To compare the efficacy and safety of sodUIm-glucose co-transporter-2 (SGLT2) inhibitors and metformin in adults with type 1 diabetes mellitus (T1DM).
MethodsRandomized clinical trials until February 2019, designed to assess the efficacy and safety of SGLT2 inhibitors/metformin in adults with T1DM, were searched on PubMed, EMBASE, the Cochrane Library, and Web of Science. Safety and efficacy data were synthesized using Bayesian network meta-analyses.
ResultsTwenty eligible studies with 5868 participants were included in network meta-analysis. SGLT2 inhibitors provided greater reductions in HbA1c than placebo [weighted mean difference (WMD) −0.40 (95 % confidence interval (CI) −0.47, −0.32)] and metformin (WMD: −0.32; 95 % CI: −0.47, −0.14). Both SGLT2 inhibitors and metformin promoted greater reductions in body weight than placebo. SGLT2 inhibitors caused greater reductions in body weight than metformin (WMD: −1.54; 95 % CI: −2.93, −0.09). Both SGLT2 inhibitors and metformin provided greater reductions in total insulin dose than placebo, while no difference between metformin and SGLT2 inhibitors was found. No difference in severe hypoglycemia was found between SGLT2 inhibitors and metformin. SGLT2 inhibitors induced a higher risk for diabetic ketoacidosis (DKA) than metformin/placebo.
ConclusionSGLT2 inhibitors provided greater reductions in HbA1c and body weight than metformin/placebo. Both SGLT2 inhibitors and metformin induced greater reductions in total insulin dosage than placebo, with no significant differences observed between SGLT2 inhibitors and metformin. SGLT2 inhibitors induced a higher risk for DKA than metformin/placebo.
Comparar la eficacia y la seguridad de los inhibidores del cotransportador-2 de sodio-glucosa (SGLT2) y de la metformina en adultos con diabetes mellitus tipo 1 (DMT1).
MétodosSe buscaron en PubMed, EMBASE, la Cochrane Library y la Web of Science, los ensayos clínicos aleatorizados llevados a cabo hasta febrero de 2019 diseñados para evaluar la eficacia y la seguridad de los inhibidores del SGLT2/metformina en adultos con DMT1. Los datos de seguridad y eficacia fueron recopilados empleando el metanálisis en red bayesiana.
ResultadosVeinte estudios, con 5.868 participantes, fueron aptos para ser incluidos en el metanálisis en red. Los inhibidores del SGLT2 proporcionaron mayores reducciones en la HbA1c que el placebo {diferencia de media ponderada (DMP) −0,40 [intervalo de confianza (IC) de 95 %: −0,47, −0,32]}, y la metformina (DMP: −0,32: IC 95 %: −0,47; −0,14). Tanto los inhibidores del SGLT2 como la metformina favorecieron mayores disminuciones de peso corporal que el placebo (DMP: −1,54; IC 95 %: −2,93; −0,09). Los inhibidores del SGLT2 produjeron mayores disminuciones de peso corporal que la metformina (DMP: −1,54; IC 95 %: −2,93; −0,09). Los inhibidores del SGLT2 y la metformina dieron lugar a mayores reducciones de la dosis total de insulina que el placebo, no encontrándose diferencias entre la metformina y los inhibidores del SGLT2. En cuanto a la hipoglicemia grave, no se encontraron diferencias entre los inhibidores del SGLT2 y la metformina. Los inhibidores del SGLT2 provocaron un mayor riesgo de cetoacidosis diabética (CAD) que la metformina/placebo.
ConclusiónLos inhibidores del SGLT2 proporcionan mayores reducciones en la HBA1c y de peso corporal que la metformina/placebo. Tanto los inhibidores del SGLT2 como la metformina provocaron mayores disminuciones en la dosificación total de insulina que el placebo, no observándose diferencias significativas entre los inhibidores del SGLT2 y la metformina. Los inhibidores del SGLT2 indujeron un mayor riesgo de CAD que la metformina/placebo.
The standard predominant therapeutic strategy for treating type 1 diabetes mellitus (DM1) has been the intensive use of insulin.1,2 Despite progress in the formulation and administration of insulin, there are many challenges still facing patients and physicians. An international comparative study showed that only approximately 14 % of children and 30 % of adults in 19countries or various regions of the world (324,501 participants) had reached the glycated hemoglobin A1c (HbA1c) objectives recommended by the American Diabetes Association.3 The intensive use of insulin predisposes individuals to hypoglycemia, weight gain and large variations in blood sugar levels, which can increase the potential for cardiovascular complications.4–6
The classical drug metformin has been typically employed for more than half a century as a biguanide agent for treating type 2 diabetes mellitus (DM2).7 Metformin acts primarily by inhibiting the production of hepatic glucose, improving muscle absorption of glucose and reducing intestinal absorption of glucose. Available data have revealed that adjuvant metformin therapy is associated with weight loss, improved lipid profiles and reduced insulin doses in patients with DM1.8,9 Sodium-dependent glucose cotransporter 2 inhibitors (SGLT2i), a new and promising therapeutic modality for DM2, acts mainly by inhibiting glucose reabsorption in the proximal renal tubule and, as a consequence, decreases blood glucose levels through glucose excretion in urine.10 According to the available data, SGLT2i provides numerous benefits, including weight loss, lower HbA1c levels and reduced need for insulin by patients with DM1.11–13
However, only one head-to-head trial to date has compared SGLT2i to metformin in terms of efficacy in DM1. When actual comparisons are scarce, network meta-analyses allow for the synthesis and comparison of treatments through available evidence to estimate direct and indirect comparisons of interest.14,15 Our objective was to compare (through the use of a network meta-analysis) the efficacy and safety of metformin and SGLT2i as a complement for standard insulin therapy for adults with DM1.
Materials and methodsResearch strategyThrough exhaustive searches on PubMed, EMBASE, Cochrane Library and Web of Knowledge, we identified all the randomized clinical trials (RCTs) that researched the safety and efficacy of metformin or SGLT2i in adults with DM1 and that had been published up to February 2019, regardless of language. The search was conducted using various combinations of keywords, such as “(canagliflozin OR dapagliflozin OR sotagliflozin OR LX4211 OR empagliflozin OR metformin)” and “(diabetes mellitus, type 1) OR (type 1 diabetes) OR (insulin-dependent diabetes)”. The exact search is available upon request to the authors. We also identified additional studies through a manual search of all references in the recovered articles.
Study selectionThe RCTs that met the following inclusion criteria were added to this meta-analysis: 1) conducted with humans; 2) the patients with DM1 were older than 18 years; and 3) changes in HbA1c, weight and insulin dose and cases of severe hypoglycemia and diabetic ketoacidosis (DKA) in the intervention and control groups were recorded. We did not consider abstracts or unpublished reports. We excluded case reports, editorials, review articles and letters to the editor. We also excluded articles that had no control group. We also excluded articles from studies that employed adjuvant drug therapy in addition to metformin and SGLT2i. This study was conducted according to the PRISM guidelines (see Appendix, additional material).
Data extractionTwo researchers (Zhang QQ and Wu YC) independently conducted a preliminary review of the studies. The next phase consisted of reviewing (from the point of view of the selection criteria) the abstracts and examination of the complete text. The final selection of articles was performed by consensus of the 2 reviewers. The following characteristics were extracted from each study: the first author, the article publication date, the study design, the study duration, the daily dose of SGLT2i or metformin, the rate of severe hypoglycemia and DKA, the baseline measures and changes in baseline HbAc1 levels, total insulin dose and body weight, both for the intervention and control groups. For the inclusion in our meta-analysis, we did not define a minimum number of interventions or controls.
Evaluation of the study quality and risk of biasTo assess the risk of bias in the included studies, we employed the Cochrane Collaboration Tool for assessing the risk of bias in randomized trials.16 Table S1 (see Appendix, additional material) shows the risk of bias for all included articles.
Statistical analysisInitially, we performed the random-effects meta-analysis by pairs within each direct comparison of treatment before the network meta-analysis, for which we employed STAT 11.0 (Stata, College Station, TX). To assess the heterogeneity, we used I2 values. We performed a network meta-analysis using the Markov chain Monte Carlo Bayesian simulation framework to assess the safety and efficacy of metformin and SGLT2i in adults with DM1. To calculate the network meta-analysis, we employed the Aggregate Data Drug Information System (version 1.17.6).17 Convergence was achieved using Brooks-Gelman-Rubin graphs and the potential scale reduction factors (PSRF) (1PSRF≤1.5). The run-length increased accordingly to improve the convergence model when PSRF was >1.05. The dichotomous values (severe hypoglycemia or DKA) were analyzed using the odds ratio (OR) with a 95 % confidence interval (CI). The continuous variables (HbA1c, body weight and total insulin dose) were analyzed using WMD with a 95 % CI. We constructed network graphs for each endpoint. We employed the node splitting technique to determine the inconsistency of the network meta-analysis when this showed closed loops (p<.05 reveals significant inconsistencies in the network). We performed subgroup analyses to assess the differences in the safety and efficacy of SGLT2i in adults with DM1. We calculated the probability range for each of the treatments in order to show the best results.
ResultsCharacteristics of the included studiesWe identified a total of 1280 entries, of which 1316 were excluded for being duplicates. After reviewing the titles and abstracts, we removed an additional 874 articles. Forty-six other articles were rejected for not meeting the inclusion criteria. Fig. 1 shows in detail the flow diagram followed in the selection process in the meta-analysis. Ultimately, we chose a total of 20 studies with 5868 participants, which included an article on canagliflozin,12 5 on dapagliflozin (2 articles in 1 RCT),18–21 3 on empagliflozin,22–24 4 on sotagliflozin,25–28 6 on metformin25–31 and 1 that directly compared metformin with empagliflozin.32 Of the included studies, 19 were randomized parallel-group trials, and 1 was a crossover controlled RCT. Table 1 shows the characteristics of the included trials. All parameters were balanced between the groups. The definitions of severe hypoglycemia were similar for all studies and were based on the criteria of the American Diabetes Association.
Characteristics of the Included Studies.
Study | Duration, weeks | Study arms, daily | Participants, n | Age, years | Duration of type 1 diabetes, years | Body weight, kg | Baseline BMI, kg/m2 | Baseline HbA1c, % | Basal insulin |
---|---|---|---|---|---|---|---|---|---|
[0,1–10]Canagliflozin | |||||||||
[3,0]Henry RR et al. (2015) | [3,0]18 | [1,0]300mg CANA | [1,0]117 | [1,0]42.8±11.0 | [1,0]21.9±10.6 | [1,0]82.9±15.0 | [1,0]28.1±3.9 | [1,0]8.0±0.5 | MDI: 44±37.6 IU/day |
CSII: 73±62.4 IU/day | |||||||||
[1,0]Placebo | [1,0]117 | [1.0]42.0±11.9 | [1.0]23.3±11.0 | [1.0]83.0±15.4 | [1.0]28.0±3.6 | [1.0]7.9±0.6 | MDI: 45±38.5 IU/day | ||
CSII: 72±61.5 IU/day | |||||||||
[0,1–10] | |||||||||
[0,1–10]Dapagliflozin | |||||||||
[1,0]Henry RR et al. (2015) | [1,0]2 | 10mg DAPA | 15 | 37.5±15.2 | 18.1±14.0 | 78.4±19.9 | 25.8±4.8 | 8.39±0.82 | NR |
Placebo | 13 | 34.5±12.2 | 16.2±9.7 | 78±11.4 | 25.3±3.0 | 8.75±0.92 | NR | ||
[1,0]Dandona P et al. (2017) /Dandona P et al. (2018) | [1,0]24/52 | 10mg DAPA | 296 | 42.7±14.1 | 19.9±11.1 | 82.0±17.3 | 28.1±5.1 | 8.52±0.64 | 59.4±28.2 IU/day |
Placebo | 260 | 42.7±13.6 | 21.2±12.2 | 84.3±18.3 | 28.6±5.2 | 8.53±0.67 | 63.1±29.3 IU/day | ||
[1,0]Mathieu C et al. (2018) | [1,0]24 | 10mg DAPA | 270 | 42.4±12.80 | 19.45±11.90 | 80.06±18.30 | 27.80±5.53 | 8.43±0.69 | 58.68±28.26 IU/day |
Placebo | 272 | 43.0±13.73 | 18.98±11.65 | 78.88±18.87 | 27.62±5.41 | 8.43±0.65 | 56.57±25.23 IU/day | ||
[1,0]Watada H et al. (2018) | [1,0]2 | 10mg DAPA | 14 | 37.1±10.2 | 14.7±12.4 | 59.8±8.6 | 22.2±2.1 | 7.9±0.6 | 38.1±14.7 IU/day |
Placebo | 14 | 42.6±10.6 | 16.9±10.5 | 57.2±10.7 | 22.9±3.4 | 8.1±0.8 | 36.9±14.8 IU/day | ||
[0,1–10] | |||||||||
[0,1–10]Empagliflozin | |||||||||
[1,0]Pieber TR et al. (2015) | [1,0]4 | 25mg EMPA | 18 | 41.9±9.7 | 23.7±14.5 | 76.9±14.5 | 25.4±3.5 | 8.15±0.54 | 0.65±0.23 IU/kg/day |
Placebo | 19 | 40.5±10.6 | 20.5±12.8 | 79.8±13.8 | 25.4±3.7 | 8.18±0.67 | 0.66±0.23 IU/kg/day | ||
[1,0]Shimada A et al. (2018) | [1,0]4 | 25mg EMPA | 12 | 46.6±10.8 | 20.8±13.5 | 60.5±10.2 | 22.6±2.7 | 7.89±0.91 | 0.66±0.15 IU/kg/day |
Placebo | 11 | 43.9±11.7 | 14.8±10.0 | 63.6±7.7 | 23.7±2.6 | 8.23±0.47 | 0.71±0.22 IU/kg/day | ||
[3,0]Rosenstock J et al. (2018) | [1.0]52 | 25mg EMPA | 241 | 45.3±13.9 | 22.5±13.0 | 85.6±18.3 | 29.5±6.0 | 8.06±0.53 | 0.74±0.26 IU/kg/day |
Placebo | 239 | 44.5±13.5 | 22.4±12.4 | 83.4±16.7 | 28.5±5.3 | 8.13±0.57 | 0.70±0.23 IU/kg/day | ||
[1,0]26 | 25mg EMPA | 242 | 44.2±13.5 | 21.2±11.4 | 83.3±18.9 | 28.4±5.6 | 8.19±0.65 | 0.71±0.24 IU/kg/day | |
Placebo | 238 | 42.2±13.2 | 21.7±13.0 | 80.7±16.9 | 27.8±5.1 | 8.19±0.58 | 0.70±0.24 IU/kg/day | ||
[0,1–10] | |||||||||
[0,1–10]Sotagliflozin | |||||||||
[1,0]Sands AT et al. (2015) | [1,0]4 | 400mg SOTA | 16 | 45.5 (21, 55)a | 16.8 (3.4, 42.9)a | 74.2 (55.6, 107.9)a | 27.1±3.1 | 7.94±0.55 | 0.60 IU/kg/day |
Placebo | 17 | 34.0 (21, 57)a | 18.5 (4.7, 40.8)a | 72.7 (55.3, 104.6)a | 26.2±3.0 | 7.98±0.51 | 0.60 IU/kg/day | ||
[1,0]Garg SK et al. (2017) | [1,0]24 | 400mg SOTA | 699 | 43.3±14.2 | 20.5±12.4 | 82.40±17.13 | 28.29±5.13 | 8.26±0.96 | 56,88±27.60 IU/day |
Placebo | 703 | 42.4±14.0 | 19.6±12.1 | 81.55±17.03 | 28.10±5.18 | 8.21±0.92 | 58,35±29.09 IU/day | ||
[1,0]Buse JB et al. (2018) | [1.0]52 | 400mg SOTA | 262 | 46.4±13.12 | 24±12.88 | 86.5±18.004 | 29.63±5.297 | 7.56±0.724 | 0.72±0.335 IU/kg/day |
Placebo | 268 | 45.2±12.72 | 24.2±12.38 | 87.3±17.709 | 29.55±5.188 | 7.54±0.712 | 0.74±0.357 IU/kg/day | ||
[1,0]Danne T et al. (2018) | [1.0]52 | 400mg SOTA | 263 | 41.7±13.23 | 18.9±11.18 | 81.97±17.963 | 27.85±4.921 | 7.71±0.819 | 0.74±0.267 IU/kg/day |
Placebo | 258 | 39.7±13.42 | 18.1±10.72 | 81.08±16.875 | 27.5±5.17 | 7.79±0.881 | 0.75±0.295 IU/kg/day | ||
[0,1–10] | |||||||||
[0,1–10]Metformin | |||||||||
[1,0]Jacobsen IB et al. (2009) | [1,0]24 | 2000mg Metformin | 12 | 43.5±13.1 | 17.8±10.3 | 87.6±13.1 | 29.5±2.7 | 8.37±0.20 | 56.8±2.9 IU/day |
Placebo | 12 | 37.3±9.6 | 20.3±10.2 | 92.0±10.2 | 29.2±2.8 | 9.17±1.13 | 70.6±30.9 IU/day | ||
[1,0]Khan AS et al. (2006) | [1,0]16 | 850∼2550mg Metformin | 15 | 48 [12] | 19 [10] | 91 [12] | 31.3±2.6 | 8.6±1.4 | NR |
Placebo | 15 | 48 [12] | 19 [10] | 91 [12] | 31.3±2.6 | 8.6±1.4 | NR | ||
[1,0]Lund SS et al. (2009) | [1.0]52 | 2000mg Metformin | 49 | 46.1±11.6 | 5–51 | 80.5±12.5 | 26.2±3.4 | 9.48±0.99 | 0.74±0.26 IU/kg/day |
Placebo | 51 | 44.9±10.8 | 6–56 | 79.0±15.3 | 25.8±4.3 | 9.60±0.86 | 0.75±0.22 IU/kg/day | ||
[1,0]Meyer L et al. (2006) | [1,0]24 | 1700mg Metformin | 31 | 39.9±12.9 | 16.9±8.9 | 80.5±12.5 | 26.4±4.6 | 7.58±0.84 | 0.72±0.21 IU/kg/day |
Placebo | 31 | 41.1±9.8 | 21.6±10.2 | 79.0±15.3 | 25.8±3.6 | 7.57±0.76 | 0.73±0.22 IU/kg/day | ||
[1,0]Pitocco D et al. (2013) | [1,0]24 | 850∼2550mg Metformin | 21 | 46±8 | 9.2±0.7 | 83 [12] | 28.7±2.1 | 7.24±0.90 | 0.61±0.22 IU/kg/day |
Placebo | 21 | 41 [10] | 8.8±0.8 | 77 [11] | 27.3±2.0 | 7.73±0.42 | 0.63±0.15 IU/kg/day | ||
[1,0]Petrie JR et al. (2017) | [1,0]144 | 2000mg Metformin | 219 | 55.2±8.5 | 33.4±11.0 | 83.9±15.4 | 28.4±4.5 | 8.1±0.9 | 0.63±0.26 IU/kg/day |
Placebo | 209 | 55.8±8.8 | 34.3±10.5 | 83.5±13.7 | 28.5±4.1 | 8.0±0.8 | 0.68±0.30 IU/kg/day | ||
[0,1–10] | |||||||||
[0,1–10]Empagliflozin vs. Metformin | |||||||||
[2,0]Lunder M et al. (2017) | [2,0]12 | 25mg EMPA | 10 | 46.0±2.3 | 22.5±3.7 | NR | NR | 7.8±0.1 | NR |
2000mg Metformin | 10 | 46.4±3.9 | 23.2±4.8 | NR | NR | 7.9±0.2 | NR | ||
Placebo | 10 | 43.1±2.1 | 22.2±3.8 | NR | NR | 7.8±0.2 | NR |
Abbreviations: CANA, canagliflozin; DAPA, dapagliflozin; EMPA, empagliflozin; HbA1c, hemoglobin A1c; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily insulin; n, number of patients in each group; NR, not reported; SOTA, sotagliflozin.
Data on the baseline characteristics are expressed as mean and standard deviation, except when otherwise indicated.
The meta-analysis by pairs of 12 trials showed that, when compared with placebo, SGLT2i resulted in significantly greater reductions in HbA1c (0.402 %) and with a high degree of heterogeneity (I2=70.6 %; p<.001, Fig. 2). The meta-analysis by pairs of 6 trials revealed that the differences in HbA1c levels between the metformin and placebo groups were not significant, with high homogeneity (I2=17.1 %; p=.306).
Paired forest diagram for the primary and secondary results. a) effect of SGLT2i on HbA1c compared with placebo; b) effect of SGLT2i on body weight compared with placebo; c) effect of SGLT2i on the total insulin dose compared with placebo; d) effect of metformin on HbA1c compared with placebo; e) effect of metformin on body weight compared with placebo; f) effect of metformin on total insulin dose compared with placebo.
Abbreviations: CI, confidence interval; HbA1c, glycated hemoglobin; SGLT2i, sodium-dependent glucose cotransporter 2 inhibitors; WMD, weighted mean difference.
Fig. 3 shows the network graph for the analysis of the safety and efficacy of metformin and SGLT2i. The results of the network meta-analysis indicated that SGLT2i were associated with higher reductions of HbA1c when compared with metformin (WMD, −0.32; 95 %CI −0.47 to −0.14) and placebo (WMD, −0.40; 95 %CI −0.47 to −0.32). However, there were no significant differences between metformin and placebo (WMD, −0.08; 95 %CI −0.24 to 0.06, Table 2). The results of the consistency test showed that there was no evidence of inconsistencies between the direct and indirect evidence for HbA1c (p=.51).
Comparison of the Safety and Efficacy Results for Metformin, Sodium-Dependent Glucose Cotransporter 2 Inhibitors and Placebo.
[0,1–3]A. HbA1c: WMD (95%CI), % | ||
[0,1–3]Metformin | ||
0.32 (0.14 to 0.47) | SGLT2i | |
−0.08 (−0.24 to 0.06) | −0.40 (−0.47 to −0.32) | Placebo |
[0,1–3] | ||
[0,1–3]B. Body weight: WMD (95%CI), kg | ||
[0,1–3]Metformin | ||
1.54 (0.09–2.93) | SGLT2i | |
−2.02 (−3.29 to −0.77) | −3.56 (−4.17 to −2.92) | Placebo |
[0,1–3] | ||
[0,1–3]C. Total insulin dose: WMD (95%CI), units | ||
[0,1–3]Metformin | ||
3.09 (−1.53 to 7.09) | SGLT2i | |
−3.88 (−7.46 to -0.59) | −6.94 (−9.55 to −4.32) | Placebo |
[0,1–3] | ||
[0,1–3]D. Severe hypoglycemia: OR (95%CI) | ||
[0,1–3]Metformin | ||
1.40 (0.63–2.90) | SGLT2i | |
1.25 (0.65–2.30) | 0.89 (0.60–1.34) | Placebo |
[0,1–3] | ||
[0,1–3]E. Diabetic ketoacidosis: OR (95%CI) | ||
[0,1–3]Metformin | ||
0.00 (0.00 to 0.01) | SGLT2i | |
0.00 (0.00 to 0.11) | 9.21 (2.69–79.82) | Placebo |
Abbreviations: CI, confidence interval; HbA1c, glycated hemoglobin; OR, odds ratio; SGLT2i: sodium-dependent glucose cotransporter 2 inhibitors; WMD, weighted mean difference.
Fig. 2 also shows the results of the meta-analysis by pairs for the secondary results. The results of the network meta-analysis revealed that both metformin (WMD, −3.88; 95 %CI −7.46 to −0.59) and SGLT2i (WMD, −6.94; 95 %CI −9.55 to −4.32) provided significantly higher reductions in the total insulin dose than placebo, with greater decreases for metformin (92 % probability of the treatment being the most effective), although there were no differences between metformin and SGLT2i (WMD, 3.09; 95 %CI −1.53 to 7.09). The results of the network meta-analysis indicated that SGLT2i is associated with greater decreases in body weight compared with metformin (WMD, −1.54; 95 %CI −2.93 to −0.09) and placebo (WMD, −3.56; 95 %CI −4.17 to −2.92). Body weight loss was lower with metformin than with placebo (WMD, −2.02; 95 %CI −3.29 to −0.77; Table 2).
Safety resultsFig. 4 shows the results of the meta-analysis by pairs for the safety results. The network meta-analysis showed no differences in terms of severe hypoglycemia between metformin, SGLT2i and placebo. DKA occurred more frequently with SGLT2i than with metformin and placebo.
Paired forest plots for severe hypoglycemia and DKA. a) Safety of SGLT2i in severe hypoglycemia; b) Safety of SGLT2i in DKA; c) Safety of metformin in severe hypoglycemia; d) Safety of metformin in DKA.
Abbreviations: CI, confidence interval; DKA, diabetic ketoacidosis; OR, odds ratio; SGLT2i, sodium-dependent glucose cotransporter 2 inhibitors.
Fig. 5 shows the network graph for the analysis of subgroups. Table 3 presents the comparative results for the safety and efficacy of the types of SGLT2i, metformin and placebo. The reduction in HbA1c levels obtained with dapagliflozin, empagliflozin and sotagliflozin was greater than that achieved with metformin and placebo. There were no differences between canagliflozin, metformin and placebo.
Comparison of the Safety and Efficacy of Canagliflozin, Dapagliflozin, Empagliflozin, Sotagliflozin, Metformin and Placebo.
[0,1–6]A. HbA1c: WMD (95%CI), % | |||||
[0,1–6]Canagliflozin | |||||
0.16 (-0.16 to 0.47) | Dapagliflozin | ||||
0.17 (-0.15 to 0.47) | 0.01 (−0.21 to 0.21) | Empagliflozin | |||
0.16 (-0.16 to 0.46) | −0.01 (−0.21 to 0.20) | −0.01 (−0.20 to 0.19) | Sotagliflozin | ||
-0.16 (-0.47 to 0.16) | −0.32 (−0.54 to −0.09) | −0.33 (−0.52 to −0.12) | −0.32 (−0.52 to −0.09) | Metformin | |
-0.25 (-0.53 to 0.03) | −0.41 (−0.56 to −0.25) | −0.42 (−0.55 to −0.27) | −0.41 (−0.54 to −0.26) | −0.09 (−0.25 to 0.06) | Placebo |
[0,1–6] | |||||
[0,1–6]B. Body weight: WMD (95%CI), kg | |||||
[0,1–6]Canagliflozin | |||||
-1.20 (-4.91 to 2.27) | Dapagliflozin | ||||
-0.96 (-3.94 to 1.86) | 0.20 (−2.77 to 3.18) | Empagliflozin | |||
-0.84 (-3.71 to 1.83) | 0.38 (−2.45 to 3.13) | 0.17 (−1.66 to 1.96) | Sotagliflozin | ||
-2.40 (-5.32 to 0.48) | −1.23 (−4.08 to 1.88) | −1.39 (−3.34 to 0.69) | −1.55 (−3.23 to 0.32) | Metformin | |
-4.35 (-6.95 to -1.87) | −3.17 (−5.74 to −0.54) | −3.37 (−4.78 to −1.90) | −3.54 (−4.55 to −2.38) | −1.97 (−3.43 to −0.58) | Placebo |
[0,1–6] | |||||
[0,1–6]C. Total insulin dose: WMD (95%CI), units | |||||
[0,1–6]Canagliflozin | |||||
0.48 (-9.19 to 9.37) | Dapagliflozin | ||||
-1.61 (-10.36 to 7.25) | −2.07 (−7.92 to 4.69) | Sotagliflozin | |||
-3.71 (-12.12 to 5.12) | −4.18 (−9.78 to 2.52) | −2.08 (−7.41 to 3.50) | Metformin | ||
-7.59 (-15.56 to 0.34) | −8.08 (−12.56 to −2.94) | −6.02 (−10.03 to −1.98) | −3.93 (−7.84 to −0.48) | Placebo | |
[0,1–6] | |||||
[0,1–6]D. Severe hypoglycemia: OR (95%CI) | |||||
[0,1–6]Canagliflozin | |||||
5.99 (0.85–49.16) | Dapagliflozin | ||||
6.30 (0.75–61.44) | 1.06 (0.27–4.39) | Empagliflozin | |||
6.94 (1.01–61.13) | 1.18 (0.45–3.45) | 1.10 (0.29–4.29) | Sotagliflozin | ||
4.02 (0.58–36.68) | 0.68 (0.25–2.13) | 0.65 (0.17–2.60) | 0.58 (0.22–1.57) | Metformin | |
5.19 (0.85–39.59) | 0.86 (0.41–1.99) | 0.83 (0.26–2.64) | 0.75 (0.37–1.46) | 1.27 (0.60–2.56) | Placebo |
[0,1–6] | |||||
[0,1–6]E. Diabetic ketoacidosis: OR (95%CI) | |||||
[0,1–6]Metformin | |||||
0.00 (0.00–0.13) | Placebo | ||||
0.00 (0.00–0.01) | 0.07 (0.00–0.67) | Sotagliflozin | |||
0.00 (0.00–0.08) | 0.37 (0.01–25.08) | 5.45 (0.07–1358.16) | Empagliflozin | ||
0.00 (0.00–0.03) | 0.18 (0.00–2.72) | 2.59 (0.03–168.02) | 0.48 (0.00–48.13) | Dapagliflozin | |
0.00 (0.00–0.00) | 0.00 (0.00–0.03) | 0.00 (0.00–0.67) | 0.00 (0.00–0.12) | 0.00 (0.00–0.30) | Canagliflozin |
Abbreviations: CI, confidence interval; HbA1c, glycated hemoglobin; OR, odds ratio; WMD, weighted mean difference.
In terms of body weight loss, the 4 SGLT2i were superior to placebo. However, there were no observed differences among the 4 SGLT2i and between SGLT2i and metformin. Once the efficacy results had been classified, canagliflozin, dapagliflozin and sotagliflozin were noteworthy in terms of body weight loss.
In terms of total insulin dose reductions, dapagliflozin, sotagliflozin and metformin were superior to placebo. However, canagliflozin, dapagliflozin and sotagliflozin were the 3 most effective therapies according to the probabilities range.
In terms of severe hypoglycemia, no differences were identified between the treatments. Canagliflozin and sotagliflozin had a higher probability of causing DKA than placebo. In addition, the incidence of DKA appeared to be higher with the 4 SGLT2i than with metformin, which was consistent with the results of the probabilities range.
DiscussionAlthough several RCTs have evaluated the safety and efficacy of SGLT2i and metformin in adults with DM1, to date only 1 RCT has been designed to compare the effects of empagliflozin and metformin on the HbA1c levels of patients with DM1.32 The present study has helped consolidate the evidence from numerous RCTs that compared SGLT2i or metformin with placebo, with the objective of achieving an indirect comparison of the overall safety and efficacy among SGLT2i and between these and metformin, using a network meta-analysis approach to this end.
Our network meta-analysis therefore demonstrated that SGLT2i presented a 92 % greater likelihood of being the most effective drugs in DM1 in terms of glycemic control (measured by HbA1c values), weight loss and reduction in total insulin dose, without increasing the risk of severe hypoglycemia. However, SGLT2i are associated with a greater risk of DKA. A meta-analysis has suggested that SGLT2i were more effective in glycemic control, in total insulin dose reduction and weight loss in DM1, and although the drugs did not induce severe hypoglycemia, they did increase the risk of DKA.11 Ketoacidosis that occurs during treatment with SGLT2i has been related to a series of biological mechanisms. One of the possible reasons for this situation is the difficulty in recognizing early symptoms, given that glucose concentrations remain very close to objective levels. In addition, the excessive reduction in insulin doses decreases its inhibitory effect on ketone production. Furthermore, there could be a direct relationship between treatment with SGLT2i and ketosis. Previous studies have shown that SGLT2i is expressed in pancreatic α-cells and that their inhibition directly triggers glucagon secretion.33
The differences, in terms of safety and efficacy, among SGLT2i in adults with DM1 have been analyzed in greater depth. The network meta-analysis thereby revealed that although canagliflozin provided reductions in body weight (with a 60 % likelihood of being the most effective treatment) not accompanied by hypoglycemia, the drug increased the rates of DKA. However, we need to consider that the previously mentioned results are based only on 1 RCT. High-quality clinical trials, with large samples and long-term follow-up, ensure the results of research studies on the safety and efficacy of canagliflozin in patients with DM1.
Sotagliflozin, a new dual inhibitor of SGLT1 and SGLT2, has been shown to reduce glucose absorption in the gastrointestinal tract through SGLT2 inhibition and reabsorption of renal glucose through SGLT2 inhibition.34,35 Our results showed that treatment with sotagliflozin translated into a significant reduction in HbA1c levels, body weight and insulin doses, although with a greater risk of DKA when compared with placebo.
Our network meta-analysis found similar reductions in HbA1c and body weight for dapagliflozin and empagliflozin, considered 2 of the best adjuvant drugs for DM1, according to the probabilities range. Furthermore, dapagliflozin showed a 46 % probability of being the best adjuvant treatment in terms of total insulin dose reduction without increasing the risk of severe hypoglycemia and DKA. In terms of the risk of severe hypoglycemia and DKA, empagliflozin showed similar results to dapagliflozin. The association between dapagliflozin and a lower risk of DKA might be explained by the DEPICT-1 study,19,20 in which participants underwent regular home monitoring through the use of meters for blood glucose and ketones and maintained direct communication with the physicians every 2 weeks to identify the risk factors for DKA and detect the first warning signs. The researchers established a simple and viable rule by which insulin doses were reduced by no more than 20 % after the use of the study drug, with adjustment subsequent to the initial dose. This rule was highly effective in reducing the risk of ketoacidosis.
In general, SGLT2i may be employed in adults with DM1, provided the patients undergo regular home check-ups and have a good understanding of how to detect the early signs of ketoacidosis. To date, there has been no study on the efficacy of SGLT2i in patients with DM1 with a follow-up longer than a year. Therefore, longer term trials are needed to clarify the safety and efficacy of SGLT2i as adjuvant therapy.
Although metformin is not more effective than SGLT2i, the reduction in body weight and insulin dose is significantly greater than that produced by the control group. The recent REMOVAL study with metformin in DM1 showed that body weight loss was maintained for at least 3years.36 Studies have shown that metformin can also reduce LDL cholesterol levels, an important cardiovascular risk factor, and that these reductions are maintained for at least 3years.36 In terms of safety, treatment with metformin does not increase the risk of severe hypoglycemia and DKA; furthermore, metformin showed a lower risk of DKA than the control group. Metformin would be a reliable complementary treatment to insulin for patients with DM1 who need a high total insulin dose, those who have a clear body weight gain and those who have high cardiovascular risk.
We need to recognize a number of limitations in this network meta-analysis. Firstly, only 13 trials with SGLT2i and 6 with metformin were ultimately included in the study, and there were even fewer RCTs employed in the subgroup analysis. Secondly, there were small differences in the populations of the included RCTs. So far, the studies have assumed that patients with DM1 are a homogeneous group of individuals with a similar response to adjuvant treatments, which is unlikely. Thirdly, we estimated high heterogeneity for HbA1c and total insulin doses for studies related to SGLT2i.
In conclusion, SGLT2i provide greater decreases in HbA1c levels and body weight when compared with metformin and placebo. SGLT2i and metformin produce greater reductions in insulin doses than placebo, but there were no significant differences between SGLT2i and metformin. However, the risk of inducing DKA was higher with SGLT2i than with metformin and placebo.
FundingThis report was supported by the hospital project of Taizhou People's Hospital (Project no. ZL201725).
Conflicts of interestThe authors declare that they have no conflicts of interest.
The authors would like to thank all participants of this study.
Please cite this article as: Zhang Q, Wu Y, Lu Y, Fei X. Eficacia y seguridad de la metformina y de los inhibidores del cotransportador-2 de sodio-glucosa en adultos con diabetes tipo 1: una revisión sistemática y metaanálisis en red. Rev Clin Esp. 2020;220:8–21.