Searches were conducted on databases including EMBASE, PubMed, Web of Science, Cochrane Central Register of Controlled Trials (CENTRAL), Wanfang Data, CNKI, VIP Database for Chinese Technical Periodicals, and the Chinese Biomedical Database. The search spanned until February 25, 2023, guided by predefined search strategies utilizing medical subject headings (MeSH) or their synonyms. We imposed no constraints on publication date, study design, or language. Results were imported into the Endnote bibliographic management tool. The specific search strategies employed in Embase are detailed in the Supplementary Table S2 .

The PICOS principle (19) shaped our selection process: (a) Population: Inclusion criteria entailed patients diagnosed with type 2 diabetes aged >18 years. (b) Intervention: RT that detailed at least one training variable (e.g., training intensity). (c) Comparator: Control groups that lacked physical intervention (e.g., health education, no intervention). (d) Outcome: Must report both pre- and post-intervention values for at least HbA1c or FBG. (e) Study Design: RCTs.

Exclusion criteria included: (a) Patients diagnosed with gestational diabetes. (b) Interventions not exclusive to RT (e.g., RT combined with aerobic exercise). (c) Previously published or duplicate literature. (d) Document types like “case reports”, “reviews”, “meta-analysis”, and “letters”.

Two researchers (SW, TK) independently screened titles and abstracts based on our selection criteria, further examining papers that initially met the criteria. Any discrepancies between their selections were addressed through discussion or, if necessary, mediation by a third researcher (ML). Extracted data covered details such as authors, country, sample size, gender, age, body mass index (BMI), disease duration, and RT training variables (e.g., period, intensity, sets, repetitions). RT groups were categorized by training intensity: high-intensity (≥70% 1RM), moderate-intensity (51%-69% 1RM), and low-intensity (≤50% 1RM) (20). After extraction, the two primary researchers cross-checked their findings. In instances where critical data was absent or unclear in the studies, we reached out to the authors for clarification.

Two independent reviewers utilized the Physical Therapy Evidence Database (PEDro) scale (21) to gauge the methodological quality of the studies. The PEDro scale, encompassing 11 criteria, integrates three from the Jadad scale (22) and nine from the Delphi list (23). RCTs are scored between 0-10 (from low to high quality) on the PEDro scale, with scores ≥6 indicating high-quality studies (21). The scale’s first item, specifying eligibility criteria, is geared towards establishing external validity and isn’t factored into the cumulative score. Previous evaluations by Maher et al. (21) confirmed an inter-rater reliability intraclass correlation coefficient of 0.68, based on consensus ratings from two or three independent raters.

To evaluate the overall effect of RT on HbA1c and FBG and establish a dose-response relationship in patients with type 2 diabetes, the standardized mean differences (SMD) was calculated using the formula: formula: SMD i = m 1 i − m 2 i S i (24). Here, SMDi represents the standardized difference between reported measures, where m1i and m2i represent the means of the intervention and control groups, respectively, while si is the pooled standard deviation. This formula was adjusted for sample size by Hedges and Olkin as: g =( 1- 3 4 N i − 9 ) (24), where N_i denotes the combined sample size of both groups. SMD signifies the difference between post-test averages of the treatment and control groups divided by the combined group’s standard deviation and is accompanied by 95% confidence intervals (CIs). Due to significant heterogeneity among studies, stemming from factors such as varying muscle groups, a random-effects model was utilized to estimate the effect of RT interventions (19). Our categorical variable meta-analysis was executed using Stata 16 software. Cohen categorized effect sizes as: 0.00 to ≤0.49 for small, 0.50 to ≤0.79 for moderate, and ≥0.80 for large effects (25). For clarity, a positive SMD was reported when RT exhibited superiority over the control group. The I² statistic assessed heterogeneity. Additionally, meta-analytic regression (metareg) investigated if the combined values of diverse training variables could predict RT’s influence on HbA1c and FBG. Subcategories were established to identify pivotal training variables, encompassing training duration, frequency, repetitions per session, repetitions per set, intensity, and rest intervals. Each subcategory was examined using a random-effects meta-regression model to discern potential enhancements in HbA1c and FBG measurements.

As presented in Table 2 , the studies exhibited an average quality score of 5.23 ± 1.31 points, with scores ranging from 3 to 8. Notably, six studies (29, 30, 40, 42, 43, 46) secured a score of 6 or above.

25 studies (15, 16, 26–43, 45–49) reported the effects of RT on HbA1c, yielding an SMD of -0.63 (95% CI: -0.97 to -0.29; I^2 = 87.0%, p<0.001). Relative to the control group, RT substantially decreased HbA1c levels in T2DM patients, as illustrated in Figure 2 . Of the 16 studies (15, 16, 28, 29, 34, 35, 37, 38, 40, 42–47, 49) that examined RT’s influence on FBG, the average SMD was -0.63 (95% CI: -1.07 to -0.20; I² = 88.5%, p<0.001). This is depicted in Figure 3 and highlights the significant effect.

Meta-regression results pinpointed the number of repetitions per set (p=0.034) as a predictor of RT’s impact on HbA1c (refer to Table 3 ). Regression analysis, however, revealed that no specific training parameters—including duration, intensity, frequency, number of sets, and number of repetitions—significantly affected FBG measurements (see Table 4 ).

Our data suggests a substantial benefit of RT on HbA1c levels in T2DM patients (-0.63 (95% CI -0.97- -0.29; I² = 87.0%, p=0.000)), aligning with previous studies (14, 17, 50). Effective HbA1c management can attenuate diabetes-related complications and mortality (51). Maintaining glucose levels at 7% or below can curtail long-term complications by up to 76% (49). A 1% decline in HbA1c can potentially decrease diabetes-related myocardial infarction risk by 14% and overall mortality by 21% (52). Our results are in line with Kelly, who reported an approximate 0.8% decrease in HbA1c through exercise, optimizing glycemic control (53). The therapeutic potential of RT may arise from augmented glucose transporter presence, increased muscle mass, and enhanced insulin receptor activity in muscle cells (54).

Furthermore, RT has been acknowledged for its efficacy in diminishing FBG in T2DM patients (-0.63 (95% CI -1.07- -0.20; I² = 88.5%, p=0.000)). This is consistent with Yaping et al., who posited that RT exerts a more potent influence on FBG enhancement (55). Brooks et al. (28), Castaneda et al. (29), and Yavari et al. (49) reported analogous findings. Mechanisms underlying these improvements include increased muscle strength, augmented GLUT-4 receptor activity, rapid glucose transportation, and refined insulin resistance and sensitivity (56). Rezaeeshirazi et al. (44) also reported increased insulin sensitivity in elderly T2DM patients’ post-resistance exercise, leading to substantial FBG reduction.

Numerous studies have established that various training variables, such as duration, frequency, and intensity, within sports training can optimize the hypoglycemic effect (14, 57). Consequently, discerning which variable correlates with a more potent positive impact of exercise on T2DM patients becomes crucial. Our analysis, encompassing 25 studies (15, 16, 26–43, 45–49), revealed that the duration of RT had no significant association with HbA1c reduction. Training durations ranging from 12-16 weeks seem to be the most effective in decreasing HbA1c. This may be attributed to the notion that a noticeable exercise effect necessitates a more extended training duration. For instance, Dunstan et al. (35) observed no marked improvement in HbA1c following 8 weeks of moderate-intensity RT. The lack of significant change in HbA1c could be due to brief training durations, specifically, 4-6 weeks (58). Conversely, even with prolonged RT cycles, some patients might not achieve the desired results, possibly due to inadequate exercise intensity or compliance. Church et al. (33) found no significant amelioration in HbA1c after 36 weeks of moderate-intensity RT, and Yamamoto et al. (48) reported 27 participants dropping out of their study. Hence, intervention duration alone doesn’t dictate HbA1c reduction but is influenced by both training intensity and adherence.

Our meta-analysis, which incorporated 20 studies (15, 16, 26–29, 33–40, 42, 43, 45–47, 49), recorded a 0.89% and 0.29% decrease in HbA1c with high and low intensity RT, respectively. This aligns with another meta-analysis (encompassing 24 trials) which suggested that high-intensity RT is superior in reducing HbA1c in T2DM patients as compared to low-to-moderate intensity RT (48). Contrarily, a separate Australian study observed that 3 months of moderate-intensity RT led to a significant decline in HbA1c levels (by 5.92%) in T2DM patients (43). In our research, it was evident that when the mean baseline HbA1c exceeded 7.5%, participants displayed greater HbA1c improvement post-RT, a finding corroborated by studies from Segal et al. (Canada) (46) and Kadoglou et al. (Greece) (59). However, the intensity of RT didn’t exhibit a direct correlation with its effect on HbA1c reduction in diabetic patients.

Moreover, our analysis indicated that conducting 2-3 sets per week and 3 sets per exercise session were most effective in HbA1c reduction. Conventionally, the advised frequency for RT is two to three times weekly, although it’s recommended to be integrated with AT (60). Excessive training frequency could lead to fatigue; hence, the number of sets might not be the primary determinant for HbA1c reduction in T2DM patients. Meta-regression analysis underscored the pivotal role of repetition count in HbA1c decline. T2DM patients witnessed the most significant drop in HbA1c with 6-10 repetitions per set (mean SMD= 1.31). While the aggregate repetitions in an RT set might induce a substantial physiological stimulus for enhanced strength and subsequent HbA1c reduction, fewer repetitions might be more potent (17). This is because repetition count is often equated with training intensity. Historically, emphasis was on “training intensity” rather than “number of repetitions”, leading to the inference that fewer repetitions correspond to augmented training intensity (61). Lastly, no discernible correlation existed between RT rest periods and HbA1c reduction. Rest intervals under 60 seconds appeared most effective in diminishing HbA1c in our study. A potential rationale is that shorter rest intervals induce more fatigue, thus offering a stimulus augmenting muscular strength and subsequently lowering HbA1c (62).

Sixteen studies (15, 16, 28, 29, 34, 35, 37, 38, 40, 42–47, 49) were incorporated into this analysis. Meta-regression identified no significant association between training variables and FBG. Nonetheless, the optimal RT program for enhancing FBG in T2DM patients is characterized by a training intensity of 70-80% of 1RM, conducted thrice weekly, encompassing 3 sets during each session, with 8-10 repetitions, and inter-set rest intervals of less than 60 seconds.

Evidence from randomized controlled trials indicates distinct glycemic responses in T2DM patients undergoing RT. Notably, a 26-week high-intensity RT (70% 1RM) regimen yielded minimal alterations in blood glucose levels (16). Conversely, a 12-week moderate-intensity RT regimen precipitated notable reductions in fasting blood glucose (15). However, a separate 12-week moderate-intensity RT program (50-60% 1RM) failed to elicit significant glucose reductions (43). A recent meta-analysis posited that older T2DM patients should prioritize RT intensity over duration and frequency to ensure optimal glycemic management (50). Collectively, the impact of RT on glycemic fluctuations appears to be heterogenous, and the standalone effect of RT intensity on glucose levels in diabetics remains nuanced.

Subgroup analysis from our research reveals that a training intensity of 70-80% of 1RM, executed thrice weekly in 3 sets, with 8-10 repetitions per set, and rest periods of less than 60 seconds between sets, is most conducive for FBG reduction. These observations align with prior studies (28, 29, 47). Ferriolli et al. (63) suggested an RT protocol for elderly diabetic patients, advocating for sessions on at least 2 days per week at a moderate-to-high intensity of 1RM. Similarly, Liu et al.’s meta-analysis (14) endorsed an RT regimen with an average high intensity (75-100% of 1RM), performed in 2-3 sets per week and 8-10 repetitions per set, as efficacious in reducing FBG among T2DM patients. Furthermore, in scenarios devoid of contraindications, Pan et al. (64) recommended diabetic patients to train thrice weekly, targeting all major muscle groups over 3 sets of 8-10 repetitions, but never exceeding 10 repetitions. These recommendations resonate with the insights of our meta-analysis. Despite these insights, it remains imperative to further probe the efficacy of resistance exercise interventions on FBG.

A notable limitation of our study is the inability to conduct a thorough analysis of the interrelationships among the suggested training variables. Our research draws from several studies that employed diverse training parameter combinations, such as intensity, training frequency, and number of sets. Given the variability in these parameters, it remains inconclusive whether the outcomes would remain optimal if each element of the RT method were tailored based on the prevailing dose-response relationship. Hence, further investigations are imperative to devise an analytical framework that provides insights into the interplay between these training parameters. One approach to overcome this limitation is through the systematic modeling of training variables, wherein one RT variable remains constant while another is varied. While our study evaluated the influence of radiation on HbA1c and FBG, the assessment did not extend to hypertrophy (specifically, alterations in lean tissue) due to the paucity of studies documenting such changes. Consequently, the direct link between exercise and augmented muscle mass remains speculative. We advocate for future studies examining the relationship between hypertrophy and HbA1c.

Moreover, the majority of the studies reviewed did not provide details on medication adjustments throughout the intervention. If the control group received a more potent medication regimen compared to the intervention group, the impact of RT on glycemic control could have been underestimated. This was particularly evident in studies with combined medication interventions. Another challenge encountered in our literature review and meta-analysis was the pronounced bias inclination of the included studies (only 6 out of 26 studies attained a PEDro score of 6), the marked heterogeneity across the studies (i.e., I^2 values ranging from 59.5% to 88.5%), and the uneven SMD reported for specific training parameters.