The evolving efficacy landscape of intra-articular injections for rotator cuff injuries over time: a network meta-analysis of randomized controlled studies Authors: Morong Wu, Chenshu Yan, Yan Xu, Zheng Yan Published: 2025 ABSTRACT BackgroundRotator cuff injury (RCI), a leading cause of shoulder disability, necessitates effective non-surgical interventions. Despite widespread use of intra-articular injections (corticosteroids, hyaluronic acid [HA], prolotherapy, platelet-rich plasma [PRP]), comparative efficacy across short-, mid-, and long-term outcomes remains unclear due to heterogeneous evidence.PurposeTo systematically review the literature to compare corticosteroids, hyaluronic acid, prolotherapy, and platelet-rich plasma for rotator cuff injuries, using network meta-analysis stratified by short-, mid-, and long-term outcomes.MethodsA comprehensive search was conducted across PubMed, Embase, Cochrane Library, Web of Science, Scopus, CNKI, VIPC, and Wanfang Data up to May 2025. Pain (VAS) and functional outcomes (e.g., Constant-Murley Score) were assessed at short- (≤ 1 month), mid- (1–3 months), and long-term (> 3 months) intervals. Frequentist models with random effects ranked interventions via SUCRA.ResultsTwenty-nine RCTs (1800 patients) were included. For pain relief, prolotherapy consistently ranked highest across all timeframes. PRP also exhibited favorable long-term analgesic efficacy. Corticosteroids showed short-term benefits but inferior long-term efficacy. For functional improvement, HA demonstrated the best short-term results, whereas prolotherapy was optimal for mid-term recovery. Long-term functional outcomes favored HA and prolotherapy. The detailed SUCRA values and rankings for all interventions and timepoints are presented in Table 2.ConclusionsComparative efficacy analysis suggests temporal variations in treatment outcomes: prolotherapy and HA offer the most favorable early (≤ 3 months) symptom control for both pain and function. For sustained pain relief (> 3 months), prolotherapy and PRP rank highest, while HA and prolotherapy are associated with better long-term functional outcomes. Corticosteroids appear most effective for short-term analgesia, with diminishing efficacy over longer durations. This evidence supports a shift from using a single injection agent towards a tailored strategy that selects different injections based on the patient’s symptoms and the specific treatment phase to achieve optimal outcomes. PROSPERO registration: PROSPERO registration number: CRD42025632416.Level of evidenceLevel I Systematic Review and MetaAnalysis.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13018-025-06475-8. Introduction Rotator cuff injury (RCI) is a prevalent and debilitating shoulder disorder that significantly impacts patients’ quality of life and functional capacity. The rotator cuff, comprising four muscles (supraspinatus, infraspinatus, teres minor, and subscapularis) and their tendons, plays a pivotal role in stabilizing the shoulder joint and facilitating a wide range of upper limb movements. Patients with RCI often experience chronic pain, weakness, and restricted mobility, which severely impair their ability to perform daily activities such as lifting objects, reaching overhead, dressing, and even sleeping [1]. The global burden of RCI is substantial, with prevalence rates increasing markedly with age: approximately 25–30% of individuals aged 50–60 years and up to 60% of those over 80 years are affected, making it one of the leading causes of shoulder-related disability worldwide [2]. The pathogenesis of RCI is multifactorial, involving a complex interplay of intrinsic factors (e.g., age-related tendon degeneration, reduced vascularity) and extrinsic factors (e.g., mechanical compression, repetitive overhead activities), which collectively contribute to tendon microtrauma, progressive degeneration, and eventual tearing. Notably, RCI can also present with an insidious onset, where structural damage visible on imaging may precede the development of significant clinical symptoms [3]. Furthermore, it is well recognized that imaging success, such as tendon healing after repair, does not necessarily correspond to clinical outcomes, underscoring the complexity of managing RCI and the importance of focusing on patient-centered measures like pain and function [4, 5]. Rotator cuff tears therefore represent a spectrum of degenerative and traumatic shoulder pathologies, with a high prevalence that increases markedly with age [1]. The management of rotator cuff injuries is guided by tear characteristics. While surgical repair is typically indicated for severe or full-thickness tears, with techniques like arthroscopic and mini-open showing similar effectiveness [6], intra-articular injection serves as a foundational conservative treatment for partial-thickness injuries. This study focuses on optimizing non-surgical care by evaluating the comparative efficacy of injectable agents for the latter group. Although intra-articular injections are widely used for RCI, the comparative efficacy of available agents (corticosteroids, HA, prolotherapy, and PRP) across different timeframes remains uncertain. Systematic reviews and network meta-analysis(NMA) have emerged as powerful tools to address this gap, enabling the synthesis of direct and indirect evidence to rank interventions based on their comparative efficacy [7]. By integrating data from multiple studies, NMA provides a comprehensive framework for evaluating treatment options, thereby informing evidence-based clinical decision-making [8]. Given the clinical importance of RCI and the ongoing need to identify the most effective injectable therapies, this study aims to conduct a systematic review and NMA. A key feature of this analysis is the explicit stratification of outcomes by follow-up interval (short-term: ≤1 month, mid-term: 1–3 months, long-term: >3 months). This stratified approach facilitates a direct comparison of the efficacy of intra-articular corticosteroids, hyaluronic acid (HA), prolotherapy, and platelet-rich plasma (PRP) for RCI, with a specific focus on evaluating how treatment effects may vary over time. This temporal dimension, while recognized as important, has not been consistently or comprehensively addressed in prior comparative syntheses. The findings are expected to generate evidence to inform the selection of injection therapies based on treatment timelines, thereby providing more nuanced guidance for clinicians and potentially improving patient outcomes. Methods Study design and registration This systematic review and NMA was conducted to evaluate the comparative efficacy of intra-articular injections of corticosteroids, HA, prolotherapy, and PRP in the treatment of RCI. The study protocol was prospectively registered in the PROSPERO international registry of systematic reviews (Registration number: CRD42025632416) and adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Network Meta-Analyses (PRISMA-NMA) guidelines [9]. This approach ensures transparency, reproducibility, and methodological rigor in the synthesis of evidence. Data sources and search strategy A comprehensive and systematic literature search was conducted across multiple electronic databases, including PubMed, Embase, Cochrane Library, Web of Science, Scopus, CNKI, VIPC, and Wanfang Data, from their inception to Aug 10, 2025. The search strategy was designed to capture all relevant studies by combining Medical Subject Headings (MeSH) terms and free-text keywords related to “rotator cuff injury,” “intra-articular injection,” “corticosteroids,” “hyaluronic acid,” “prolotherapy,” “platelet-rich plasma,” and “randomized controlled trials (RCTs).” No language restrictions were imposed to minimize selection bias. The detailed search strategy for each database is provided in Supplementary Fig. S1. Eligibility criteria Studies were included if they met the following criteria: Population Adults (≥ 18 years) diagnosed with rotator cuff injuries (partial rotator cuff tears). This focus on partial-thickness tears was chosen because non-surgical interventions, including intra-articular injections, are primarily indicated and commonly utilized for this patient group, whereas full-thickness tears often require surgical consideration [1, 4]. To ensure the comparability of the patient population across studies, included RCTs were required to have a diagnosis of partial-thickness tear confirmed by standardized imaging modalities, such as magnetic resonance imaging (MRI) or ultrasonography. Interventions Intra-articular injections of corticosteroids, hyaluronic acid, hypertonic glucose, or PRP. Comparators Placebo, routine functional exercise, no treatment, or any of the active interventions listed above. Outcomes Primary outcome: Pain relief, measured using the Visual Analog Scale (VAS); Secondary outcome: Improvement in shoulder function, assessed by validated tools such as the Constant-Murley Score (CMS), Shoulder Pain and Disability Index (SPADI), American Shoulder and Elbow Surgeons Score (ASES), University of California Los Angeles Shoulder Scale(UCLA), and Disabilities of the Arm, Shoulder and Hand (DASH) questionnaire, among others. (short-term follow-up ≤ 1 month, 1 month< mid-term follow-up ≤ 3 months, long-term follow-up >3 months) Study design RCTs published in peer-reviewed journals. Population Adults (≥ 18 years) diagnosed with rotator cuff injuries (partial rotator cuff tears). This focus on partial-thickness tears was chosen because non-surgical interventions, including intra-articular injections, are primarily indicated and commonly utilized for this patient group, whereas full-thickness tears often require surgical consideration [1, 4]. To ensure the comparability of the patient population across studies, included RCTs were required to have a diagnosis of partial-thickness tear confirmed by standardized imaging modalities, such as magnetic resonance imaging (MRI) or ultrasonography. Population Adults (≥ 18 years) diagnosed with rotator cuff injuries (partial rotator cuff tears). This focus on partial-thickness tears was chosen because non-surgical interventions, including intra-articular injections, are primarily indicated and commonly utilized for this patient group, whereas full-thickness tears often require surgical consideration [1, 4]. To ensure the comparability of the patient population across studies, included RCTs were required to have a diagnosis of partial-thickness tear confirmed by standardized imaging modalities, such as magnetic resonance imaging (MRI) or ultrasonography. Interventions Intra-articular injections of corticosteroids, hyaluronic acid, hypertonic glucose, or PRP. Interventions Intra-articular injections of corticosteroids, hyaluronic acid, hypertonic glucose, or PRP. Comparators Placebo, routine functional exercise, no treatment, or any of the active interventions listed above. Comparators Placebo, routine functional exercise, no treatment, or any of the active interventions listed above. Outcomes Primary outcome: Pain relief, measured using the Visual Analog Scale (VAS); Secondary outcome: Improvement in shoulder function, assessed by validated tools such as the Constant-Murley Score (CMS), Shoulder Pain and Disability Index (SPADI), American Shoulder and Elbow Surgeons Score (ASES), University of California Los Angeles Shoulder Scale(UCLA), and Disabilities of the Arm, Shoulder and Hand (DASH) questionnaire, among others. (short-term follow-up ≤ 1 month, 1 month< mid-term follow-up ≤ 3 months, long-term follow-up >3 months) Outcomes Primary outcome: Pain relief, measured using the Visual Analog Scale (VAS); Secondary outcome: Improvement in shoulder function, assessed by validated tools such as the Constant-Murley Score (CMS), Shoulder Pain and Disability Index (SPADI), American Shoulder and Elbow Surgeons Score (ASES), University of California Los Angeles Shoulder Scale(UCLA), and Disabilities of the Arm, Shoulder and Hand (DASH) questionnaire, among others. (short-term follow-up ≤ 1 month, 1 month< mid-term follow-up ≤ 3 months, long-term follow-up >3 months) Study design RCTs published in peer-reviewed journals. Study design RCTs published in peer-reviewed journals. Exclusion criteria included non-RCTs, studies lacking essential data for meta-analysis (e.g., missing outcome data [VAS or functional scores], insufficient statistical detail [e.g., missing standard deviations, sample sizes per group for outcomes of interest], or inability to extract data specific to partial rotator cuff tears after contacting authors if necessary), and those focusing on surgical interventions or mixed populations (e.g., combined with other shoulder pathologies such as adhesive capsulitis, glenohumeral osteoarthritis, subacromial impingement syndrome, acromioclavicular joint disorders, or biceps tendinopathy). Study selection and data extraction The literature screening process was conducted following a standardized protocol. Initially, all identified records were imported into EndNote X9 for systematic management and duplicate removal. Three independent reviewers (Morong Wu, Chenshu Yan, Yan Xu) performed the initial screening by evaluating titles and abstracts against predefined inclusion criteria. Studies that potentially met the eligibility criteria proceeded to full-text review, where both reviewers independently assessed their suitability for inclusion. Any discrepancies between reviewers during the screening process were resolved through consensus discussions or, when necessary, by consulting a fourth reviewer(Zheng Yan) to ensure objectivity and minimize bias. The literature screening process was conducted in accordance with the PRISMA 2020 Flow Diagram, a widely recognized framework for systematic reviews and meta-analyses [4]. Data extraction was systematically performed using a standardized form to capture key study characteristics, including authorship, publication year, country of origin, as well as detailed information on patient demographics, intervention protocols, outcome measures, and follow-up durations. In cases where data were incomplete or unclear, we contacted the corresponding authors to seek necessary clarifications. If key outcome data (e.g., means, standard deviations, or sample sizes for the outcomes of interest) remained unavailable after attempting to contact the authors, the study was excluded from the quantitative synthesis for that specific outcome and timeframe. No data imputation was performed, in order to preserve the integrity of the analysis and avoid potential biases associated with estimating missing values. This rigorous approach ensured the accuracy and comprehensiveness of the collected data for subsequent analysis. As illustrated in Fig. 1, the screening process encompassed the initial identification of records, screening, eligibility assessment, and final inclusion of studies. Risk of bias assessment The methodological quality and risk of bias in the included studies were rigorously evaluated using the Cochrane Risk of Bias Tool for Randomized Trials (RoB 2). Each study was systematically assessed across five critical domains: (1) the randomization process, (2) deviations from the intended interventions, (3) missing outcome data, (4) measurement of outcomes, and (5) selective reporting of results. Based on the evaluation within these domains, studies were classified as having a “low risk of bias,” “some concerns regarding bias,” or “high risk of bias.” According to the scoring criteria, if all five domains were rated as “low risk,” the overall risk of bias was classified as “low risk.” If the evaluation results indicated “some concerns” in one or more domains but no “high risk,” the overall risk of bias was classified as “some concerns.” If any domain was rated as “high risk,” the overall risk of bias was classified as “high risk.” Each domain included three levels of assessment: “low risk,” “some concerns,” and “high risk.” The risk of bias assessment was conducted following a standardized protocol to ensure rigor and consistency. Three independent reviewers (Morong Wu, Chenshu Yan, and Yan Xu) performed the methodological quality evaluation using the Cochrane RoB 2 tool. Each reviewer independently assessed all included studies across the five critical domains. Any discrepancies in domain judgments between the reviewers were resolved through consensus discussions. In cases where consensus could not be reached, a fourth reviewer (Zheng Yan) was consulted to make a final arbitration. This multi-step process was implemented to minimize subjectivity and ensure the objectivity and reliability of the bias assessment. Statistical analysis All outcome measures were analyzed using either random- or fixed-effects models, selected based on heterogeneity levels. Heterogeneity was assessed using Cochran’s Q-test (p-value) and the I² statistic. A fixed-effects model was applied when heterogeneity was low (p ≥ 0.10 and I² < 50%); otherwise, a random-effects model was employed (p < 0.10 or I² ≥ 50%) [10]. Continuous outcomes were expressed as standardized mean differences (SMD) with 95% confidence intervals (CIs). The network meta-analysis was performed within the frequentist framework. This approach was selected for its computational efficiency, direct provision of confidence intervals for effect estimates, and the proven reliability of its ranking metrics (SUCRA values), which are consistent with those derived from Bayesian methods for the purpose of treatment hierarchy evaluation [11]. Prior to conducting the NMA, the underlying assumption of transitivity was critically evaluated. This was assessed by examining the distribution of potential effect modifiers across the available treatment comparisons. Key patient and study characteristics, including mean age, sex distribution, VAS, functional scores were compared qualitatively across the different treatment arms in the evidence network. No major imbalances that would violate the transitivity assumption were identified. NMA was conducted in R software (version 4.2.1) using the gemtc and netmeta packages under a frequentist framework [11, 12]. Evidence networks were constructed to visualize direct and indirect comparisons between interventions, with node sizes proportional to sample sizes and edge thickness reflecting the number of studies for each comparison. Ranking probabilities were estimated using the SUCRA, where higher values (0–100%) indicate a greater likelihood of being the optimal intervention [13]. To ensure model convergence, trace plots and Gelman-Rubin diagnostics were performed. Trace plots were visually inspected for stability, and the potential scale reduction factor (PSRF) was calculated, with values close to 1.0 indicating satisfactory convergence [13]. Global and local inconsistency tests (node-splitting method) were performed to evaluate disagreement between direct and indirect evidence. Sensitivity analyses were conducted by excluding studies with high risk of bias or studies with small sample sizes. For outcomes reported in ≥ 10 studies, funnel plots were generated to assess publication bias, complemented by Egger’s regression test (p < 0.05 indicating significance) [14]. To explore the potential impact of geographic or ethnic variability on the treatment effects, we planned to conduct subgroup analyses or meta-regression based on the region of origin. However, when stratifying the already limited number of studies from non-Asian regions further across the three predefined timeframes (short-, mid-, and long-term), the sample size in many resultant subgroups was insufficient to support a statistically robust comparison. Consequently, these planned analyses were not deemed feasible. As an alternative assessment, we qualitatively examined the distribution of study-level covariates within each treatment comparison in the evidence networks. This assessment did not identify any clear association between geographic region and the magnitude of treatment effects, which provides some indirect support for the transitivity assumption regarding this factor. Results Literature retrieval results A total of 6411 studies were initially identified through database searches and supplementary sources, including PubMed (n = 909), Embase (n = 1105), Cochrane Central Register of Controlled Trials (n = 923), Web of Science (n = 1451), Scopus (n = 1617), CNKI (n = 116), VIPC (n = 137), and WanFang Data (n = 150), with three additional records identified from other sources. After removing 4983 duplicates (4862 detected by Covidence and 119 manually identified), 1428 studies underwent title and abstract screening. During this phase, 1329 studies were excluded due to irrelevance to the research question, such as non-interventional designs, mismatched outcomes, or ineligible populations. Subsequently, 99 full-text articles were retrieved for eligibility assessment, of which one study could not be accessed. Following a detailed evaluation, 69 studies were excluded for the following reasons: duplicates (n = 9), insufficient follow-up duration (n = 13), inappropriate outcome measures (n = 14), incorrect intervention types (n = 4), inadequate sample size (n = 16), insufficient data for analysis (n = 11), or non-target patient populations (n = 2). Ultimately, 29 studies met the inclusion criteria and were incorporated into the quantitative synthesis. The complete screening process is summarized in Fig. 1, adhering to the PRISMA 2020 Flow Diagram. Fig. 1PRISMA flow diagram PRISMA flow diagram Characteristics of included studies Twenty-nine RCTs involving 1800 patients with partial rotator cuff tears were included in this network meta-analysis [15–43]. The studies spanned 13 countries across Asia (China, Iran, Japan, Malaysia, Pakistan, Thailand), Europe (Switzerland, Turkey), Africa (Egypt), North America (Canada), and South America (Argentina), with publication years ranging from 2001 to 2024. The sample sizes per study varied from 24 to 100 participants (median: 60 patients), with a pooled cohort of 943 males (47.4%) and 857 females (52.6%). Patient age ranged from 38.9 ± 7.4 to 62.4 ± 8.6 years across studies. All trials were classified as Level I evidence according to therapeutic study design criteria. The interventions evaluated included intra-articular injections of platelet-rich plasma (PRP, n = 16 studies), hyaluronic acid (HA, n = 9), prolotherapy (n = 10), and corticosteroids (n = 14), compared against placebo (n = 7) or active controls. Follow-up durations were stratified into short-term (≤ 1 month, 100% of studies), mid-term (1–3 months, 93.1%), and long-term (>3 months, 86.2%), with 10 studies extending beyond 6 months. The primary outcome was pain relief, assessed using the visual analogue scale (VAS). Secondary outcomes focused on shoulder function improvement, including the shoulder pain and disability index (SPADI), single assessment numeric evaluation score (SANE score), American shoulder and elbow surgeons score (ASES), Constant-Murley score (CMS), overall shoulder satisfaction, Western Ontario rotator cuff index (WORC), disabilities of the arm, shoulder and hand score (DASH), simple shoulder test (SST), Oxford shoulder score (OSS), University of California Los Angeles shoulder score (UCLA), and shoulder disability questionnaire score (SDQ). A summary of the characteristics of the included studies is presented in Fig. 2 (Table 1). Fig. 2Results of quality evaluation of included studies “+” means “Low risk”; “!” means “Some concerns”; “?” means “High risk”. ① Randomization process; ② Deviations from intended interventions; ③ Missing outcome data; ④ Measurement of the outcome; ⑤ Selection of the reported result; ⑥ Overall Bias Results of quality evaluation of included studies “+” means “Low risk”; “!” means “Some concerns”; “?” means “High risk”. ① Randomization process; ② Deviations from intended interventions; ③ Missing outcome data; ④ Measurement of the outcome; ⑤ Selection of the reported result; ⑥ Overall Bias Table 1Characteristics of included studiesStudy IDAuthorYearCountryLevel of evidenceGroupGroup peopletotalSex(Male/Female)Patient ageFollow-upOutcome index1Adrien et al2020,JanSwitzerlandTherapeutic Level IPRP418022/1948.2 ± 16.27 months, 12 months① + ③ + ④ + ⑤Therapeutic Level IPlacebo3923/1647.6 ± 11.12Asadollah et al2019,FebIranTherapeutic Level ICorticosteroids295712/1756.3 ± 8.31 month①Therapeutic Level IProlotherapy2813/1559.4 ± 9.23YU et al2018,SepChinaTherapeutic Level IHA448924/2038.93 ± 7.351 month, 3 months, 6 months, 12 months④Therapeutic Level IPRP4522/2340.56 ± 7.854Wen et al2010,JunChinaTherapeutic Level IHA25519/1651.16 ± 7.841 week, 2 weeks, 3 weeks, 4 weeks, 5 weeks, 6 weeks, 11 weeks① + ⑤Therapeutic Level IPlacebo2610/1652.38 ± 8.955Brandi et al2017,AprAustraliaTherapeutic Level IProlotherapy173613/451 ± 166 weeks, 3 months, 6 months⑥Therapeutic Level ICorticosteroids1914/546 ± 156Haleh et al2021,MayIranTherapeutic Level IPRP30585/2557.33 ± 9.801 week, 1 month, 3 months① + ⑦ + ⑧Therapeutic Level ICorticosteroids286/2253.60 ± 7.247Samihah et al2023,MayMalaysiaTherapeutic Level IPRP224417/1557.8 ± 11.53 weeks, 6 weeks, 3 months, 6 months① + ②Therapeutic Level IProlotherapy2219/1351.1 ± 12.68Tirthankaret al2023,NovIndiaTherapeutic Level IPRP5010020/3048 ± 116 weeks, 3 months, 6 months⑤Therapeutic Level ICorticosteroids5022/2845 ± 109Shou et al2022,MarChinaTherapeutic Level IPRP24486/1861.8 ± 2.51, 3 months① + ② + ⑤Therapeutic Level IHA245/1961.5 ± 2.110Doaaet al2018,JulyEgyptTherapeutic Level IPRP15306/946.8 ± 10.62 months① + ⑫Therapeutic Level ICorticosteroids157/841.5 ± 12.511Ilkeret al2015,SepTurkeyTherapeutic Level IPRP30629/2159.16 ± 10.7612 months① + ⑧Therapeutic Level IPlacebo328/2459.68 ± 9.8812Chris et al2020,OctKoreaTherapeutic Level IPRP306011/1955.3 ± 10.31 week, 1 month, 3 months, 6 months① + ④ + ⑤ + ⑨ + ⑪ + ⑫Therapeutic Level ICorticosteroids309/2152.5 ± 11.213Serdar et al2013,JulyTurkeyTherapeutic Level IPRP20407/1345.5 ± 11.83 weeks, 6 weeks, 12 weeks, 24 weeks, 1 year① + ② + ⑦Therapeutic Level IPlacebo206/1451.4 ± 10.014Coryet al2021,FebCanadaTherapeutic Level IPRP479916/3149.94 ± 9.706 weeks, 3 months, 12 months① + ④ + ⑦Therapeutic Level ICorticosteroids5219/3349.08 ± 9.5415Che et al2022,FrbChinaTherapeutic Level IProlotherapy295716/1349.10 ± 8.442 weeks, 6 weeks, 12 weeks① + ②Therapeutic Level IPlacebo2814/1452.18 ± 9.8316Alireza et al2013,DecIranTherapeutic Level IHA20408/1248.14 ± 9.231 week, 2 weeks, 3 weeks, 12 weeks② + ⑤Therapeutic Level IPlacebo206/1417Morteza et al2021,JunIranTherapeutic Level IPlacebo336517/1656.90 ± 13.62 weeks, 3 months②Therapeutic Level IProlotherapy3213/1952.5 ± 13.918Merih et al2010,JulyTurkeyTherapeutic Level IHA12243/958.67 ± 6.323 weeks, 3 months, 4 years①Therapeutic Level IPlacebo123/952.5 ± 5.6919Radwa S. et al2023,OctEgyptTherapeutic Level ICorticosteroids306014/1652.17 ± 12.492 weeks, 4 weeks, 6 weeks, 8 weeks, 12 weeks, 4 months④Therapeutic Level IPRP3012/1853.77 ± 12.4420Luciano et al2024,JunArgentinaTherapeutic Level IPRP5010025/2547.6 ± 7.31 month, 3 months, 6 months, 12 months① + ③ + ④Therapeutic Level ICorticosteroids5023/2747.7 ± 7.521Hala M. et al2020,MayEgyptTherapeutic Level IProlotherapy20606/1445.40 ± 15.563 months① + ⑦Therapeutic Level IPRP206/1441.85 ± 10.21Therapeutic Level ICorticosteroids206/1441.85 ± 10.2122Umairet al2022,MarPakistanTherapeutic Level ICorticosteroids306022/855.9 ± 4.33 months① + ④Therapeutic Level IPRP3024/655.2 ± 5.223Mehmet et al2017,MayTurkeyTherapeutic Level IProlotherapy5710123/1950.19 ± 12.133 weeks, 6 weeks, 12 weeks, 1 year① + ② + ⑦Therapeutic Level IPlacebo4419/1640.31 ± 10.6024Ahmed et al2016,AugEgyptTherapeutic Level IPRP204010/1052 ± 126 weeks, 12 weeks, 6 months④ + ⑤ + ⑨Therapeutic Level ICorticosteroids2011/950 ± 1025Yozo et al2001,JunJapanTherapeutic Level IHA387827/1159.5 ± 9.14 weeks①Therapeutic Level ICorticosteroids4028/1262.4 ± 8.626Mehrdad et al2022,FrbIranTherapeutic Level IProlotherapy408017/2145.89 ± 12.221 week, 1 month, 3 months① + ②Therapeutic Level ICorticosteroids4017/2347.38 ± 11.9027Marvin et al2021,SepThailandTherapeutic Level IPRP15313/1251.3 ± 10.31 month, 6 months① + ⑧Therapeutic Level ICorticosteroids163/1362.4 ± 10.528Zhi et al2022,AprChinaTherapeutic Level IPRP5010026/2456.4 ± 9.53 months, 6 months① + ⑤Therapeutic Level ICorticosteroids5029/2155.4 ± 8.729Chang et al2023,JuneChinaTherapeutic Level IProlotherapy255016/942.32 ± 1.651 week, 3 week, 3 months① + ⑪Therapeutic Level ICorticosteroids2514/1145.34 ± 1.25①VAS ②SPADI ③SANE score ④ASES ⑤CMS ⑥Overall shoulder satisfaction ⑦WORC ⑧DASH ⑨SST ⑩OSS ⑪UCLA ⑫SDQ score①VAS: visual analogue scale; ②SPADI: shoulder pain and disability index; ③SANE score: single assessment numeric evaluation score; ④ASES: American shoulder and elbow surgeons score; ⑤CMS: Constant-Murley score; ⑥Overall shoulder satisfaction ⑦WORC: Western Ontario rotator cuff index; ⑧DASH: shoulder and hand score; ⑨SST: simple shoulder test; ⑩OSS: Oxford shoulder score; ⑪UCLA: University of California Los Angeles shoulder score; ⑫SDQ score: shoulder disability questionnaire score Table 1Characteristics of included studiesStudy IDAuthorYearCountryLevel of evidenceGroupGroup peopletotalSex(Male/Female)Patient ageFollow-upOutcome index1Adrien et al2020,JanSwitzerlandTherapeutic Level IPRP418022/1948.2 ± 16.27 months, 12 months① + ③ + ④ + ⑤Therapeutic Level IPlacebo3923/1647.6 ± 11.12Asadollah et al2019,FebIranTherapeutic Level ICorticosteroids295712/1756.3 ± 8.31 month①Therapeutic Level IProlotherapy2813/1559.4 ± 9.23YU et al2018,SepChinaTherapeutic Level IHA448924/2038.93 ± 7.351 month, 3 months, 6 months, 12 months④Therapeutic Level IPRP4522/2340.56 ± 7.854Wen et al2010,JunChinaTherapeutic Level IHA25519/1651.16 ± 7.841 week, 2 weeks, 3 weeks, 4 weeks, 5 weeks, 6 weeks, 11 weeks① + ⑤Therapeutic Level IPlacebo2610/1652.38 ± 8.955Brandi et al2017,AprAustraliaTherapeutic Level IProlotherapy173613/451 ± 166 weeks, 3 months, 6 months⑥Therapeutic Level ICorticosteroids1914/546 ± 156Haleh et al2021,MayIranTherapeutic Level IPRP30585/2557.33 ± 9.801 week, 1 month, 3 months① + ⑦ + ⑧Therapeutic Level ICorticosteroids286/2253.60 ± 7.247Samihah et al2023,MayMalaysiaTherapeutic Level IPRP224417/1557.8 ± 11.53 weeks, 6 weeks, 3 months, 6 months① + ②Therapeutic Level IProlotherapy2219/1351.1 ± 12.68Tirthankaret al2023,NovIndiaTherapeutic Level IPRP5010020/3048 ± 116 weeks, 3 months, 6 months⑤Therapeutic Level ICorticosteroids5022/2845 ± 109Shou et al2022,MarChinaTherapeutic Level IPRP24486/1861.8 ± 2.51, 3 months① + ② + ⑤Therapeutic Level IHA245/1961.5 ± 2.110Doaaet al2018,JulyEgyptTherapeutic Level IPRP15306/946.8 ± 10.62 months① + ⑫Therapeutic Level ICorticosteroids157/841.5 ± 12.511Ilkeret al2015,SepTurkeyTherapeutic Level IPRP30629/2159.16 ± 10.7612 months① + ⑧Therapeutic Level IPlacebo328/2459.68 ± 9.8812Chris et al2020,OctKoreaTherapeutic Level IPRP306011/1955.3 ± 10.31 week, 1 month, 3 months, 6 months① + ④ + ⑤ + ⑨ + ⑪ + ⑫Therapeutic Level ICorticosteroids309/2152.5 ± 11.213Serdar et al2013,JulyTurkeyTherapeutic Level IPRP20407/1345.5 ± 11.83 weeks, 6 weeks, 12 weeks, 24 weeks, 1 year① + ② + ⑦Therapeutic Level IPlacebo206/1451.4 ± 10.014Coryet al2021,FebCanadaTherapeutic Level IPRP479916/3149.94 ± 9.706 weeks, 3 months, 12 months① + ④ + ⑦Therapeutic Level ICorticosteroids5219/3349.08 ± 9.5415Che et al2022,FrbChinaTherapeutic Level IProlotherapy295716/1349.10 ± 8.442 weeks, 6 weeks, 12 weeks① + ②Therapeutic Level IPlacebo2814/1452.18 ± 9.8316Alireza et al2013,DecIranTherapeutic Level IHA20408/1248.14 ± 9.231 week, 2 weeks, 3 weeks, 12 weeks② + ⑤Therapeutic Level IPlacebo206/1417Morteza et al2021,JunIranTherapeutic Level IPlacebo336517/1656.90 ± 13.62 weeks, 3 months②Therapeutic Level IProlotherapy3213/1952.5 ± 13.918Merih et al2010,JulyTurkeyTherapeutic Level IHA12243/958.67 ± 6.323 weeks, 3 months, 4 years①Therapeutic Level IPlacebo123/952.5 ± 5.6919Radwa S. et al2023,OctEgyptTherapeutic Level ICorticosteroids306014/1652.17 ± 12.492 weeks, 4 weeks, 6 weeks, 8 weeks, 12 weeks, 4 months④Therapeutic Level IPRP3012/1853.77 ± 12.4420Luciano et al2024,JunArgentinaTherapeutic Level IPRP5010025/2547.6 ± 7.31 month, 3 months, 6 months, 12 months① + ③ + ④Therapeutic Level ICorticosteroids5023/2747.7 ± 7.521Hala M. et al2020,MayEgyptTherapeutic Level IProlotherapy20606/1445.40 ± 15.563 months① + ⑦Therapeutic Level IPRP206/1441.85 ± 10.21Therapeutic Level ICorticosteroids206/1441.85 ± 10.2122Umairet al2022,MarPakistanTherapeutic Level ICorticosteroids306022/855.9 ± 4.33 months① + ④Therapeutic Level IPRP3024/655.2 ± 5.223Mehmet et al2017,MayTurkeyTherapeutic Level IProlotherapy5710123/1950.19 ± 12.133 weeks, 6 weeks, 12 weeks, 1 year① + ② + ⑦Therapeutic Level IPlacebo4419/1640.31 ± 10.6024Ahmed et al2016,AugEgyptTherapeutic Level IPRP204010/1052 ± 126 weeks, 12 weeks, 6 months④ + ⑤ + ⑨Therapeutic Level ICorticosteroids2011/950 ± 1025Yozo et al2001,JunJapanTherapeutic Level IHA387827/1159.5 ± 9.14 weeks①Therapeutic Level ICorticosteroids4028/1262.4 ± 8.626Mehrdad et al2022,FrbIranTherapeutic Level IProlotherapy408017/2145.89 ± 12.221 week, 1 month, 3 months① + ②Therapeutic Level ICorticosteroids4017/2347.38 ± 11.9027Marvin et al2021,SepThailandTherapeutic Level IPRP15313/1251.3 ± 10.31 month, 6 months① + ⑧Therapeutic Level ICorticosteroids163/1362.4 ± 10.528Zhi et al2022,AprChinaTherapeutic Level IPRP5010026/2456.4 ± 9.53 months, 6 months① + ⑤Therapeutic Level ICorticosteroids5029/2155.4 ± 8.729Chang et al2023,JuneChinaTherapeutic Level IProlotherapy255016/942.32 ± 1.651 week, 3 week, 3 months① + ⑪Therapeutic Level ICorticosteroids2514/1145.34 ± 1.25①VAS ②SPADI ③SANE score ④ASES ⑤CMS ⑥Overall shoulder satisfaction ⑦WORC ⑧DASH ⑨SST ⑩OSS ⑪UCLA ⑫SDQ score①VAS: visual analogue scale; ②SPADI: shoulder pain and disability index; ③SANE score: single assessment numeric evaluation score; ④ASES: American shoulder and elbow surgeons score; ⑤CMS: Constant-Murley score; ⑥Overall shoulder satisfaction ⑦WORC: Western Ontario rotator cuff index; ⑧DASH: shoulder and hand score; ⑨SST: simple shoulder test; ⑩OSS: Oxford shoulder score; ⑪UCLA: University of California Los Angeles shoulder score; ⑫SDQ score: shoulder disability questionnaire score Characteristics of included studies ① + ④ + ⑤ + ⑨ + ⑪ + ⑫ ①VAS: visual analogue scale; ②SPADI: shoulder pain and disability index; ③SANE score: single assessment numeric evaluation score; ④ASES: American shoulder and elbow surgeons score; ⑤CMS: Constant-Murley score; ⑥Overall shoulder satisfaction ⑦WORC: Western Ontario rotator cuff index; ⑧DASH: shoulder and hand score; ⑨SST: simple shoulder test; ⑩OSS: Oxford shoulder score; ⑪UCLA: University of California Los Angeles shoulder score; ⑫SDQ score: shoulder disability questionnaire score Quality evaluation results of included studies The methodological quality of the 29 included randomized controlled trials was evaluated using the RoB 2. Results indicated that 16 studies (55.17%) were rated as low risk, 3 studies (10.34%) as high risk, and the remaining 10 studies (34.48%) as having some concerns. All studies reported randomized allocation, among which 21 (72.41%) specified the method of random sequence generation (e.g., random number table or computer-generated randomization). Regarding blinding, 26 studies (89.66%) blinded participants; however, the 3 high-risk studies failed to blind patients, potentially introducing bias due to psychological expectations (e.g., overestimating therapeutic effects or underestimating improvements), thereby compromising the accuracy of outcome measurements. Additionally, 18 studies (62.07%) implemented blinding for both participants and investigators. All studies demonstrated complete outcome data with no evidence of attrition or selective reporting, resulting in an overall low risk of bias. The detailed results of the methodological quality and risk of bias assessment for all included studies are presented in Fig. 2. Heterogeneity assessment Substantial heterogeneity was observed across studies for both pain (VAS) and functional improvement outcomes at short-, mid-, and long-term follow-ups, with most comparisons exhibiting moderate-to-high heterogeneity (I2 ≥ 50%). To explore potential sources of heterogeneity, a qualitative assessment of study protocols was conducted, focusing on concomitant therapies such as physiotherapy and analgesic use. This approach was employed after concluding that a meta-regression analysis—which we had initially considered—would not be feasible or reliable. The stratification of studies by three follow-up timeframes meant that the number of studies available for each specific treatment comparison was insufficient to support a stable meta-regression model. Our qualitative assessment revealed that, while minor variations in specific exercise protocols were present, a standardized core of physical therapy or a structured home exercise program was consistently implemented across all comparative arms within the vast majority of individual trials. This internal consistency in the application of core co-interventions within each RCT serves to mitigate their role as major confounding factors in the network meta-analysis. Consequently, while the influence of unmeasured variation in adjunctive therapies cannot be entirely discounted, the comparative treatment effects and the resultant hierarchy of interventions are likely to be valid reflections of the efficacy of the injectates themselves. To address this issue and incorporate the heterogeneity into the model, a random-effects model was uniformly applied for all interventional comparisons, as it provides a more conservative estimate of the pooled effect size by accounting for between-study variation. Forest plots detailing heterogeneity patterns across follow-up periods for pain and functional outcomes are provided in Supplementary Fig. S2. Subsequently, we will conduct sensitivity analysis to exclude studies with large heterogeneity, so as to verify the accuracy of the results. Model convergence diagnostics Convergence of the network meta-analysis models was rigorously assessed using trace plots and the PSRF [13]. For all outcomes, trace plots demonstrated stable fluctuations around the posterior means after an initial burn-in period of 50,000 iterations, indicating adequate mixing of Markov chains (Supplementary Fig. S3). The PSRF values for all parameters were ≤ 1.05 (range: 1.01–1.05), well below the recommended threshold of 1.10, confirming satisfactory convergence. Inconsistency tests Global inconsistency tests using the design-by-treatment interaction model revealed no significant disagreement between direct and indirect evidence for any outcomes across follow-up periods (p > 0.05 for all VAS and functional improvement comparisons). Local inconsistency, evaluated via the node-splitting method, demonstrated consistency between direct and indirect estimates for all treatment comparisons. Detailed forest plots comparing direct versus indirect evidence and associated P-values are provided in Supplementary Fig. S4. Network meta-analysis Figure 3a–c present the network plots for the VAS scores in the short-term, medium-term, and long-term, respectively. These figures illustrate the comparative effectiveness of different interventions in reducing pain intensity over varying follow-up periods. Figure 4a–c depict the network plots for the improvement of shoulder joint function in the short-term, medium-term, and long-term, respectively. These figures provide a comprehensive visualization of the relative efficacy of various treatments in enhancing functional outcomes across different timeframes. Fig. 3a VAS(short-term) b VAS(mid-term) c VAS(long-term) a VAS(short-term) b VAS(mid-term) c VAS(long-term) Fig. 4a Functional improvement (short-term). b Functional improvement (mid-term) c functional improvement (long-term). Network plots at different times. The plots (a), (b), and (c) represent these plots at follow-up ≤ 1 month, 1 month< follow-up ≤ 3 months, and follow-up > 3 months, respectively. Each dot represents a different intervention, with larger dots representing that a larger sample size was included. Lines represent direct comparisons, with thicker lines indicating more studies that performed direct comparisons a Functional improvement (short-term). b Functional improvement (mid-term) c functional improvement (long-term). Network plots at different times. The plots (a), (b), and (c) represent these plots at follow-up ≤ 1 month, 1 month< follow-up ≤ 3 months, and follow-up > 3 months, respectively. Each dot represents a different intervention, with larger dots representing that a larger sample size was included. Lines represent direct comparisons, with thicker lines indicating more studies that performed direct comparisons VAS (short-term) A total of 984 participants reported VAS following short-term treatment (follow-up ≤ 1 month) [16–18, 20, 21, 23, 26, 27, 29, 30, 34, 37, 39–41, 43]. Compared to the control group, significant efficacy was observed for prolotherapy (SMD − 1.7, 95% confidence interval [CI] − 2.7 to − 0.74), HA (SMD − 1.6, 95% CI − 2.7 to − 0.60), and corticosteroids (SMD − 1.5, 95% CI − 2.6 to − 0.50) (Fig. 5a). Fig. 5a VAS(short-term). b VAS(mid-term). c VAS(long-term). d functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). Forest plots of comparative efficacy for pain reduction (VAS) and functional improvement across follow-up periods. a Short-term VAS outcomes (follow-up ≤ 1 month); b Mid-term VAS outcomes (1 month < follow-up ≤ 3 months); c Long-term VAS outcomes (follow-up > 3 months); d Short-term functional improvement (follow-up ≤ 1 month); e Mid-term functional improvement (1 month < follow-up ≤ 3 months); f Long-term functional improvement (follow-up > 3 months). Effect sizes are expressed as SMD with 95% confidence intervals (CI), derived from a random-effects network meta-analysis. The vertical line represents the null effect (SMD = 0) a VAS(short-term). b VAS(mid-term). c VAS(long-term). d functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). Forest plots of comparative efficacy for pain reduction (VAS) and functional improvement across follow-up periods. a Short-term VAS outcomes (follow-up ≤ 1 month); b Mid-term VAS outcomes (1 month < follow-up ≤ 3 months); c Long-term VAS outcomes (follow-up > 3 months); d Short-term functional improvement (follow-up ≤ 1 month); e Mid-term functional improvement (1 month < follow-up ≤ 3 months); f Long-term functional improvement (follow-up > 3 months). Effect sizes are expressed as SMD with 95% confidence intervals (CI), derived from a random-effects network meta-analysis. The vertical line represents the null effect (SMD = 0) The SUCRA values and corresponding rankings results of reducing short-term VAS were as follows: Prolotherapy(0.765) > HA(0.690) > Corticosteroids(0.661) > PRP(0.380) > Placebo(0.004) (Table 2). The comparative league table, SUCRA cumulative probability plots, and rank probability plots for short-term VAS outcomes across intra-articular injections are presented in Figs. 6a, 7a and 8a, respectively. Table 2SUCRA values and rankings of interventions for pain and functional improvement across time periodsOutcomePeriodTreatmentSUCRARankingVASShort-termProlotherapy0.7651HA0.6902Corticosteroids0.6613PRP0.3804Placebo0.0045Mid-termProlotherapy0.8751HA0.7082PRP0.4653Corticosteroids0.3584Placebo0.0095Long-termProlotherapy0.8611PRP0.6652Placebo0.3893HA0.3004Corticosteroids0.2885Functional improvementShort-termHA0.6521Prolotherapy0.6052PRP0.6033Placebo0.5284Corticosteroids0.1125Mid-termProlotherapy0.9871Corticosteroids0.5962PRP0.3963Placebo0.3174HA0.2045Long-termHA0.8711Prolotherapy0.7452Corticosteroids0.4013Placebo0.2504PRP0.2335SUCRA values and corresponding rankings for each intervention in pain reduction (VAS) and functional improvement across short-term (follow-up ≤ 1 month), mid-term (1 month < follow-up ≤ 3 months), and long-term (follow-up > 3 months) periods. Higher SUCRA values (0–1) indicate a higher probability of being the optimal intervention. The ranking (1–5) represents the relative effectiveness, with rank 1 being the best Table 2SUCRA values and rankings of interventions for pain and functional improvement across time periodsOutcomePeriodTreatmentSUCRARankingVASShort-termProlotherapy0.7651HA0.6902Corticosteroids0.6613PRP0.3804Placebo0.0045Mid-termProlotherapy0.8751HA0.7082PRP0.4653Corticosteroids0.3584Placebo0.0095Long-termProlotherapy0.8611PRP0.6652Placebo0.3893HA0.3004Corticosteroids0.2885Functional improvementShort-termHA0.6521Prolotherapy0.6052PRP0.6033Placebo0.5284Corticosteroids0.1125Mid-termProlotherapy0.9871Corticosteroids0.5962PRP0.3963Placebo0.3174HA0.2045Long-termHA0.8711Prolotherapy0.7452Corticosteroids0.4013Placebo0.2504PRP0.2335SUCRA values and corresponding rankings for each intervention in pain reduction (VAS) and functional improvement across short-term (follow-up ≤ 1 month), mid-term (1 month < follow-up ≤ 3 months), and long-term (follow-up > 3 months) periods. Higher SUCRA values (0–1) indicate a higher probability of being the optimal intervention. The ranking (1–5) represents the relative effectiveness, with rank 1 being the best SUCRA values and rankings of interventions for pain and functional improvement across time periods SUCRA values and corresponding rankings for each intervention in pain reduction (VAS) and functional improvement across short-term (follow-up ≤ 1 month), mid-term (1 month < follow-up ≤ 3 months), and long-term (follow-up > 3 months) periods. Higher SUCRA values (0–1) indicate a higher probability of being the optimal intervention. The ranking (1–5) represents the relative effectiveness, with rank 1 being the best Fig. 6a VAS(short-term). b VAS(mid-term). c VAS(long-term). d Functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). League tables of pairwise comparisons for VAS and functional improvement across follow-up periods. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short, mid, and long-term). Values represent SMD with 95% CI from direct and indirect comparisons. Upper diagonal: SMD (95% CI) for row-defining interventions versus column-defining interventions; lower diagonal: reciprocal comparisons a VAS(short-term). b VAS(mid-term). c VAS(long-term). d Functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). League tables of pairwise comparisons for VAS and functional improvement across follow-up periods. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short, mid, and long-term). Values represent SMD with 95% CI from direct and indirect comparisons. Upper diagonal: SMD (95% CI) for row-defining interventions versus column-defining interventions; lower diagonal: reciprocal comparisons Fig. 7a VAS(short-term). b VAS(mid-term). c VAS(long-term). d functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). SUCRA cumulative probability plots for intervention rankings. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short-, mid-, and long-term). Higher SUCRA values (0–100%) indicate a greater probability of being the optimal intervention. Colored bands represent cumulative probabilities for each treatment a VAS(short-term). b VAS(mid-term). c VAS(long-term). d functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). SUCRA cumulative probability plots for intervention rankings. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short-, mid-, and long-term). Higher SUCRA values (0–100%) indicate a greater probability of being the optimal intervention. Colored bands represent cumulative probabilities for each treatment Fig. 8a VAS(short-term). b VAS(mid-term). c VAS(long-term). d functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). Rank probability plots for intervention hierarchies across outcomes and timeframes. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short-, mid-, and long-term). Bar heights represent the probability of each intervention occupying a specific rank (1 = best, 5 = worst). Colored bands represent cumulative probabilities for each treatment a VAS(short-term). b VAS(mid-term). c VAS(long-term). d functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). Rank probability plots for intervention hierarchies across outcomes and timeframes. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short-, mid-, and long-term). Bar heights represent the probability of each intervention occupying a specific rank (1 = best, 5 = worst). Colored bands represent cumulative probabilities for each treatment VAS (mid-term) A total of 990 participants were included in the assessment of mid-term VAS (1 month< follow-up ≤ 3 months) [17, 18, 20, 21, 23, 24, 26–28, 34–37, 42, 43]. Compared to Corticosteroids (SMD − 0.81, 95% confidence interval [CI] − 2.6 to 1.1) and PRP (SMD − 0.97, 95% confidence interval [CI] − 2.6 to 0.77), the Prolotherapy (SMD − 1.8, 95% confidence interval [CI] − 3.7 to − 0.13) and HA (SMD − 1.5, 95% confidence interval [CI] − 3.3 to 0.44) treatment groups demonstrated a more significant reduction in mid-term VAS among patients (Fig. 5b). The SUCRA values and corresponding rankings results of reducing mid-term VAS were as follows: Prolotherapy(0.875) > HA(0.708) > PRP(0.465) > Corticosteroids(0.358) > Placebo(0.009)(Table 2). The comparative league table, SUCRA cumulative probability plots, and rank probability plots for mid-term VAS outcomes across intra-articular injections are presented in Figs. 6b, 7b and 8b, respectively. VAS (long-term) In the long-term follow-up assessment of VAS (follow-up >3 months), 863 patients reported their outcomes [15, 17, 21, 25–29, 34, 37, 41, 42]. Prolotherapy (SMD − 1.2, 95% confidence interval [CI] − 2.9 to 0.52) and PRP (SMD − 0.49, 95% confidence interval [CI] − 2.0 to 0.99) demonstrated superior long-term pain relief efficacy, whereas no significant difference was observed between Corticosteroids (SMD 0.24, 95% confidence interval [CI] − 1.7 to 2.2)/HA (SMD 0.43, 95% confidence interval [CI] − 2.7 to 3.5)and Placebo in long-term pain reduction (Fig. 5c). The SUCRA values and corresponding rankings results of reducing long-term VAS were as follows: Prolotherapy(0.861) > PRP(0.665) > Placebo(0.389) > HA(0.300) > Corticosteroids(0.288)(Table 2). The comparative league table, SUCRA cumulative probability plots, and rank probability plots for long-term VAS outcomes across intra-articular injections are presented in Figs. 6c, 7c and 8c, respectively. Improvement in shoulder function (short-term) Data from 1142 patients indicated short-term shoulder joint functional improvement (follow-up ≤ 1 month) based on standardized assessment scales [17, 18, 20, 21, 23, 26, 27, 29, 31, 33, 34, 37, 40, 41]. HA (SMD − 0.22, 95% confidence interval [CI] − 2.1 to 1.6) showed moderate efficacy in short-term functional outcomes, whereas no statistically significant differences were found between Prolotherapy (SMD − 0.12, 95% confidence interval [CI] − 1.5 to 1.3), PRP (SMD − 0.075, 95% confidence interval [CI] − 1.6 to 1.5), and placebo. Notably, Corticosteroids (SMD 0.76, 95% confidence interval [CI] − 0.89 to 2.4) demonstrated lower efficacy compared to placebo in this domain (Fig. 5d). In summary, HA(SUCRA = 0.652) appeared to be the most effective intervention for short-term shoulder joint functional improvement (follow-up ≤ 1 month), followed by Prolotherapy(SUCRA = 0.605), PRP(SUCRA = 0.603), Placebo (SUCRA = 0.528) ,Corticosteroids(SUCRA = 0.112). (Table 2; Figs. 6d, 7d and 8d). Improvement in shoulder function (mid-term) A cohort of 1258 patients participated in the mid-term follow-up assessment (1 month< follow-up ≤ 3 months) of shoulder joint functional outcomes [17–24, 26–28, 30, 33–38, 42, 43]. The Prolotherapy groups (SMD − 1.2, 95% confidence interval [CI] − 2.1 to − 0.2) demonstrated significantly greater improvement in functional recovery during this period, while Corticosteroids (SMD − 0.29, 95% confidence interval [CI] − 1.1 to 0.58), PRP (SMD − 0.13, 95% confidence interval [CI] − 0.96 to 0.73), and HA (SMD 0.15, 95% confidence interval [CI] − 0.91 to 1.2) exhibited limited efficacy, with no significant differences observed compared to the control group (Fig. 5e). All in all, Prolotherapy(SUCRA = 0.987) appeared to be the most effective intervention for mid-term shoulder joint functional improvement (1 month< follow-up ≤ 3 months), followed by Corticosteroids(SUCRA = 0.596), PRP(SUCRA = 0.396), Placebo(SUCRA = 0.317),HA(SUCRA = 0.204). (Table 2; Figs. 6e, 7e and 8e). Improvement in shoulder function (long-term) Long-term shoulder joint functional outcomes (follow-up >3 months) were evaluated in a cohort of 1146 patients [15, 17, 19, 21, 22, 25–30, 33, 34, 37, 38, 41, 42]. Standardized functional assessment scales revealed that both HA and Prolotherapy treatment groups demonstrated superior improvement in functional recovery during this period compared to PRP and Corticosteroids. Specifically, HA (SMD − 1.7, 95% confidence interval [CI] − 4.1 to 0.65) and Prolotherapy (SMD − 1.0, 95% confidence interval [CI] − 2.4 to 0.38) were associated with significantly better functional outcomes, whereas PRP (SMD − 0.0081, 95% confidence interval [CI] − 1.1 to 1.0) and Corticosteroids (SMD − 0.21, 95% confidence interval [CI] − 1.4 to 0.96) showed no statistically significant advantage over baseline control measurements (Fig. 5f). In conclusion, HA(SUCRA = 0.871) appeared to be the most effective intervention for long-term shoulder joint functional improvement (follow-up > 3 months), followed by Prolotherapy(SUCRA = 0.745), Corticosteroids(SUCRA = 0.401), Placebo(SUCRA = 0.250), PRP(0.233). (Table 2; Figs. 6f, 7f and 8f). Publication bias Publication bias was assessed through visual inspection of funnel plots and quantified using Egger’s linear regression test. For all outcomes (VAS and functional improvement across short-, mid-, and long-term follow-ups), funnel plots demonstrated symmetrical distributions of study effect sizes around the pooled estimates, with no evidence of missing studies in the lower or upper quadrants. Egger’s test further confirmed the absence of significant small-study effects, with p-values ranging from 0.4556 to 0.8782, well above the significance threshold of 0.05. For example, mid-term VAS outcomes yielded an Egger’s test p-value of 0.4902, while long-term functional improvement showed p = 0.8372. These findings collectively suggest minimal risk of publication bias, supporting the robustness of the synthesized evidence ( Fig. 9a–f) [44]. Fig. 9a VAS(short-term). b VAS(mid-term). c VAS(long-term). d Functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). Funnel plots and Egger’s regression tests. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short-, mid-, and long-term). Egger’s regression test (p > 0.05) indicated no significant publication bias a VAS(short-term). b VAS(mid-term). c VAS(long-term). d Functional improvement(short-term). e Functional improvement(mid-term). f Functional improvement(long-term). Funnel plots and Egger’s regression tests. a–c VAS outcomes (short-, mid-, and long-term); d–f Functional improvement outcomes (short-, mid-, and long-term). Egger’s regression test (p > 0.05) indicated no significant publication bias Sensitivity analysis To test the stability and reliability of the NMA results, we performed sensitivity analyses for short-, mid-term and long-term VAS and functional improvement. First, 3 papers that were evaluated as high-risk in terms of literature quality were excluded [22, 23, 32], and sensitivity analyses were performed before and after exclusion. Second, as RCTs with small sample sizes may have affected the accuracy of the results, 4 studies with sample sizes of < 40 were excluded from the sensitivity analysis [19, 24, 32, 41]. The results showed little difference between the results before and after the exclusion of the sensitivity analysis, indicating that the quality of the literature was good and that the results of the network meta-analysis were solid and stable. The results of the sensitivity analysis are shown in Supplementary Figs. S5a, b, S6a, b. Discussion Network meta-analysis revealed distinct temporal patterns in the efficacy of intra-articular interventions for both pain relief and functional improvement in RCI. Specifically, while prolotherapy, HA, and corticosteroids demonstrated comparable efficacy in short-term pain reduction (≤ 1 month), prolotherapy and HA emerged as superior to other treatments for pain alleviation in the mid-term phase (1–3 months). Notably, long-term pain relief (> 3 months) was most effectively sustained by prolotherapy and PRP. Functional recovery exhibited a similar time-dependent profile: HA provided the greatest benefit in early functional restoration (≤ 1 month), prolotherapy showed the strongest efficacy for mid-term functional outcomes (1–3 months), and both HA and prolotherapy demonstrated statistically superior long-term functional improvement (> 3 months) compared to corticosteroids and PRP. The temporal efficacy patterns observed in this analysis could potentially be explained by mechanisms that are theorized to align with distinct tissue repair stages. Preclinical evidence suggests that the early benefits of prolotherapy may derive, in part, from osmotic stress triggering a localized repair response [45], while those of HA are frequently attributed to its lubricating and potential anti-inflammatory properties [46]. Corticosteroids are widely regarded as effective for short-term analgesia, primarily through their anti-inflammatory action [47], though some evidence indicates they might inadvertently impede longer-term healing processes. The long-term efficacy of PRP is suggested to involve tissue regeneration mediated by growth factors [48]. This process may rely on the capacity of concentrated platelets to orchestrate a coordinated healing response through the release of a complex network of signaling proteins that regulate key processes such as inflammation, angiogenesis, and matrix synthesis [49, 50]. Regarding functional improvement, early HA effects have been linked to improved lubrication [51], mid-term prolotherapy benefits are hypothesized to result from stimulated collagen deposition [52]. Long-term outcomes are potentially supported by HA’s metabolic regulation [53] alongside PRP’s postulated regenerative capacity [54]. Collectively, these proposed mechanisms offer a theoretical framework that may provide a rationale for temporal treatment selection, though this requires further validation. When interpreting these findings and their clinical applicability, several contextual factors regarding the evidence base merit consideration. The overall strength of the conclusions is inevitably shaped by the methodological characteristics of the included trials. Some degree of heterogeneity was observed, which may reflect variations in clinical practice, such as differences in injection technique or concomitant rehabilitation protocols. Additionally, while the geographic distribution of the included studies reflects the global research landscape for this intervention, it invites consideration of the generalizability of the results across diverse healthcare settings and patient populations. These aspects do not invalidate the observed temporal patterns but highlight that the proposed framework represents the most robust synthesis of current evidence, which future research can further refine and validate. The SUCRA rankings offer a preliminary hierarchy to guide clinical decision-making, while their interpretation is enriched by considering the confidence intervals. For instance, the long-term VAS comparison between HA and placebo (SMD 0.43, 95% CI − 2.7 to 3.5) highlights that the precise effect size for some comparisons remains to be more precisely estimated in future studies. This reflects the current state of evidence and underscores the value of our analysis in pinpointing where future research is most needed to refine these estimates. The rankings should therefore be viewed as a robust indicator of the relative potential of treatments, with the understanding that their exact order will be refined as the evidence base matures. This temporal, efficacy-based framework provides a foundational strategy for personalizing injection therapy, which can be adapted as more precise data become available. Supported by the observed temporal efficacy patterns, a stepwise approach to intra-articular injection therapy for RCI may be proposed as a hypothesis for future investigation: For patients seeking rapid symptom relief in the acute phase (≤ 1 month), prolotherapy or HA could be considered favorable initial options, offering significant early pain reduction and, in the case of HA, functional improvement. Prolotherapy might be particularly advantageous for continued management into the mid-term phase (1–3 months), where it demonstrated the highest efficacy for both pain relief and functional recovery in this analysis. For sustained long-term (> 3 months) pain control, PRP emerges as a candidate based on its observed superiority in this timeframe. HA also showed notable long-term functional benefits. While corticosteroids provide effective short-term analgesia, their diminishing returns over time and potential inhibitory effects on tissue repair suggest their use might be reserved for scenarios where immediate pain relief is paramount and other options are unsuitable. Ultimately, this proposed temporal selection algorithm should be integrated with individual patient factors such as symptom severity, functional goals, comorbidities, and personal preferences. This framework warrants validation in prospective studies specifically designed to test sequential or combination therapies tailored to these distinct phases. Our results align with certain recommendations from current guidelines and meta-analyses, yet al.so provide critical refinements. The latest Rotator Cuff Tendinopathy Diagnosis, Nonsurgical Medical Care, and Rehabilitation: A Clinical Practice Guideline indicates that corticosteroids should only be used for short-term pain and functional improvement in rotator cuff disorders, while acknowledging that HA can alleviate pain and improve function in the short to medium term, and that PRP is a first-line option with certain effects on pain and function [55]. Our network meta-analysis refines these recommendations by revealing a distinct temporal efficacy pattern: PRP demonstrates superior long-term pain relief, which corroborates the findings of the pairwise meta-analysis by Mohamad et al. [56]. More importantly, our analysis critically expands the current paradigm by identifying prolotherapy—a modality not sufficiently covered in major guidelines or prior large-scale non-trial syntheses—as a consistently top-ranking intervention across multiple timeframes. Furthermore, although HA is often associated with osteoarthritis, our data underscore its significant value in improving both short- and long-term functional outcomes in rotator cuff injuries. Thus, while our study confirms the roles of corticosteroids and PRP as established in the highest levels of evidence, it also challenges and extends the current therapeutic hierarchy by introducing prolotherapy and HA as key considerations for a stage-adapted treatment strategy. When integrating these efficacy findings into clinical practice, the safety profile and potential adverse events across different injectates must also be considered to ensure a balanced risk-benefit assessment. Our qualitative review of the included studies indicated that serious adverse events were rare across all intervention types. The most commonly reported events were transient, minor reactions such as post-injection pain and localized inflammation. While all intra-articular injections carry a baseline risk of infection or procedural complication, the overall safety data support that the efficacy rankings discussed herein are not offset by major differential safety concerns. Beyond efficacy and safety, the translation of these findings into routine clinical practice is influenced by significant economic and health-system considerations. The cost of injectates varies substantially; corticosteroids are generally widely accessible and low-cost, whereas HA and particularly PRP are associated with higher direct costs due to product preparation and acquisition. Furthermore, accessibility is not uniform, as coverage for PRP and HA injections by national or private insurance plans differs across healthcare systems and regions, potentially creating disparities in patient access. These economic barriers may be compounded by broader social determinants of health. A recent study by Timoteo et al. demonstrated that lower socioeconomic status is independently associated with worse pain and functional outcomes after rotator cuff repair [57], highlighting that financial constraints and neighborhood-level deprivation can significantly influence treatment efficacy and recovery. Therefore, while our analysis identifies PRP and HA as high-performing options for specific long-term goals, the cost-benefit ratio from both a healthcare system and a patient out-of-pocket perspective must be evaluated in local contexts. The compelling efficacy of prolotherapy, which often utilizes low-cost dextrose, presents a potentially cost-effective alternative in settings where resource constraints are a primary concern. These economic and socioeconomic realities are crucial for clinicians and policymakers when prioritizing treatment options within specific healthcare environments. Limitations The interpretation of this study’s findings requires careful consideration of its limitations. The inclusion of high-quality studies notwithstanding, differences in adjunctive therapies (including anti-inflammatory medications and physical therapy regimens) between the study cohorts may have confounded the observed treatment effects. Notably, beyond the technical aspects of preparation, the biological activity of PRP is intrinsically linked to platelet function. Future research should consider patient-specific factors, such as concomitant use of medications known to modulate platelet activity (e.g., NSAIDs, anticoagulants) [58], as these represent a potential and important source of heterogeneity in treatment response and a key to personalizing regenerative therapies [59]. Furthermore, the substantial out-of-pocket cost of PRP may contribute to response bias in unblinded patients, potentially influencing patient-reported outcomes; this effect could be compounded by the lack of routine coverage for biologic agents like PRP under most government or private insurance plans. It is important to note that our study did not incorporate formal cost-benefit analysis. Furthermore, HA injections may only be approved by certain insurances at specific time intervals. The length of efficacy of each of these injections warrants further study, and an ongoing issue in the literature pertains to the volume and formulation of various injections, particularly for PRP, as well as their associated dosing schedule. For example, multiple HA injections available on the market may require up to 3 injections staged over multiple weeks. Similarly, the impact of different platelet and leukocyte concentrations among PRP injections and outcomes continues to be evaluated. Furthermore, while the included studies spanned several continents (Asia, Europe, Africa, America), a substantial proportion of the RCTs originated from Asian populations (e.g., China, Iran, Japan, Malaysia, Pakistan, Thailand). Although rotator cuff pathology is a global concern, potential variations in genetic predisposition, body habitus, activity levels, healthcare access, or cultural perceptions of pain and function could theoretically influence treatment responses. While our analysis did not identify overt geographical heterogeneity signals within the available data, the relative underrepresentation of certain geographic regions (particularly North America and Western Europe) in the pooled sample suggests that the observed efficacy patterns should be interpreted with some caution across diverse global populations. Future large-scale, multi-regional trials would be valuable to confirm the robustness of these findings across broader ethnic and geographical demographics. Notably, while range of motion (ROM) is an objective measure to reflect functional improvement across different phases, we initially planned to conduct a network meta-analysis on ROM outcomes; However, insufficient data from the included studies precluded this analysis. In future research, it is essential to include a sufficient number of ROM outcome data across different timeframes to enable a more systematic evaluation of the dynamic recovery of shoulder joint function. Conclusions This network meta-analysis demonstrates distinct temporal efficacy patterns for intra-articular injections in partial-thickness rotator cuff injuries. The findings advocate for a shift in clinical paradigm from relying on a single injectate throughout the care continuum to adopting a tailored strategy based on treatment timeframes and specific patient goals. Based on this network meta-analysis, the following evidence-informed recommendations can be formulated to guide clinical practice: Prolotherapy and HA are optimal for early-stage pain control and functional recovery, while PRP provides superior sustained long-term pain relief. HA and prolotherapy show advantages for long-term functional improvement. Corticosteroids are effective for short-term analgesia but should be used judiciously given limited long-term benefits and potential tissue effects. To validate and refine this proposed framework, future multicenter trials should move beyond simple head-to-head comparisons of single injections. The most critically needed studies are those investigating sequential treatment strategies, which directly test the temporal paradigm by comparing, for instance, an initial course of prolotherapy followed by PRP booster injections against a single-modality approach. Furthermore, pragmatic trials conducted in diverse healthcare settings are essential to confirm the real-world effectiveness and cost-effectiveness of this time-stratified approach. These studies must prioritize the standardization of critical variables—including injection techniques (e.g., ultrasound guidance), rehabilitation protocols, and PRP preparation methods—to minimize heterogeneity and pinpoint the true effect of the injectates themselves. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1