The protocol of this study was registered in PROSPERO (CRD 42022289985) and conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.7

The literature search was conducted in the electronic databases CINAHL, Cochrane Library, EMBASE, Google Scholar, LILACS, MEDLINE/Ovid, SciELO, SPORTDiscus, and Web of Science, from the inception of the databases until May 8, 2024. The search included studies with human subjects, without language restrictions. The search strategy combined DeCS/MeSH descriptors and their synonyms "Diabetes Mellitus, Type 2″ AND "Exercise" AND "Glucose," as shown in the Supplementary material – Table S.1.

The search was structured according to the PICOS criteria: 1) Population: men and/or women diagnosed with T2D (≥ 18 years of age); 2) Intervention: a single session of AE and/or RE; with AE involving large muscle groups in rhythmic or dynamic movements, characterized as continuous (CAE) or interval (IAE).8 In comparison, RE involves the use of muscular strength to move a weight, is a brief activity, and engages isolated muscle groups, including isometric, concentric, or eccentric contractions against a body segment load or an external load.8 3) Control: no exercise or any type of exercise that did not meet the characteristics of the intervention and that was performed by individuals with T2D; 4) Outcome: blood glucose levels in response to a single session of AE and/or RE, with glycemia measurements taken before and within a period of up to 24 h after the session; 5) Study type: Randomized controlled trials, including parallel or crossover designs with a washout period exceeding 72 h, and non-randomized clinical trials.

The exclusion of duplicate articles and the evaluation of titles and abstracts of the articles retrieved from the databases were performed using the Ryyan9 tool, and the eligible articles were stored using the Mendeley Desktop software.

Two independent reviewers (JAA, CVO) read the studies in their entirety and systematically extracted data from the included articles. Disagreements were resolved by a third reviewer (LPS) and the data were analyzed using qualitative and quantitative syntheses, where possible meta-analysis.

The data were extracted using a standardized spreadsheet developed by the authors using Microsoft Excel software, and any missing data were requested by email from the authors of the studies. The following data were extracted: 1) characteristics of the study population (age, sex, body mass index (BMI), time elapsed since diagnosis, medications, glycated hemoglobin, diabetes mellitus (DM) complications, comorbidities, smoking, and level of physical activity); 2) aspects of the intervention; and 3) outcome of interest (pre- and post-exercise glucose values). Additionally, information regarding any changes in diet and medication use before the session was also extracted.

The studies' risk of bias was assessed independently by two authors (JAA, APDBB) and they were classified as "low", "uncertain", or "high" risk.10 The Cochrane "Risk of Bias" tool (RoB 2) was used to assess the risk of bias in randomised clinical trials with a parallel design and the "Revised Cochrane Risk of Bias” tool for randomized trials (RoB 2) - Additional considerations for crossover trials" was used to assess the risk of bias of randomized and non-randomized crossover design studies.10 In the present study, funnel plot asymmetry was not performed to assess publication bias in the meta-analyses because the meta-analyses presented included fewer than 10 studies.11

The certainty of the evidence and the strength of the recommendations found in the meta-analysis for the investigated outcome were assessed according to GRADE and was categorised as "high", "moderate", "low", or "very low".12

For studies that reported glucose values in mg/dL, the values were converted to mmol/L, considering that 1 mmol/L is equivalent to 18.02 mg/dL.13 Studies with different glucose measurement interventions and/or post-exercise session time points, as well as non-randomized clinical trials (quasi-experimental) and those where data extraction was not possible, had their results qualitatively analyzed. Data were combined for meta-analysis using a minimum of two randomized clinical trials assessed as clinically homogeneous, considering the type of exercise performed and the timing of post-intervention outcome (glucose) measurement. Statistical analyses were conducted following the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions.11 When applicable, dispersion data were converted from standard error to standard deviation.

The meta-analysis was conducted using Stata 17.0 software (STATA Corp., College Station, TX, USA), considering the DerSimonian and Laird random-effects model. For each study, the mean, standard error or standard deviation, and sample size were extracted. The mean differences and 95 % confidence intervals (CI) were calculated. Effect measures were derived from post-exercise glucose values, and studies were analyzed separately based on the type of exercise performed. A significance level of 5 % was established.

Statistical heterogeneity between studies was quantified using Cochran's Q test and inconsistency (I²),14 and the estimates of the variation in heterogeneity (τ²) to assess the variance between studies, it is a measure of clinical relevance.15 The statistical significance was defined as a P < 0.05 in Cochran's Q test, as for the inconsistency test, I² >75 % was considered to indicate high heterogeneity, I² between 25 % and 75 % to indicate moderate heterogeneity, and I² <25 % to indicate low heterogeneity.14 The results were represented in forest plots.

The literature search identified 7210 studies. After removing duplicates, 5793 were screened by title and abstract, of which 77 articles were assessed for eligibility. After searching for and reading the full texts, 51 studies were not eligible for the following reasons: two of them included participants without T2D (participants); seven did not assess the acute effect of exercise, five involved interventions that were not AE or RE, and 12 had a washout period shorter than 72 h (intervention); nine did not include a control group (control); two were excluded due to translation issues (unable to translate); two were literature reviews (study design); and 12 were conference abstracts.

Twenty-six studies were included in this review, with 13 of them (52 %) being grouped through meta-analysis techniques (Fig. 1).

Fig. 1.

Flowchart, PRISMA 2020.

(0.52MB).

Of the 26 studies included in this review, 13 investigated CAE vs control,16-28 six investigated IAE vs control,29-34 three investigated RE vs control,35-37 three investigated CAE vs IAE vs control,38-40 and one study41 investigated CAE vs IAE vs AE followed by RE vs RE followed by AE vs control. For the studies by Bellini et al.41 and Marcotte-Chénard et al.,34 we considered the comparison of each exercise session with the control session and not the comparison between exercise sessions.

The studies included were published between 1997 and 2024, and most of them were conducted in the USA (24 %), followed by Canada (15 %), China (11 %), Denmark (11 %), the Netherlands (11 %), Brazil (8 %), Italy (8 %), Iran (4 %), Portugal (4 %), and the United Kingdom (4 %). Three studies reported no adverse effects from the exercise session investigated.28,32,40 In the other 23 studies16-27,29-31,33-39,41 the occurrence or non-occurrence of adverse effects of the exercise session investigated was not reported. Concerning adherence to the exercise session investigated and the control session, in all the studies the experimental protocol was completed in full by the participants and the statistical analyses were performed per protocol.

The total sample was made up of 438 participants aged between 21 and 73 years, with the mean BMI ranging between 22.2 ± 2.3 and 37.0 ± 5.7 kg/m2; with the majority being male (290 men, 133 women, and 15 not reported). When reported, the average time elapsed since T2D diagnosis ranged from 1.8 ± 1.0 to 11.6 ± 1.9 years, and the mean value of glycated hemoglobin (A1c) varied between 6.0 ± 0.3 % and 10.4 ± 3.0 %. Regarding diabetes complications, 15 studies reported that participants did not have complications,16,18-20,22-26,28,29,33,34,38,41 and this information was not reported in 11 studies.17,21,27,30-32,35-37,39,40 Seven studies reported no diabetes comorbidities,16,23,24,33,35,38,41 and this information was not available in 18 studies.16,18-22,25-27,29-32,34,36,37,39,40 The sociodemographic and clinical characteristics of participants from each study are described in Table 1.

Exercise sessions lasted from 1031 to 60 min.21,23,27,32 Most of the studies that evaluated CAE performed a single exercise session at moderate intensity,16-19,21-24,26-29,38-40 only one study performed the exercise session at low intensity25 and one study conducted the exercise session with high intensity.17 Among the studies that evaluated a session of IAE, only one study conducted a single session with moderate intensity,32 while the remaining studies performed the exercise at high intensity.30-34,38-40 Of the four studies that assessed a single session of RE, three of them conducted the exercise session with moderate intensity,36,37,41 and only one study conducted a session with both low and high intensities.35 Regarding the time of day, the majority of interventions were conducted in the morning (68 %).16,17,21-24,27-30,32-35,38-41

Glucose was measured from interstitial fluid in 13 studies,18,21,22,23,25-27,29-31,34,35,40 from blood collected through a venous catheter in eight studies,16,19,24,32,33,36-38 and from capillary blood measured using a glucometer in five studies.17,20,28,39,41 Regarding pre-exercise diet, 92 % of the included studies17-31,33-41 standardized the diet ingested on intervention days, with one study not reporting whether there was standardization of the diet,16 and one study did not standardize the diet ingested by participants pre-exercise.32 The characteristics of the investigated exercise sessions are described in the Supplementary material (tables S.2.1, S.2.2, and S.2.3).

Of the 13 studies that assessed the effect of a session of CAE on blood glucose,16-28 ten of them reported a significant reduction in post-CAE glucose,16,17,19-26 with no difference in three studies.18,27,28 Of the six studies that investigated a session of IAE vs. control session,29-34 two of them found a significant reduction in post-IAE glucose,29,31 with no difference in four studies.30,32,33,34 Of the three studies that investigated a session of CAE vs. IAE vs. control session,38-40 two of them showed that both CAE and IAE sessions significantly reduced post-exercise glucose,38,39 and in one study there was no significant change in glucose in response to any of the investigated sessions.40 The three studies that investigated RE vs. control sessions did not find a significant reduction in post-exercise glucose.35-37 In the study that investigated a single session of CAE vs. RE vs. CAE followed by RE vs. RE followed by CAE vs. control session,41 the CAE and RE sessions, both individually and combined, significantly reduced post-exercise glucose compared to the control session.

The reduction in blood glucose persisted for up to 24 h post-CAE, up to 30 min post-IAE, up to 60 min post-RE, and up to 45 min after combined CAE and RE. The mean and standard deviation (SD) values and glucose values from studies that were not included in the meta-analysis are presented in Table S.3, and the reasons for non-inclusion are detailed in Table S.4 of the Supplementary material.

The risk of bias was assessed as low in one study,26 uncertain in 22 studies,16-25,27,28,30,33-41 and high in three studies.29,31,32 Although 84 % of the included studies in this systematic review were RCTs,16-28,33,35-41 the randomization method and allocation sequence concealment were not always clearly reported. Only five studies reported the methods used for randomization order,16,26,28,38,40 and only one study reported allocation sequence concealment.26

Considering the nature of the investigated intervention (exercise), it was not possible to blind participants and researchers. On the other hand, the evaluated outcome (glucose) is not influenced by the lack of blinding to the intervention. The results obtained from the Cochrane Risk of Bias tool are illustrated in Fig. 2.

Fig. 2.

Cochrane Risk of Bias Tool. D1: bias due to the randomization process; DS: bias from period and carry-over effects; D2: bias due to deviations from intended interventions; D3: bias due to missing outcome data; D4: bias due to outcome measurement; and D5: bias due to selective outcome reporting.

(0.69MB).

Fig. 3 (Panel A) presents the results of the comparison between a CAE session and a control session, showing a statistically significant reduction in glucose of 1.48 mmol/L (26.7 mg/dL) in the first-minute post-CAE. Ten minutes post-session of CAE there was a significant reduction in glucose of 1.61 mmol/L (29.01 mg/dL) (Fig. 3B). There was a decrease in glucose by 0.57 mmol/L (10.27 mg/dL) at the 30-minute mark post-CAE and a reduction of 1.15 mmol/L (20.72 mg/dL) in glucose 60 min after the CAE session (Fig. 3C; Fig. 3D), although the difference was not statistically significant.

Fig. 3.

Forest plots comparing post-session glucose levels. (A) CAE vs. control session at the first minute. (B) CAE vs. control session at 10 min. (C) CAE vs. control session at 30 min. (D) CAE vs. control session at 60 min. (E) IAE vs. control session at the first minute. (F) IAE vs. control session at 30 min. (G) RE vs. control session at the first minute. (H) RE vs. control session at 30 min. (I) RE vs. control session at 60 min. (J) RE vs. control session at 90 min. (L) RE vs. control session at 120 min.

When comparing IAE vs. control session, a statistically significant difference was observed at both analyzed moments, with a decrease in glucose values of 2.66 mmol/L (47.92 mg/dL) in the first-minute post-IAE (Fig. 3E) and a reduction of 1.11 mmol/L (20.00 mg/dL) 30 min post-IAE (Fig. 3F).]

When comparing RE vs. control session, there was a significant reduction in glucose of 1.18 mmol/L (21.26 mg/dL) in the first-minute post-RE (Fig. 3G). Glucose levels were also evaluated at 30 min post-session (−0.27 mmol/L) (Fig. 3H), 60 min (0.37 mmol/L) (Fig. 3I), 90 min (0.86 mmol/L) (Fig. 3J), and 120 min post-RE (0.57 mmol/L) (Fig. 3L); however, the difference was not statistically significant, and from 60 min post-session, glucose tends to ascent.

At all analyzed time points (CAE vs. control session, IAE vs. control session, and RE vs. control session), the included studies showed low heterogeneity (I²: 0 %).

According to GRADE,12 the comparisons of CAE vs. control, IAE vs. control, and RE vs. control have moderate certainty of evidence (Table 2).