Choice of Glucose-Lowering Drugs as Initial Monotherapy for Type 2 Diabetes Patients with Contraindications or Intolerance to Metformin: A Systematic Review and Meta-Analysis

Background: There are multiple glucose-lowering drugs available as alternative initial monotherapy for type 2 diabetes patients with contraindications or intolerance to metformin. However, little comparative and systematic data are available for them as initial monotherapy. This study estimated and compared the treatment effects of glucose-lowering drugs as initial monotherapy for type 2 diabetes. Methods: PubMed, Web of Science, Embase, CNKI, Chongqing VIP, and WanFang Data from 1 January 1990 until 31 December 2020 were searched for randomized controlled trials which compared a glucose-lowering drug with placebo/lifestyle-intervention for type 2 diabetes. Drug classes included metformin, sulfonylureas (SUs), thiazolidinediones (TZDs), glinides (NIDEs), α-glucosidase inhibitors (AGIs), dipeptidyl peptidase-4 inhibitors (DPP-4is), sodium-glucose cotransporter-2 inhibitors (SGLT2is), insulins (INSs), and glucagon-like peptide-1 receptor agonists (GLP-1RAs). Results: A total of 185 trials were included, identifying 38,376 patients from 56 countries across six continents. When choosing an initial drug monotherapy alternative to metformin, SUs were most efficacious in reducing HbA1c (−1.39%; 95% CI −1.63, −1.16) and FPG (−2.70 mmol/L; 95% CI −3.18, −2.23), but increased hypoglycemia risks (5.44; 95% CI 2.11, 14.02). GLP-1RAs were most efficacious in reducing BMI (−1.05 kg/m2; 95% CI −1.81, −0.29) and TC (−0.42 mmol/L; 95% CI −0.61, −0.22). TZDs were most efficacious in increasing HDL-C (0.12 mmol/L; 95% CI 0.07, 0.17). SGLT2is were most efficacious in lowering SBP (−4.18 mmHg; 95% CI −4.84, −3.53). While AGIs conferred higher risk of AE-induced discontinuations (2.57; 95% CI 1.64, 4.03). Overall, only GLP-1RAs showed an integrated beneficial effect on all outcomes. Our results also confirmed the intraclass differences in treatment effects across drugs. Most trials were short-term, and no significant differences in mortality, total vascular events, myocardial infarction, heart failure, stroke, or diabetic nephropathy were observed across drug classes. Conclusions: Our results suggest a potential treatment hierarchy for decision-makers, with GLP-1RAs being the preferred alternative therapy to metformin regarding their favorable efficacy and safety profiles.


Introduction
Diabetes is among the top ten causes of death, imposing serious threats to global health and economy [1].
Globally, one in eleven adults were living with diabetes (463 million), 4.2 million deaths were due to diabetes, and 10% of the health expenditure (USD 760 billion) was spent on diabetes in 2019 [1]. Type 2 diabetes accounts for around 90% of all diabetes, which is characterized by chronic hyperglycemia [2]. Hyperglycemia, if left poorly managed, can lead to severe diabetes-related vascular events and even death. Conversely, if appropriately managed, the events can be prevented or delayed [1,3].
Regarding the heavy disease burden and frequent therapy alteration, it is of value to choose a rational drug as alternative initial monotherapy when metformin is contraindicated or intolerant. Evidence shows that focusing solely on glucose control has a limited effect on reducing the risks of diabetes-related vascular events and death, while a comprehensive control of cardiovascular risk factors (e.g., hypertension, dyslipidemia) can bring greater benefits, particularly among patients with special characteristics and needs [3,6]. In addition, the diabetes associations recommend a patient-centered approach to guide the choice of drugs, considering efficacy, weight impact, hypoglycemia risk, effect on cardiovascular and renal comorbidities, and patient preferences [4,7]. It is essential to fully understand the advantages and disadvantages of each drug before making a choice. However, there is little systematic and comparative data available for all the alternative drugs as initial monotherapy of type 2 diabetes [4]. Therefore, this study systematically estimated and compared the treatment effects of all available glucose-lowering drugs when used as initial monotherapy for type 2 diabetes, so as to enable decision-makers to make informed choice when choosing an alternative drug as initial monotherapy for patients with contraindications or intolerance to metformin.

Methods
This study was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [8] and was registered at PROSPERO (CRD42020170769). We estimated the drug-specific effects of each drug after removing the effects of placebo (PBO) and LI and compared the drug classes against each other.

Results
The initial search identified 30,124 records, of which 420 records were assessed in full texts. Finally, 185 trials were included , which identified 38,376 patients randomly assigned to 23 drugs from eight classes or PBO/LI (Figure 1). The sample size of the trials varied from 16 to 888. The study duration ranged from 12 weeks to 108 weeks. Metformin was studied in 40 trials, with pioglitazone (23 trials), acarbose (22 trials), and rosiglitazone (21 trials) being the three next most studied drugs. No eligible trial was detected for gliquidone, mitiglinide, beinaglutide, or insulin. The participants were from 56 countries across six continents. Mean age varied from 30.8 years to 74.5 years, with a diabetes duration of 0.23 year to 9.9 years. The characteristics with the risk of bias of the trials are presented in Supplementary Table S3. The publication bias is shown in Supplementary Figure S1.       For HbA1c, FPG, BMI, TC, HDL-C and SBP, data are WMD (95% CIs); for hypoglycemia, death, AE-induced discontinuations, total vascular events, myocardial infarction, heart failure, stroke, and diabetic nephropathy, data are RR (95% CIs). * Statistically significant differences. Across the drug classes, SUs produced significantly greater HbA1c reductions than all other drug classes, ranging from 0.40% (95% CI 0.09, 0.72) vs. GLP-1RAs to 0.95% (95% CI 0.61, 1.29) vs. NIDEs. In addition, GLP-1RAs, metformin, and TZDs were superior to DPP-4is, AGIs, and NIDEs. SGLT2is worked better than DPP-4is and NIDEs. The greatest contrast among all comparisons was detected between SUs and NIDEs ( Table 2). Table 2. Treatment effects of glucose-lowering drugs compared with each other.    Across the drug classes, metformin, GLP-1RAs, and AGIs produced significantly greater BMI reductions than DPP-4is, TZDs, and SUs, whilst SGLT2is were superior to SUs. The largest difference was seen between metformin and SUs (−2.50 kg/m 2 ; 95% CI −3.96, −1.04), with the smallest difference seen between AGIs and DPP-4is (−0.96 kg/m 2 ; 95% CI −1.87, −0.05) ( Table 2).  (Table 1; Supplementary Figures S26-S33).

Mortality and Vascular Outcomes
A total of 116 trials (24,976 patients) provided data on mortality, while most of them (101/116) reported that no death occurred during the trial. For the fatal events, most were considered by the investigator as not related to the study drug. A total of 32 trials (11,397 patients) provided data on vascular outcomes, 11 of which reported that no vascular events occurred. The main events reported were myocardial infarction (17 trials), heart failure (10 trials), stroke (9 trials), and diabetic nephropathy (9 trials). Compared with PBO/LI, all drugs had a neutral effect on mortality, total vascular events, myocardial infarction, heart failure, stroke, and diabetic nephropathy (Table 1; Supplementary Figures S58-S95).
Across the drug classes, there was no significant difference observed regarding mortality, total vascular events, myocardial infarction, heart failure, stroke, or diabetic nephropathy (Table 2; Supplementary Table S4).
Across the drug classes, AGIs conferred significantly higher risk of AE-induced discontinuations than TZDs (

Sensitivity Analysis
The sensitivity analyses results did not significantly change the overall results ( Supplementary Figures S104-S110).

Discussion
Evidence-based healthcare decision-making needs comparisons of all relevant competing therapies, on a full range of intermediate outcomes (e.g., blood glucose, blood lipids, and blood pressure), important AEs (e.g., hypoglycemia), and long-term outcomes (e.g., vascular events and mortality). This study identified important interclass and intraclass differences in these outcomes across all available glucose-lowering drugs when used as initial monotherapy for type 2 diabetes, to help decision makers rationally choose an alternative initial drug when metformin is contraindicated or intolerant.
The core of managing type 2 diabetes is to control blood glucose [3,4]. Intensive glucose control plays an important role in reducing the risks of diabetes-related vascular events and death [200]. This study showed that SUs generated the greatest HbA1c reductions, followed by GLP-1RAs, metformin, TZDs, SGLT2is, DPP-4is, AGIs, and NIDEs; and the HbA1c reductions observed were around the recommendation of clinical guidelines [3]. This is partly consistent with the results of Sherifali's (2010) study, which found SUs to have greater HbA1c effect than metformin, TZDs, DPP-4is, AGIs, and NIDEs by synthesizing 61 English-language studies that compared an oral drug with placebo in patients with or without a background of other drugs [201]. Regarding FPG effects, SUs were also shown to have the largest FPG reductions, followed by TZDs, metformin, GLP-1RAs, SGLT2is, AGIs, DPP-4is, and NIDEs. In addition, we found the HbA1c reduction with liraglutide (1.17%) was more similar to SUs (1.39%) than other GLP-1RAs, such as lixisenatide (0.60%). We also found large differences in FPG reductions within GLP-1RAs, with 1.97 mmol/L for liraglutide vs. 0.97 mmol/L for lixisenatide. These findings confirmed important intraclass differences in glucose effect across drugs. Thus, optimal therapeutic decisions should also consider the comparative effects across individual drugs, especially when there is a need to choose from drugs belonging to the same classes.
Overweight or obesity is a common comorbidity of type 2 diabetes (85%) [202]. Weight gain may increase medical costs, while weight loss can improve cardiovascular risk factors, and reduce mortality and costs in patients with both type 2 diabetes and overweight/obesity [203][204][205][206]. This study reported that metformin and GLP-1RAs reduced BMI; while NIDEs, AGIs, DPP-4is, and SGLT2is maintained BMI; and SUs and TZDs increased BMI. The results were roughly consistent with that of Maruthur's (2016) study, which compared the weight effect across monotherapies and metformin-based combinations of six drug classes from 85 English-language studies, and found weight was maintained or reduced with metformin, GLP-1RAs, DPP-4is, and SGLT2is, but increased with SUs and TZDs [207].
Dyslipidemia and/or hypertension, as independent risk factors for cardiovascular diseases, are also common comorbidities of type 2 diabetes (72%) [3,6]. Patients with type 2 diabetes, dyslipidemia, and hypertension are six times more likely to have cardiovascular diseases compared to those with diabetes alone [6]. Their integrated controls play important roles in reducing vascular events and costs, especially for those with long diabetes duration, old age, a history of cardiovascular diseases, or multiple risk factors [3]. This study showed that GLP-1RAs, AGIs, and metformin reduced TC, but SGLT2is increased TC; GLP-1RAs were superior to other drug classes in reducing TC. We also noted that within DPP-4is, alogliptin and sitagliptin showed opposite TC effects, confirming important intraclass differences across drugs. Regarding HDL-C effects, only TZDs, SGLT2is, and metformin increased HDL-C, and TZDs were the most efficacious drugs. This result is partly consistent with that of Bolen's (2007) study, which found TZDs to have beneficial effects on HDL-C compared with other drugs by synthesizing 53 English-language studies on comparing monotherapies or combination therapies, although it did not include DPP-4is, SGLT2is, or GLP-1RAs [208].
Blood pressure lowering is related to improved mortality and other clinical outcomes for type 2 diabetes patients, and even a small reduction of 2.4 mmHg in SBP has an effect in reducing cardiovascular events [209,210]. As this study showed that SGLT2is and GLP-1RAs conferred SBP reductions by 2.74~5.36 mmHg, they may have beneficial effects on cardiovascular events. Our findings roughly correspond to that of Tsapas's (2021) study, which reported that treatment with SGLT2is or GLP-1RAs with or without a metforminbased background therapy were more efficacious than other drug classes in reducing SBP by estimating 204 English-language studies [211].
A patient-centered approach recommends minimizing AEs, vascular events, and mortality beyond optimizing efficacy. Hypoglycemia as a common AE in treating type 2 diabetes is associated with multiple complications and high economic burden, and fear of it may reduce treatment adherence and prevent optimal glucose control [212,213]. This study showed that SUs had higher hypoglycemia risks than other drug classes, which is consistent with the previous studies [207,208]. In addition, large intraclass differences were noted regarding hypoglycemia risk. For example, within DPP-4is, alogliptin was reported to increase the risk but linagliptin reduced the risk. Overall, amongst all monotherapies, only GLP-1RAs produced a comprehensive beneficial effect on HbA1c, FPG, BMI, TC, and SBP, while not increasing hypoglycemia risk. As GLP-1RAs did not have a significant beneficial effect on HDL-C, while TZDs conferred the best effect on HDL-C, GLP-1RAs and TZDs may be the best mixed therapy to control HbA1c, FPG and lipid profile all together for patients with contraindications or intolerance to metformin. In terms of longterm outcomes, no significant difference was observed regarding total vascular events, myocardial infarction, heart failure, stroke, diabetic nephropathy, or mortality across the drug classes. This may be because when patients start the initial drug monotherapy they are most likely in the early stages of diabetes and may not develop serious vascular events in the short term. However, most of the included studies were short-term trials, lasting less than one year, thus with limited ability to observe long-term vascular outcomes and deaths.
Only a few systematic reviews were detected to have compared several drug classes, however, they were not focused on monotherapy, not covered all available drug classes, and only included English-language studies [201,207,208,211]. Thus, our study has noteworthy strengths in that it is the first attempt to comprehensively and simultaneously estimate and compare the treatment effects of all available glucose-lowering drugs when used as initial monotherapy for type 2 diabetes across multiple clinically meaningful intermediate and long-term outcomes, by merging data from English-and Chinese-language studies. The inclusion of 62 Chinese-language studies allowed the inclusion of populations and drugs that had not been studied previously in other settings. Our observations expanded the knowledge by supplementing with evidence from China. Our results may enable decision makers to better understand the similarities and differences across drugs, as to make comparisons between drugs at a glance when choosing an alternative monotherapy to metformin for type 2 diabetes patients.
Certain limitations should be acknowledged. First, from the methodological point of view, a meta-analysis has some weakness, including high risk of bias within studies, potential publication bias, and clinical heterogeneity. Second, most of the included trials were short-term trials, lasting less than one year, with limited ability to observe long-term clinical outcomes, such as vascular events and mortality. Ideally, as therapeutic decisionmaking should depend on long-term effectiveness, the generalizability of our results may be limited due to paucity of long-term evidence. Third, we only searched PubMed, Web of Science, Embase, CNKI, Chongqing VIP, and WanFang Data for eligible RCTs, thus RCTs indexed only in other databases (e.g., Scopus) may not be included in our review, which may induce potential bias.

Conclusions
Rational choice of drugs should be individualized, based on unique patient characteristics and benefit-risk profile of each drug. When choosing an alternative drug as initial monotherapy for type 2 diabetes patients with contraindications or intolerance to metformin, our results suggest a potential treatment hierarchy, with GLP-1RAs being preferred in terms of their favorable efficacy and safety profiles. In addition, SUs are most suggested for patients whose HbA1c or FPG is far away from the target, but least suggested for those wishing to minimize hypoglycemia. GLP-1RAs are most suggested for patients for whom BMI management or TC management is an emphasis. TZDs are most suggested for patients for whom HDL-C control is a priority. SGLT2is are most suggested for patients having a need to reduce SBP. While AGIs are least suggested for patients wishing to avoid AE-induced discontinuations. Our observations also corroborated the intraclass differences in treatment effects among drugs. Future studies should pay more attention to long-term studies to obtain more precise data on vascular outcomes and mortality. Moreover, this study also provided comparative and systematic clinical data for international researchers to conduct cost-effectiveness modelling studies, which will add long-term costs and effectiveness evidence for the rational choice of initial monotherapies alternative to metformin and improving the diabetes management as well as healthcare resource allocation.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jcm11237094/s1. Author Contributions: S.G. contributed to the study design, data search and collection, statistical analysis, data interpretation, and writing and review of the manuscript. H.D. contributed to the study design, data interpretation, and review of the manuscript. X.H., X.S. and Y.G. contributed to the data collection, statistical analysis, and data interpretation. L.S. and X.Z. contributed to review of the manuscript. M.H. contributed to the data collection and data interpretation. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement:
The datasets used and/or analyzed during the current study are available from the corresponding authors on reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.