The effect of exercise intervention on IGF axis and metabolic markers regulation in breast cancer patients: a systematic review with pairwise, network, and dose-response meta-analyses Authors: Bing-Ao Chen, Wen-Hui Qin, Hong-Bao Chen, Qi-Hang Yang, Yang Yuan Published: 2025 ABSTRACT AbstractBackgroundBreast cancer is one of the most common malignancies among women today. Insulin resistance is closely associated with the occurrence and progression of breast cancer. The interplay between the insulin-like growth factor (IGF) axis and metabolic processes plays a critical role in the progression of breast cancer. Moderate exercise interventions can improve the overall health of breast cancer patients by regulating the IGF axis and associated metabolic markers.ObjectiveThis study aims to evaluate the effects of various types and doses of exercise on the IGF axis and metabolic in breast cancer patients through pairwise meta-analysis, network meta-analysis (NMA), and dose-response NMA, and to identify the optimal exercise modalities and dosage range.MethodsA comprehensive search identified randomized controlled trials (RCTs) for inclusion. Exercise interventions were categorized into continuous aerobic exercise (CAE), combined aerobic and resistance exercise (CE), and resistance exercise (RE). Statistical analyses were conducted on insulin, IGF axis, insulin resistance, and fasting glucose. Dose-response NMA was used to assess the relationship between exercise dose and IGF-1, insulin, and insulin resistance levels.ResultsA systematic search of PubMed, Web of Science, and other databases included 26 RCTs with a total of 1803 participants. Exercise significantly reduced IGF-1, insulin, BMI, HOMA-IR, IGFBP-1, triglycerides, and low-density lipoprotein (LDL), while increasing IGFBP and high-density lipoprotein (HDL). Network meta-analysis revealed that CAE was most effective in inhibiting IGF-1 (SUCRA = 81.42), while CE showed the best improvement for insulin (SUCRA = 89.87) and HOMA-IR (SUCRA = 85.77). Dose-response analysis indicated that the recommended exercise dose for breast cancer patients ranged from 560 to 1080 METs·min/week. Specifically, for IGF-1 suppression, CAE with 560–1080 METs·min/week was recommended; for insulin, CE with 460–1400 METs·min/week was optimal; and for HOMA-IR reduction, CE with 460–930 METs·min/week was most effective.ConclusionExercise interventions effectively regulate the IGF axis and metabolic markers in breast cancer patients. Continuous aerobic exercise is most effective in reducing IGF-1, while combined aerobic and resistance exercise is more advantageous for improving insulin levels and insulin resistance.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12885-025-15475-0.Keywords: Exercise, Breast cancer, IGF-1, Fasting insulin, Insulin resistance Introduction Breast cancer is the most common type of cancer among women worldwide and is the second leading cause of cancer-related deaths in women [1]. In recent years, with lifestyle changes and the aging population, the incidence of breast cancer has been rising annually. Among the various factors influencing the onset and prognosis of breast cancer, metabolic abnormalities, particularly insulin-related metabolic disorders such as obesity and type 2 diabetes, have attracted increasing attention [2–4]. Multiple studies have shown that women with diabetes (primarily type 2 diabetes) have a 20%-27% increased risk of developing breast cancer, and this risk is not limited to specific tumor subtypes [5]. An increasing body of research indicates a close relationship between insulin resistance (IR) and cancer. In IR patients, hyperinsulinemia can activate the insulin-insulin-like growth factor (IGF) signaling pathway, which further activates cancer-related pathways such as mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PIK3CA), promoting cell proliferation and survival, thus supporting tumor growth [6–9]. In a study on hormone receptor–positive (HR+) / human epidermal growth factor receptor 2-negative (HER2-) early breast cancer, researchers found that although relapse scores (RS) were not significantly associated with body mass index (BMI), insulin and its related metabolic markers (such as C-peptide, GLP-1, and glucagon) were positively correlated with tumor size [7, 10, 11]. Exercise not only improves weight control in breast cancer patients but also significantly affects the function of the insulin-like growth factor system [12–15]. Studies have shown that high levels of physical activity are associated with lower insulin and IGF-1 levels and higher IGFBP-3 levels, and these changes help inhibit the IGF signaling pathway, thereby reducing cancer cell proliferation and metastasis [14, 16, 17]. The IGF axis and metabolic markers are closely interconnected in the development and progression of breast cancer [18]. Metabolic abnormalities such as obesity, insulin resistance, and hyperinsulinemia can enhance IGF signaling activity, promoting tumor cell proliferation and metastasis, suggesting that the IGF axis is a key bridge linking metabolic disorders with tumor progression [19–21]. This study employs pairwise meta-analysis, network meta-analysis, and dose-response analysis to systematically identify the optimal exercise regimens that regulate the insulin-like growth factor (IGF) axis-including IGF-1, insulin-like growth factor binding protein-1 (IGFBP-1), and insulin-like growth factor binding protein-3 (IGFBP-3)-as well as related metabolic markers such as insulin, homeostatic model assessment of insulin resistance (HOMA-IR), BMI, fasting glucose, triglycerides, low-density lipoprotein (LDL), and high-density lipoprotein (HDL), thereby achieving optimal health outcomes for breast cancer patients [22]. Materials and methods Protocol and registration This meta-analysis was conducted following the guidelines of the PRISMA-NMA extension for network meta-analysis and Cochrane Handbook for Systematic Reviews of Interventions. The study protocol was registered in PROSPERO with the registration number 1,030,309. As the data used in this study were from previously published studies, ethical approval was not required. Search strategy and study selection A comprehensive literature search was conducted using the PubMed, Web of Science, Embase, and Cochrane databases to identify studies evaluating the effects of exercise interventions on insulin, the IGF axis, and metabolic markers in patients with breast cancer. The search covered all publications from database inception to April 2025, without language or regional restrictions. The search strategy primarily utilized Medical Subject Headings (MeSH) and free-text terms, including “Exercise,” “Physical Activity,” “Breast Cancer,” “Insulin,” “IGF-1,” and “HOMA-IR.” Detailed search strategies are provided in the Supplementary Materials. Two reviewers (BA-C and WH-Q) independently screened the titles and abstracts for eligibility and assessed full texts according to predefined inclusion and exclusion criteria. Any disagreements were resolved through consultation with a third reviewer (YY), who has expertise in this field. All references were managed and screened using EndNote X9 (Thomson ISI Research Soft, Philadelphia, PA, USA). Inclusion and exclusion criteria The inclusion criteria were as follows: Female patients diagnosed with breast cancer (≥18 years old). Female patients diagnosed with breast cancer (≥18 years old). Exercise interventions were defined as planned, purposeful, and repeatable physical activities, classified into three types with specific definitions provided in Table 1. Additionally, to account for the sustained physiological changes induced by exercise, interventions must last at least 4 weeks. Exercise interventions were defined as planned, purposeful, and repeatable physical activities, classified into three types with specific definitions provided in Table 1. Additionally, to account for the sustained physiological changes induced by exercise, interventions must last at least 4 weeks. The control group must have received routine care without specific interventions. The control group must have received routine care without specific interventions. Studies must report at least one of the following outcomes: Insulin, IGF-1, and HOMA-IR, and include metabolic-related markers such as BMI, fasting glucose, triglycerides, IGFBP-1, IGFBP-3, low-density lipoprotein (LDL), and high-density lipoprotein (HDL). The units for all biomarkers were standardized according to the most commonly used units reported in the included studies. Studies must report at least one of the following outcomes: Insulin, IGF-1, and HOMA-IR, and include metabolic-related markers such as BMI, fasting glucose, triglycerides, IGFBP-1, IGFBP-3, low-density lipoprotein (LDL), and high-density lipoprotein (HDL). The units for all biomarkers were standardized according to the most commonly used units reported in the included studies. The study design must be a randomized controlled trial (RCT). The study design must be a randomized controlled trial (RCT). Table 1Definitions of different exercise interventionsExerciseOperational definitionContinuous Aerobic ExerciseSteady-state, rhythmic, large-muscle aerobic activity performed continuously within a sessionResistance ExerciseMulti-joint or major muscle-group resistance training using machines, free weights, or bands; structured sets/reps.Combined Aerobic and Resistance ExerciseProgramme including both aerobic (steady-state) and resistance components within the weekly plan. Definitions of different exercise interventions The exclusion criteria were as follows: Interventions lasting less than 4 weeks. Interventions lasting less than 4 weeks. Studies with methodological weaknesses, incomplete or unclear data reporting, or results that do not provide sufficient scientific evidence based on established research standards. Studies with methodological weaknesses, incomplete or unclear data reporting, or results that do not provide sufficient scientific evidence based on established research standards. Cross-over or complex experimental designs that did not allow for separate extraction of the effects of exercise interventions. Cross-over or complex experimental designs that did not allow for separate extraction of the effects of exercise interventions. Conference abstracts, experimental protocol descriptions, or systematic reviews. Conference abstracts, experimental protocol descriptions, or systematic reviews. Data extraction Two researchers (BA-C and WH-Q) independently extracted relevant information from all included studies. In case of disagreements, a third researcher (YY) was consulted to resolve the issue. Extracted data included: the first author’s name, publication year, basic intervention information (type, duration, frequency, and intensity), and the main outcomes (Insulin, IGF-1, and HOMA-IR) along with secondary outcomes (BMI, fasting glucose, triglycerides, IGFBP-1, IGFBP-3, LDL, and HDL). Incomplete or unclear data were clarified by contacting the corresponding author. Risk of bias assessment The risk of bias for the included studies was assessed using the RevMan 5.4 tool by two researchers (BA-C and WH-Q). Any disagreements were resolved by consulting a third reviewer (YY). Since the exercise interventions could not meet blinding conditions for both experimenters and participants, blinding was not assessed for potential bias. Measurement of treatment effect The mean difference (MD) and its corresponding standard deviation (SD) were used to assess changes before and after treatment. If SD was not directly reported, it was estimated based on standard error (SE), 95% confidence intervals (CIs), P-values, or t-statistics. For calculating the SD of the difference before and after the intervention, a correlation coefficient of 0.5 was assumed, which is widely accepted in the literature for medium-level measurement reliability, ensuring a balance between potential associations between pre- and post-measurements and robustness of estimation. Statistical analysis Pairwise meta-analysis and publication bias For the primary outcomes (IGF-1, Insulin, and HOMA-IR) and secondary outcomes (BMI, fasting glucose, triglycerides, IGFBP-1, IGFBP-3, LDL, and HDL), all exercise interventions were analyzed as a single exercise group. Subgroup analyses were performed when there were more than three exercise interventions for a primary outcome. Random-effects models were used to perform pairwise meta-analysis, combining mean differences (MD) and calculating corresponding 95% credible intervals (Crl). I² statistics were used to assess heterogeneity, with I² values exceeding 50% indicating high heterogeneity. Publication bias was assessed using funnel plots, Egger’s regression test, and Begg’s rank test. Network meta-analysis Network meta-analysis was conducted using MetaInsight v6.0.1 and R software (version 4.4.1), generating network diagrams and intervention ranking plots. Frequency statistics were performed using MetaInsight, and Bayesian statistical analysis was conducted using the gemtc package. Initial analysis was conducted in MetaInsight, while random-effects model analysis and visualization were performed using the network package in R combined with Bayesian Markov Chain Monte Carlo (MCMC) methods. Data were standardized prior to analysis. MCMC simulations were conducted with four chains running in parallel, with an initial value of 2.5, step size of 10, and 25,000 iterations, discarding the first 5,000 iterations to reduce bias from the initial values. Model fit was assessed using the Deviance Information Criterion (DIC), and model convergence was evaluated using Gelman-Rubin diagnostic indices. Cumulative ranking curves under the surface area under the curve (SUCRA) were used to rank interventions. Local inconsistency was evaluated using the node-splitting method and comparing the DIC values between consistency and inconsistency models. A difference of less than 5 was considered acceptable for model fit. Bayesian ranking plots were generated for Insulin, IGF-1, and HOMA-IR outcomes based on SUCRA values. Dose-response network meta-analysis A Bayesian model-driven network meta-analysis (MBNMA) was used to evaluate the dose-response effects of different exercise interventions on Insulin, IGF-1, and HOMA-IR. To ensure outcome heterogeneity, MD was used to represent effect size, and 95%Crl was used to assess credibility. The MBNMAdose framework was employed to combine dose-response modeling with Bayesian network meta-analysis, simulating multiple dose-response functions and integrating evidence networks with heterogeneity to improve precision in estimating intervention effects. The MBNMAdose method, based on random-effects models, estimated all parameters using Bayesian methods and Markov Chain Monte Carlo (MCMC) algorithms. Dose-response modeling incorporated commonly used functions such as the Emax model. Model fitting and diagnostics were performed using the MBNMAdose package in R, supplemented with rjags, rstan, and coda packages to process multi-source data, control heterogeneity, and estimate dose-dependent effects. Finally, the dose-response curves were visualized using the ggplot2 package. Results Study selection and characteristics A total of 8,530 articles were identified through electronic database searches. After removing duplicates and conducting full-text screening and evaluation, 26 randomized controlled trials (RCTs) were ultimately included. These studies involved a total of 1,803 participants, all of whom were female. The mean age of the participants was 52.76 ± 10.15 years, and the average body mass index (BMI) was 28.67 ± 6.19. All participants were diagnosed with stage 0-III breast cancer. The most common exercise intervention was combined aerobic and resistance exercise (CE, n = 13), followed by aerobic exercise (CAE, n = 14). The average exercise dose across all intervention groups was 1,026.5 METs·min/week Fig. 1. PRISMA flow diagram Risk of bias assessment Figure 2 summarizes the risk of bias assessment for the 26 included studies, conducted using the RevMan 5.4 tool across five independent sources. Due to the inability to meet double-blind experimental conditions for exercise interventions, the Binding of participants and personnel (performance bias) analysis was not performed. The assessment revealed that the risk of bias was generally low, with the following results: Random sequence generation (selection bias) showed a low-risk rate of 96.15%, Allocation concealment (selection bias) had a low-risk rate of 61.54%, Binding of outcome assessment (detection bias) had a low-risk rate of 73.07%, Incomplete outcome data (attrition bias) had a low-risk rate of 80.77%, and Selective reporting (reporting bias) showed a low-risk rate of 84.62%. Summary of risk of bias Pairwise meta-analysis A meta-analysis was conducted on 26 studies evaluating 10 metabolic markers related to the IGF axis (Tables 2 and 3). All exercise interventions were grouped as the Physical Activity (PA) group, with the control group receiving routine care. As shown in Table 1, exercise interventions resulted in significant changes in the following markers: IGF-1 was significantly downregulated, IGFBP-1 was significantly downregulated, IGFBP-3 was significantly upregulated, BMI was significantly reduced, fasting insulin was significantly decreased, HOMA-IR was significantly decreased, triglycerides were significantly reduced, while fasting insulin showed no significant change. Additionally, low-density lipoprotein (LDL) was significantly reduced, and high-density lipoprotein (HDL) was significantly increased. Table 2Demographic and study characteristics of included studiesStudyInterventionmethod = nAge(mean ± sd)BMI(mean ± sd)Time(week)Frequenc(n/week)Time(min/n)doseCancerstagingOperation(Y/N)Chemotherapy(Y/N)RegionAzamian2015MBE = 1345.05 ± 8.2330.33 ± 5.45123601050II-IIIYYIranCG = 1443.05 ± 7.0828.05 ± 4.17000II-IIIYYIranBruno2018CAE = 1855.8 ± 6.825.4 ± 3.8120601630I-IIIY/MilanCG = 2057.1 ± 6.724.6 ± 3.47300I-IIIY/MilanChang2020CAE = 1751.4 ± 7.522.7 ± 2.612360672I-III/NKoreaCG = 1750 ± 6.124.6 ± 4.4000I-III/NKoreaD’Alonzo2021CE = 5059.3 ± 8.834.5 ± 748//1860//YAmericaCG = 5160.5 ± 8.933.3 ± 5.3000//YAmericaDieli2018CE = 5052.8 ± 10.633.1 ± 5.716343.331284I-III/YAmericaCG = 5053.6 ± 10.133.4 ± 5.2000I-III/YAmericaDieli2019CG = 2946.9 ± 10.235.1 ± 6.116343.3300-IIIYYAmericaCG = 2755.6 ± 10.533.2 ± 5.90000-IIIYYAmericaCE = 1946.7 ± 1034.7 ± 6.4343.330-IIIYYAmericaCG = 2255.9 ± 10.333 ± 5.50000-IIIYYAmericaEhret2022CAE = 2736.5 ± 5.525.9 ± 5.42011501095/N/AmericaCAE = 2733.8 ± 6.524.2 ± 4.911502190/N/AmericaCG = 2735.6 ± 6.727.5 ± 6.600/N/AmericaFairey2003CAE = 2459 ± 529.4 ± 7.4153352250I-IIIAYYCanadaCG = 2858 ± 629.1 ± 6.1000I-IIIAYYCanadaFebvey2021CAE = 4053.8 ± 10.925.2 ± 5.824252.5785II-IIIYYFranceCG = 1848.9 ± 11.125.1 ± 6.1000II-IIIYYFranceGuinan2013CAE = 1650.05 ± 8.27/82281048I-IIIYYFranceCG = 1045.05 ± 9.04/000I-IIIYYFranceHarvie2019CE = 13454.6 ± 11.226.9 ± 4.824//540I-III//BritainCE = 13754 ± 9.227 ± 5.1//180I-III//BritainCG = 13855.3 ± 10.527.6 ± 6.1000I-III//BritainIrwin2009CAE = 3656.4 ± 9.530.4 ± 624330578.50-IIIA/YAmericaCG = 3255.6 ± 7.730.1 ± 7.40000-IIIA/YAmericaIsane2023CAE = 1044 ± 9.1426.05 ± 3.4812333759I-IIIYYIranCAE = 1046.29 ± 11.0728.1 ± 5.14341615I-IIIYYIranCG = 1044.9 ± 5.0227.37 ± 4.88000I-IIIYYIranJanelsins2011MBE = 954.33 ± 10.6424.89 ± 5.78123608490-IIIBY/AmericaCG = 1052.7 ± 6.6724.97 ± 4.933000-IIIBY/AmericaKim2017CE = 1156 ± 6.523.9 ± 2.7123601098I-IIIYYKoreaCG = 1349.3 ± 4.825 ± 4.7000I-IIIYYKoreaLahart2016CE = 4052.4 ± 10.327.25 ± 4.69245301532I-IIIYYBritainCG = 4054.7 ± 8.326.67 ± 4.04000I-IIIYYBritainLigibel2008CE = 4052 ± 930.3 ± 5.916/1301025I-IIIYYAmericaCG = 4253 ± 931.4 ± 6.8000I-IIIYYAmericaLigibel2009CE = 515330.3 ± 5.916//962I-III/YAmericaCG = 505331.4 ± 6.8000I-III/YAmericaLigibel2019CE = 2752.3 ± 9.630.7 ± 6.144551554I-IIIYYAmericaCG = 2253.1 ± 7.929.1 ± 7.4000I-IIIYYAmericaNuri2012CE = 1458.27 ± 6.3127.9 ± 3.515447.5866I-IIIBYYIranCG = 1558.27 ± 6.3127.4 ± 3.4000I-IIIBYYIranNuri2014CE = 1557.22 ± 6.42/15452.5640II-IIIYYIranCG = 1557.14 ± 6.21000II-IIIYYIranSchmiz2005RE = 4253.3 ± 8.725.94 ± 0.73122607200-IIIYYAmericaCG = 4352.8 ± 7.625.76 ± 0.730000-IIIYYAmericaShahrokhian2022RE:45.2 ± 3.0529.83 ± 0.8310/3687I-IIY/IranCG=43.6 ± 2.7629.71 ± 1.1410000I-IIY/IranSprod2012MBE = 1154.33 ± 3.5524.89 ± 1.93123608490-IIIBY/AmericaCG = 1052.7 ± 2.1124.97 ± 1.390000-IIIBY/AmericaThomas2013CE = 3556.5 ± 9.830.8 ± 5.9243307500-IIIA/YAmericaCG = 3055.1 ± 7.629.4 ± 7.40000-IIIA/YAmericaUth2020MBE = 4647.4 ± 9.425.5 ± 4.948260885I-IIIYYDenmarkCG = 2250 ± 9.326.4 ± 4.6000I-IIIYYDenmark Demographic and study characteristics of included studies Azamian 2015 Bruno 2018 Chang 2020 D’Alonzo 2021 Dieli 2018 Dieli 2019 Ehret 2022 Fairey 2003 Febvey 2021 Guinan 2013 Harvie 2019 Irwin 2009 Isane 2023 Janelsins 2011 Kim 2017 Lahart 2016 Ligibel 2008 Ligibel 2009 Ligibel 2019 Nuri 2012 Nuri 2014 Schmiz 2005 Shahrokhian 2022 Sprod 2012 Thomas 2013 Uth 2020 Table 3Summary of meta-analysis results: PA vs. control (10 indicators)StudiesParticipantsStatistical MethodEffect Estimate P IGF-112658Mean Difference (Ⅳ, Random, 95% CI)-6.09 [-12.17, -0.02]0.05IGFBP-14170Mean Difference (Ⅳ, Fixed, 95% CI)-2.94 [-4.19, -1.70]0.001IGFBP-39460Mean Difference (Ⅳ, Random,95% CI)3.52 [2.01, 5.03]0.001BMI14631Mean Difference (Ⅳ, Fixed, 95% CI)-0.91 [-1.63, -0.20]0.01Insulin281656Mean Difference (Ⅳ, Random, 95% CI)-1.65 [-2.32, -0.98]0.001HOMA-IR211441Mean Difference (Ⅳ, Random, 95% CI)-0.11 [-0.20,-0.02]0.01Triglycerides161201Mean Difference (Ⅳ, Random, 95% CI)-31.80 [-44.19, -19.41]0.001Blood Glucose231551Mean Difference (Ⅳ, Random, 95% CI)0.48 [-0.29, 1.26]0.22LDL12898Mean Difference (Ⅳ, Random, 95% CI)-0.23 [-0.33, -0.12]0.001HDL151112Mean Difference (Ⅳ, Random, 95% CI)4.81 [2.66, 6.96]0.001 Summary of meta-analysis results: PA vs. control (10 indicators) Network meta-analysis Table 4 and Fig. 3 present the results of the network meta-analysis assessing the effects of continuous aerobic exercise (CAE), combined aerobic And resistance exercise (CE), And resistance exercise (RE) on IGF-1, Insulin, And HOMA-IR markers. For the IGF-1 marker, CAE demonstrated the highest SUCRA value (81.42), followed by CE (75.53), while RE (10.06) was lower than the control group (CG, 32.99). For Insulin, CE had the highest SUCRA value (89.87), followed by CAE (54.97) And RE (38.17). Regarding HOMA-IR, CE showed the highest SUCRA value (85.77), followed by CAE (54.63) And RE (41.02). These findings highlight the relative effectiveness of each exercise modality in improving metabolic outcomes in breast cancer patients Table 4Network meta-analysis of the effects of different exercise modalities on IGF-1, insulin, and HOMA-IR Insulin-like Growth Factor 1 (IGF-1) CAE 2.02(-21.85, 21.53)10.95(-4.98, 26.58)23.23(-10.64, 56.64)-2.02(-21.53, 21.85) CE 8.91(-3.71, 25.66)20.99(-9.64, 56.66)-10.95(-26.58, 4.98)-8.91(-25.66, 3.71) CG 12.31(-17.66, 42.09)-23.23(-56.64, 10.64)-20.99(-56.66, 9.64)-12.31(-42.09, 17.66) RE Fasting Plasma Insulin (Insulin) CAE -2.43(-7.38, 1.98)2.06(-1.36, 5.63)1.18(-7.11, 9.49)2.43(-1.98, 7.38) CE 4.51(1.66, 8)3.6(-4.25, 12.07)-2.06(-5.63, 1.36)-4.51(-8, -1.66) CG -0.89(-8.51, 6.6)-1.18(-9.49, 7.11)-3.6(-12.07, 4.25)0.89(-6.6, 8.51) RE Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) CAE -0.17(-1.8, 0.32)0.2(-0.32, 0.81)0.09(-1.31, 1.21)0.17(-0.32, 1.8) CE 0.37(0.04, 1.86)0.19(-0.59, 2.21)-0.2(-0.81, 0.32)-0.37(-1.86, -0.04) CG -0.08(-1.42, 0.85)-0.09(-1.21, 1.31)-0.19(-2.21, 0.59)0.08(-0.85, 1.42) RE The data shown in the table represent the mean differences and 95% credible intervals, derived from the results of the network meta-analysis. The groups included are: CAE (continuous aerobic exercise), CE (combined aerobic and resistance exercise), RE (resistance exercise), and CG (conventional care group) Network meta-analysis of the effects of different exercise modalities on IGF-1, insulin, and HOMA-IR 2.02 (-21.85, 21.53) 10.95 (-4.98, 26.58) 23.23 (-10.64, 56.64) -2.02 (-21.53, 21.85) 8.91 (-3.71, 25.66) 20.99 (-9.64, 56.66) -10.95 (-26.58, 4.98) -8.91 (-25.66, 3.71) 12.31 (-17.66, 42.09) -23.23 (-56.64, 10.64) -20.99 (-56.66, 9.64) -12.31 (-42.09, 17.66) -2.43 (-7.38, 1.98) 2.06 (-1.36, 5.63) 1.18 (-7.11, 9.49) 2.43 (-1.98, 7.38) 4.51 (1.66, 8) 3.6 (-4.25, 12.07) -2.06 (-5.63, 1.36) -4.51 (-8, -1.66) -0.89 (-8.51, 6.6) -1.18 (-9.49, 7.11) -3.6 (-12.07, 4.25) 0.89 (-6.6, 8.51) -0.17 (-1.8, 0.32) 0.2 (-0.32, 0.81) 0.09 (-1.31, 1.21) 0.17 (-0.32, 1.8) 0.37 (0.04, 1.86) 0.19 (-0.59, 2.21) -0.2 (-0.81, 0.32) -0.37 (-1.86, -0.04) -0.08 (-1.42, 0.85) -0.09 (-1.21, 1.31) -0.19 (-2.21, 0.59) 0.08 (-0.85, 1.42) The data shown in the table represent the mean differences and 95% credible intervals, derived from the results of the network meta-analysis. The groups included are: CAE (continuous aerobic exercise), CE (combined aerobic and resistance exercise), RE (resistance exercise), and CG (conventional care group) Presents the network meta-analysis results, including the Radial SUCRA plot (Ⅰ) and the SUCRA Ranking Panel (Ⅱ). Panel (A) shows the results for IGF-1, Panel (B) for fasting insulin, and Panel (C) for HOMA-IR. In the Radial SUCRA plot, higher SUCRA values indicate better treatments, with the size of the nodes representing the number of participants and the thickness of the lines indicating the number of trials conducted. The SUCRA Ranking Panel highlights that higher SUCRA values and cumulative ranking curves closer to the top left reflect better performance Dose-response network meta-analysis The dose-response NMA revealed the optimal exercise dose range for female breast cancer patients based on total exercise volume (Physical Activity, Continuous Aerobic Exercise, and Combined Aerobic and Resistance Exercise) and various metabolic markers (IGF-1, Insulin, and HOMA-IR). When exercise modality was not differentiated, a total exercise volume of 560–1080 METs·min/week was found to be the most effective dose range for improving metabolic outcomes in breast cancer patients. For Continuous Aerobic Exercise, the recommended optimal dose range was also 560–1080 METs·min/week. For Combined Aerobic and Resistance Exercise, the optimal dose range for IGF-1 was 760–1600 METs·min/week, for Insulin was 460–1600 METs·min/week, and for HOMA-IR was 460–930 METs·min/week. These findings suggest that moderate to high levels of exercise, particularly within these dose ranges, are most beneficial for improving insulin-related metabolic markers in breast cancer patients (Fig. 4). Dose-response relationships between total physical activity (PA) and different exercise interventions with various indicators: (Ⅰ) Dose-response relationship between total physical activity (PA) and indicators, (Ⅱ) Dose-response relationship between continuous aerobic exercise (CAE) and indicators, (Ⅲ) Dose-response relationship between combined aerobic and resistance exercise (CE) and indicators. A IGF-1, (B) Fasting Insulin, (C) HOMA-IR. The x-axis represents the dose level of the exercise intervention, while the y-axis shows the predicted response values for each indicator at the corresponding dose. Higher response values indicate stronger treatment effects. The black solid lines represent the predicted response curves at different exercise doses, while the dashed lines indicate the 95% confidence intervals, reflecting the uncertainty in the predicted responses. The vertical lines highlight the predicted changes in each indicator at specific exercise doses. The dark green area emphasizes the variability in dose effects within the key dose range Sensitivity analysis and network meta-regression To validate the reliability of the study results and the robustness of the prediction models, a sensitivity analysis was performed on the included studies. Using the network meta-analysis (NMA) model, along with Gelman convergence assessment, bias reporting, residual bias analysis, and leverage value analysis, the impact of exercise interventions on the metabolic markers and IGF axis in breast cancer patients was comprehensively evaluated. These analytical methods ensured the stability of the model and identified the goodness of fit and key influencing factors of each study. The sensitivity analysis involved two key steps: (1) excluding studies with high risk of bias, and (2) excluding studies with high residuals. Detailed findings are provided in the supplementary information. The sensitivity analysis demonstrated that excluding studies with high bias risk or high residuals did not alter the initial SUCRA rankings of exercise interventions. Discussion Mechanistic analysis of IGF axis and metabolic markers in the progression of breast cancer The IGF axis plays a pivotal role in breast cancer progression by promoting tumor proliferation and survival through the PI3K/Akt and MAPK pathways2 [23–25]. Its regulation is closely intertwined with metabolic factors such as insulin, BMI, and triglycerides, and dysregulation of this signaling–metabolism network amplifies tumor growth and therapeutic resistance [26]. Therefore, understanding how exercise modulates the IGF axis and related metabolic markers is essential to elucidate its potential anti-tumor mechanisms in breast cancer. Exercise intervention on IGF axis and metabolic markers in breast cancer patients This study provides a multi-dimensional analysis of the effects of exercise intervention on the IGF axis and metabolic markers in female breast cancer patients. The analysis primarily focused on IGF-1, IGFBP-1, IGFBP-3, BMI, insulin, insulin resistance, fasting blood glucose, triglycerides, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) [27]. Results revealed that exercise significantly upregulated IGFBP-3 and HDL, while significantly downregulating IGF-1, IGFBP-1, BMI, insulin, insulin resistance, fasting blood glucose, triglycerides, and LDL. The exercise intervention did not produce significant changes in fasting blood glucose levels. However, analysis of the effects of different exercise modalities showed that only CE significantly lowered fasting blood glucose levels. This may be related to the insufficient intensity of single-mode exercise in cancer patients [28–30]. Physical activity, as an endogenous, non-pharmacological intervention, may reduce the risk of breast cancer partly through modulation of the IGF signaling pathway; however, this mechanism requires further confirmation, particularly considering other potential pathways such as hormonal regulation [28–31]. Regular physical activity has been shown to reduce insulin levels and improve insulin sensitivity, thereby indirectly inhibiting the synthesis and activity of IGF-1 [32, 33]. Additionally, physical activity improves overall energy metabolism, inhibits fat accumulation [34, 35], reduces BMI and triglyceride levels, and alleviates the metabolic burden associated with elevated IGF-1. In postmenopausal women, insulin resistance is more common, and exercise intervention not only reduces the stimulation of insulin and IGF-1 by adipose tissue but also effectively reverses abnormal IGF-1 signaling activation in the context of metabolic syndrome [36]. Exercise’s impact on the IGF axis and metabolic markers is considered a potential anti-cancer function that inhibits tumor cell division, metastasis, and stemness maintenance [37]. Therefore, physical activity not only downregulates insulin and IGF-1 levels to suppress IGF signaling activation but also stabilizes IGFBP levels (such as IGFBP-1 and IGFBP-3), thereby delaying IGF signaling, providing biological mechanism support for reducing breast cancer incidence, slowing disease progression, and improving patient prognosis. This also provides theoretical evidence for exercise intervention strategies in breast cancer patients [1, 38]. Effectiveness and dose analysis of exercise interventions In the network meta-analysis (NMA), this study selected three representative indicators (IGF-1, insulin, and HOMA-IR) from randomized controlled trials (RCTs) for analysis of the effects of different exercise interventions. All exercise types were categorized into continuous aerobic exercise (CAE), combined aerobic and resistance exercise (CE), and resistance exercise (RE). The results revealed that, for IGF-1, continuous aerobic exercise was most effective in inhibition, followed by combined aerobic and resistance exercise, while resistance exercise upregulated IGF-1 expression. For fasting insulin and HOMA-IR, combined aerobic and resistance exercise showed the best inhibitory effect, followed by continuous aerobic exercise and resistance exercise. In the subsequent dose analysis, PA (regardless of exercise form), CAE, and CE were analyzed for their dose-response effects. When the exercise volume reached 560 METs·min/week, PA, CAE, and CE all entered the lowest threshold of the optimal dose, while the highest optimal dose was achieved at 1080 METs·min/week. For combined aerobic and CE, the optimal dose range varied by outcome: 760–1600 METs·min/week was most effective for reducing IGF-1 levels, 460–1600 METs·min/week for lowering fasting insulin, and 460–930 METs·min/week for improving HOMA-IR. The findings suggest that CAE has the most significant intervention effect on breast cancer-related metabolic indicators, especially in lowering IGF-1 levels [39–41]. The IGF-1 levels in the CAE group were significantly lower than those in other groups, suggesting that continuous aerobic exercise may play a potential role in attenuating tumor-related growth signaling. However, this interpretation should be considered cautiously, as further mechanistic studies are needed to confirm whether IGF-1 reduction directly contributes to antitumor effects [42], which may be closely related to its ability to effectively reduce insulin levels, decrease IGF-1 synthesis, and enhance IGFBP regulation [43, 44]. In contrast, RE alone, while improving insulin levels and HOMA-IR, showed negative effects on IGF-1 regulation, with IGF-1 levels higher than in the CAE group and even the control group (CG) [27, 45–47]. This suggests that RE may activate anabolic signaling pathways (such as AKT/mTOR) and stimulate IGF-1 production [44, 48–50]. Therefore, for breast cancer patients where IGF-1 is a key oncogenic pathway, the sole use of resistance exercise may not be the optimal choice. Compared to aerobic exercise alone, CE showed superior effects on improving insulin levels and HOMA-IR, likely due to resistance training’s promotion of muscle mass, enhanced glucose uptake, and improved insulin sensitivity [51–55]. However, in terms of IGF-1 intervention, CE was still less effective than CAE, suggesting that the resistance component of CE partially counteracts the inhibitory effects of aerobic exercise on IGF-1 [56–58]. In summary, the effectiveness of exercise interventions in remodeling metabolic markers in breast cancer patients is highly dependent on the type of exercise modality. Aerobic exercise, particularly, plays a critical role in inhibiting IGF axis activity, making it especially beneficial for breast cancer patients with clinically elevated IGF-1 levels, such as those whose IGF-1 concentrations exceed the normal reference range for age-matched healthy individuals. On the other hand, CE appears to be more effective for patients with significant metabolic abnormalities, such as insulin resistance, elevated fasting insulin, or increased BMI, as it targets both insulin sensitivity and overall metabolic health. While RE offers notable metabolic benefits, particularly in improving insulin sensitivity and muscle mass, its impact on IGF-1 levels warrants careful consideration. Resistance training may inadvertently stimulate IGF-1 production, thus requiring a more cautious approach in patients where IGF-1 is a critical oncogenic pathway. Therefore, tailored exercise prescriptions, considering both metabolic status and IGF-1 regulation, are essential to optimize therapeutic outcomes in breast cancer patients. Limitations and future perspectives This study employed a review approach, utilizing pairwise meta-analysis, NMA, and dose-response meta-analysis to assess the effects of various exercise interventions on the IGF axis and metabolic markers in breast cancer patients. Sensitivity analysis was conducted to exclude potential factors that may affect result stability, such as high bias risk and residual deviance. However, this study has several limitations. First, as exercise is a non-pharmacological intervention that cannot meet double-blind standards, systemic errors may arise in patient treatment. Second, some of the studies included did not provide detailed descriptions of exercise interventions, including intervention time, frequency, and intensity, which presented challenges in dose analysis. Third, due to the limited number of randomized controlled trials (RCTs) and supporting literature on IGFBP-1 and IGFBP-3, a comprehensive analysis of the effects of different exercise interventions could not be conducted, which may reduce the accuracy and generalizability of our study results. Lastly, this study did not conduct subgroup analyses for different regions, cultures, ethnicities, or age groups, which may limit the generalizability of the findings. As high-quality research on exercise intervention for breast cancer patients’ IGF axis and metabolic markers remains relatively scarce, existing literature exhibits heterogeneity in intervention types, evaluation indicators, and study designs, which limits the stability and applicability of the conclusions. To more comprehensively and systematically elucidate the mechanisms and effects of exercise interventions, future research may integrate large clinical databases (such as SEER, TCGA, UK Biobank) for data mining and validation, enhancing the breadth and depth of analysis, and providing data support and theoretical foundations for constructing individualized exercise intervention strategies based on real-world evidence. Conclusion This study demonstrates that exercise has a beneficial intervention effect on the IGF axis and metabolic markers in breast cancer patients. Specifically, continuous aerobic exercise is most effective in inhibiting IGF-1, followed by combined aerobic and resistance exercise. For fasting insulin and HOMA-IR, combined aerobic and resistance exercise shows the best inhibitory effect, with continuous aerobic exercise and resistance exercise exhibiting lesser effects. Dose-response analysis further identified a non-linear positive relationship between exercise and IGF-1, insulin resistance, and fasting insulin levels, and determined the optimal effective dose ranges for different exercise modalities. These findings lay the foundation for the development of personalized exercise prescriptions for breast cancer patients and provide valuable insights for optimizing exercise interventions to improve IGF-1, insulin resistance, and fasting insulin levels in clinical settings. However, given the limited number of studies included in this meta-analysis, larger-scale studies are necessary to further elucidate and expand our understanding of the relationship between exercise dose and IGF-1, insulin resistance, and fasting insulin levels in this population. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Supplementary Material 5. Supplementary Material 6. Supplementary Material 7. Supplementary Material 8. Acknowledgements We would like to thank the author services for their assistance on deeply polishing of this work. Authors’ contributions In this study, BCA was responsible for drafting the manuscript and collecting data. QWH contributed to data collection and performed data analysis. HBC and QHY were involved in literature collection and data analysis. YY provided technical guidance and conducted data analysis. Funding This work is supported by author Yuan’s projects. Taishan Scholars Talent Program (No.tsqn202312181), Shandong province Youth Innovation Team Project (No.2023RW102). Data availability All data generated or analyzed during this study are included in this published article and its supplementary information files. Data extracted for the meta-analysis are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Informed consent was obtained from all individual participants included in the study. All participants were provided with detailed information about the study's objectives, procedures, potential risks, and benefits before giving their consent. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Bing-Ao Chen, Wen-Hui Qin and Hong-Bao Chen contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Supplementary Material 5. Supplementary Material 6. Supplementary Material 7. Supplementary Material 8. Data Availability Statement All data generated or analyzed during this study are included in this published article and its supplementary information files. Data extracted for the meta-analysis are available from the corresponding author on reasonable request.