicated that participants had to be “well controlled” . Most (n = 7/10) trials excluded patients with a history of recurrent hypoglycemia, severe hypoglycemia, or hypoglycemia unawareness [33, 45–48, 50], but others made no such exception [38, 39, 49]. Exclusion of patients with at least some serious systemic diseases was reported in all but three trials, but the level of stringency reported varied [39, 47, 49].
There was also a considerable range in the number of patients participating in the observational studies (23–1333 patients; Table 2). With the exception of four reports [54–56, 58], all of the observational studies mentioned patient inclusion/exclusion criteria to some extent. However, as with RCTs, the stringency of these criteria varied considerably among studies. For example, some studies indicated that all patients with T2D were included  or that all consecutive patients were enrolled , whereas others specified more detailed inclusion criteria such as age, type of prior treatment, and/or willingness to fast. Most studies did not specify any HbA1c cutoff for eligibility, although one required HbA1c ≤9.0% , two required HbA1c ≤8.5% [40, 43], one required HbA1c ≤8.0% , and one just indicated that patients had to be in “good control” .
The same variation was noted with respect to exclusion criteria, with some observational studies indicating no exclusion criteria other than insulin treatment (e.g., [44, 60]) and others listing very specific exclusion criteria (e.g., duration of diabetes, level of glycemic control, prior treatments, incomplete fasting, incomplete medical records, and/or serious renal, hepatic, cerebrovascular or cardiovascular disease, pregnancy or lactation; retinopathy, and uncontrolled hypertension [29, 40, 41, 53, 61, 62]). In contrast to the RCTs, few observational studies mentioned excluding patients with severe hypoglycemia , recurrent hypoglycemia [61, 62], or hypoglycemia unawareness , but otherwise this appeared uncommon. One study using continuous glucose monitoring (CGM) indicated that “low-risk” patients were enrolled .
Choice and Measurement of End Points
The critical factor of end points was not consistent across studies, either with respect to the choice or efficacy or safety as a primary end point, or how those end points were measured. Three RCTs used a primary efficacy end point [34, 49, 50] and four RCTs indicated that the primary end point was a safety end point [38, 45–47], and in two, a primary end point was not explicitly stated or clear, with multiple efficacy and safety end points being listed [39, 48]. In one trial, a specific composite end point was used  (Table 1). In some studies where efficacy was indicated to be the primary end point, safety end points were also included. Conversely, in some studies where safety was listed as the primary end point, some efficacy measures were also reported.
When glycemic efficacy end points were reported, there were also considerable differences in how they were measured, ranging from CGM to self-monitored blood glucose (SMBG), to clinic visits with laboratory measures, to patient diaries with or without review by a physician, to phone interviews and retrospective review of medical records. Secondary end points such as weight change, blood pressure, lipids, and quality of life were inconsistently reported.
Definition of Safety (Hypoglycemia) Outcomes
In the RCTs, the definition of hypoglycemia, as well as how hypoglycemia was measured, varied. Two trials used symptoms only [47, 50]. When a cutoff value was mentioned, most trials specified a value of <3.9 or ≤3.9 mmol/L, although two trials utilized other values (<3.1 mmol/L  and <2.8 mmol/L ; Table 1
). In one trial, the cutoff point was unclear . There were also cases where a primary end point used symptoms as the criteria, but the results for additional hypoglycemic end points were also reported (e.g., [46, 47]).
There was also considerable variability in how hypoglycemia was defined and measured in the observational studies (Table 2). Some studies used patient-reported symptoms only [29, 51, 57, 60], but more typically used symptoms with or without blood glucose (BG) measurements [40, 41, 43, 44, 52, 56, 62]. In one of these cases, the cutoff was not specified . The BG cutoff values used were also heterogeneous (<3.9 mmol/L ; ≤3.9 mmol/L ; 2.8–3.9 mmol/L ; <3.9 mmol/L ; <3.88 mmol/L ; <3.9 mmol/L ; <3.5 mmol/L ).
Duration of Treatment and Follow-Up
There was great variation in the amount of time patients were assessed, treated, and followed across studies. In the RCTs (Table 1), the run-in period for the non-sulfonylurea treatment varied considerably, even where patients were switching to the identical drug. For example, Anwar et al.  reported a 3-month run-in period for patients switching to repaglinide, whereas Mafauzy  reported 6 weeks. For three trials using incretins, one with vildagliptin  and two with sitagliptin [46, 47], the duration of the run-in period was unclear. The duration of follow-up post-Ramadan, when fasting ceased and normal diet and lifestyle patterns resumed, was also variable and sometimes difficult to discern. For the observational studies (Table 2), there was also great variation in the length of time during which patients were assessed and data obtained.
Blinding of subjects and evaluators during a clinical trial is considered an important protection against bias. Of the 10 RCTs, only two trials were blinded [38, 48] (Table 1). One was a double-blind, double-dummy trial , and the other was double-blinded . By definition, blinding was not possible in the observational studies.
Definition and Characteristics of Fasting
Fasting was not always explicitly defined in studies. In the RCTs, the intent to fast was either stated or assumed based on screening criteria in nine out of ten trials, but only one trial indicated a specific minimum amount of time (i.e., at least 10 days) for fasting . One trial indicated that participants would be excluded if they expected to break their fast for >3 days . It has been reported that the ability to keep a fast may also depend on the type of treatment , but this was not regularly reported and, when reported, not in sufficient detail to allow for meaningful interpretation of the likely impact of non-adherence on treatment. Adherence to fasting was not described for six RCTs [33, 34, 45, 48–50]. When adherence was reported, the statistic used varied (Table 1). For example, one study reported the mean number of days of fasting for the two treatment groups (i.e., 28.3 ± 3.0 and 28.1 ± 3.8 days) , whereas two other trials reported the proportion of people in each treatment group who did not break their fast except to treat symptoms of hypoglycemia (i.e., 89.7% to 98.4% of participants) [46, 47]. It was often unclear what constituted an interruption of fasting, whether an interruption was temporary or permanent, or under what circumstances an interruption might be of sufficient duration to be considered a break in the fast (non-adherence). For example, one RCT indicated that eight individuals in one of the treatment arms broke their fast, but did not quantify or otherwise describe what that meant .
In the observational studies, intention to fast (or actual fasting) during Ramadan was either an implicit or explicit inclusion criteria. In some cases, the period of intending to fast was pre-specified in some detail (>10 days [43, 55, 57]; ≥15 days ; ≥20 days ; or throughout Ramadan [59, 62]). In general, there was more attention to describing adherence to fasting in the observational studies, with 13/20 studies reporting some measure of adherence (Table 2). This was most frequently framed as the mean number of days of fasting. However, an inspection of these mean values in the table (e.g., from approximately 20.4 to 28.6 days) makes it clear that some participants did not fast the entire month of Ramadan.
Ramadan-focused Education and Quality of Life Assessment
The potential value of patient education has been evaluated in a number of clinical studies. For example, patients with T2D using OADs who received a Ramadan-focused education and awareness in diabetes (READ) program demonstrated a decrease in hypoglycemic episodes during Ramadan . Another comparative trial in Thailand showed that patients in the group receiving education prior to Ramadan significantly reduced the number of hypoglycemic episodes (P = 0.013), diastolic blood pressure, (P = 0.041), and consumption of sweetened food (P = 0.002), and the number of patients with hypoglycemic symptoms was also lower (P = 0.013) . A comparison of patients recruited from clinics in four Muslim nations who received an individualized educational program vs. those who did not showed significant improvements in a variety of diabetes management outcomes, such as modifying their treatment plan (P < 0.001), performing SMBG at least twice daily (P < 0.0001), and having improved knowledge about hypoglycemic signs and symptoms (P = 0.0007), as well as fewer severe hypoglycemic events (P = 0.0017) . One uncontrolled study described the practicality of implementing a Ramadan-specific educational program through diabetes-specialist centers . The study reported that drug dose and timing were modified in 90.5% of patients with T2D.
The extent to which Ramadan-focused education is included in trials whose primary goal is to compare drug treatment regimens is unclear. Among the randomized trials summarized in Table 1, only one study explicitly mentioned providing Ramadan-focused education, which was given to patients in both arms of the trial . However, two other trials mentioned providing medical counseling regarding risks of hypoglycemia during fasting [46, 47]. It is unclear to what degree these issues may have been raised during the routine informed consent process in other trials. Among the observational studies (Table 2), three mentioned providing Ramadan-focused advice about diabetes management [42, 44] or information about the risks associated with fasting , again provided to all patients participating in the studies. Future trials should explicitly include and describe the extent of Ramadan-focused education for participants.
With regard to patient-reported outcomes and quality of life assessment, only one of the reviewed studies assessed treatment satisfaction and one assessed lifestyle changes [56, 69].
Reporting of Diet and Exercise
Changes in diet and exercise were reported infrequently, and when reported, typically, this was not in sufficient detail to enable adequate account for confounding effects of these changes when evaluating the efficacy and/or safety of the treatments being studied. Only one RCT provided such data, with physical activity assessed by the International Physical Activity Questionnaire (IPAQ), and diet information reported in the form of changes in score for total Metabolic Equivalent Tasks (METS) for each treatment group .
Among the observational studies, as with the RCTs, reporting of data on diet and exercise was very limited, with only 3/20 studies providing information. Siaw et al.  reported general trends in diet and exercise during Ramadan, describing these only as “more,” “less,” or “unchanged”. An analysis of the proportion of patients reporting reduction (64.7%), increase (5.9%), or no changes (29.4%) in dietary intake, as well as data on physical exercise, indicated that there were no significant associations between these categories and improvements in mean HbA1c during fasting. Sahin et al.  also reported similar categorical groupings of changes in diet and exercise, whereas Ahmadani et al.  reported caloric intake for 1 day before and after fasting, in addition to a simple dichotomous outcome of diet having changed or not and exercise having changed or not. In some cases when it was mentioned in the Methods section of a paper that information on diet and/or exercise was obtained from patients, the results for those variables were not presented in the trial report. For example, Bonakdaran et al.  indicated that patients were required to record dietary programs, time of meals, quality and quantity of meals, and time and extent of exercise, but those data were neither reported nor apparently factored into the analysis of BG readings or hypoglycemic events. Zargar et al.  also mentioned that data on meals were obtained, but no results were presented.
Classes of Diabetes Treatments Studied
The 10 RCTs were heterogeneous with respect to diabetes treatments evaluated, with five comparing sulfonylureas to DPP-4 inhibitors, two comparing sulfonylureas to GLP-1 analogs, two comparing sulfonylurea to repaglinide, and one comparing sulfonylurea to pioglitazone (Table 1). In some cases, only one sulfonylurea was used, and in others, a variety of sulfonylureas was permissible. Use of metformin was inconsistent, with some trials allowing it, others mandating it, and others excluding metformin. No RCTs have compared receptor GLP-1 receptor agonists and DPP-4 inhibitors head-to-head in the treatment of T2D during Ramadan. This is in contrast to at least seven such studies published in patients with T2D not focused on Ramadan fasting (see ). Being a more recently introduced class of drugs, sodium-glucose co-transporter 2 inhibitors have not yet been studied in fasting patients with T2D during Ramadan.
Observational studies were even more heterogeneous with respect to the treatments studied and it was not uncommon to group together patients using treatments ranging from lifestyle management to OADs to insulin, either individually or in combination (Table 2).
Adherence to Treatment
In the RCTs, adherence to treatment was difficult to discern. One RCT reported the proportion of missed doses , and others reported the proportion of patients who changed the dose or timing of their medication [46, 47]. One study reported the proportion of completers reaching the optimal dose , and another reported the proportion of patients exposed to treatment for >6 and >10 weeks .
Limited reporting of adherence to treatment was also typical in the observational studies. When mentioned (6/20 studies), this was usually brief and in the form of proportion of patients with missed doses, proportion making adjustments to dose, or changes in the timing of dose (e.g., [43, 51, 54, 56, 57, 60]). Few studies provided detailed measures of adherence (e.g., overall adherence per treatment group, proportion of patients with treatment change in each group, type of treatment change, number of times the treatment was not taken, and/or number of times the treatment was not taken due to hypoglycemia or fear of hypoglycemia) [40, 41]. Across these studies, analysis of how these various measures of adherence to each treatment were related to the frequency of hypoglycemic events was generally lacking.