The clinical efficacy of digital health interventions in patients with polycystic ovary syndrome: a systematic review and meta-analysis Authors: Min Xie, Xuemei Qing, Hailong Huang, Bangfang Xie, Jing Zhang Published: 2025 ABSTRACT AbstractBackgroundPolycystic Ovary Syndrome (PCOS) is the most prevalent metabolic and endocrine disorder among women of reproductive age. Long-term management is crucial, yet the poor compliance and low sustainability of lifestyle interventions limit their effectiveness. Digital health interventions have demonstrated the potential to improve patient compliance. Therefore, this systematic review and meta-analysis evaluates the efficacy of digital health interventions in improving PCOS management.MethodsWe conducted a systematic review and meta-analysis of randomized controlled trials (RCTs) on digital health interventions for PCOS by searching databases from inception until November 2025. The quality of included studies was evaluated using the Cochrane Risk of Bias Tool, and publication bias was examined with funnel plots, Egger’s test, and Trim and Fill method. We used RevMan 5.4 software for the meta-analysis to evaluate the efficacy of these interventions. Subgroup analyses based on intervention duration and outcome measurement tools were performed to further clarify their benefits for PCOS patients.ResultsTwenty-five RCTs were included. The meta-analysis demonstrated that digital health interventions significantly improved treatment compliance (P = 0.007), compliance behavior (P < 0.05), quality of life (P = 0.04), and healthy lifestyle modifications-including dietary rhythm, sports consciousness, dietary content, and rhythm of life (all P < 0.05). Significant reductions were observed in body mass index (P = 0.02), waist circumference (P < 0.00001), serum testosterone (P = 0.002), anxiety (P < 0.00001), and depressive symptoms (P < 0.00001). However, no significant differences were observed in luteinizing hormone (P = 0.68), follicle-stimulating hormone levels (P = 0.97), or body awareness of lifestyle changes (P = 0.24). Subgroup analysis indicated that digital health interventions significantly improved body mass index after ≥ 6 months (P < 0.05) and effectively alleviated anxiety and depression regardless of intervention duration.ConclusionDigital health interventions represent a valuable adjunct in the long-term management of PCOS, demonstrating efficacy across clinical, behavioral, and psychological domains.Systematic review registrationRegistration number: CRD 420,251,002,149.Supplementary InformationThe online version contains supplementary material available at 10.1186/s41043-025-01167-9.Keywords: Digital health intervention, Polycystic ovary syndrome, Compliance, Meta-analysis Introduction Polycystic ovary syndrome (PCOS) is the most common endocrine-metabolic disorder among reproductive-aged women. It is characterized by oligomenorrhea or amenorrhea, ovulatory dysfunction, hyperandrogenemia (HA), and polycystic ovarian morphology (PCOM), and frequently co-occurs with obesity, insulin resistance (IR), dyslipidemia, and infertility [1]. Additionally, PCOS is also associated with an elevated risk of depression and anxiety, which impair both quality of life and fertility. Furthermore, the condition can progress to long-term complications including abnormal uterine bleeding, type 2 diabetes, endometrial carcinoma, and cardiovascular disease, posing lifelong health risks [2]. With a global prevalence of 5%–20%, PCOS poses substantial health and economic burdens worldwide [3]. Its rising prevalence, driven by increased awareness, improved diagnostics, and lifestyle changes, shows regional variation-for instance, in China, rates increased from 5.6% (2007–2010) to 7.8% in 2022 [4, 5]. PCOS is a multisystem disorderdisorder whose pathogenesis, though not fully elucidated, centers on IR and HA. According to the 2023 International Evidence-Based PCOS Guideline, diagnosis in adults requires at least two of the following criteria after excluding related disorders: (1) clinical/biochemical hyperandrogenism; (2) ovulatory dysfunction; (3) polycystic ovarian morphology or elevated anti-Müllerian hormone [6]. PCOS is a leading cause of reproductive dysfunction, accounting for 30–40% of female infertility cases, with 80% of affected women experiencing infertility [7]. Obesity, present in approximately 90% of infertile PCOS patients, independently predicts infertility by exacerbating anovulation and metabolic disturbances, making it a key therapeutic target [8]. Psychologically, PCOS patients face a fourfold increased risk of anxiety and depression, with reported prevalence rates of 37% and 43%, respectively [9, 10], underscoring the need for integrated mental health care. A core pathophysiological feature, IR, affects about 75% of patients and represents a critical treatment target [11, 12]. While no cure exists, management focuses on ameliorating IR, HA, and restoring ovulation [13]. The 2023 International Evidence-Based Management Guidelines for PCOS advocate lifestyle interventions as a primary treatment option for PCOS, particularly in cases of overweight or obese PCOS [6]. It is hypothesized that moderate weight loss (only 5%) will improve both the metabolic and reproductive problems associated with PCOS [14]. However, the efficacy of lifestyle interventions is often limited by poor patient compliance and high dropout rates, which compromise short-term outcomes and increase the risk of long-term recurrence [15]. his is substantiated by studies reporting a wide variation in adherence rates among PCOS patients, ranging from 21.7% to 86.0% [16]. Consequently, enhancing adherence represents a critical objective in PCOS management. In this context, digital health interventions emerge as a transformative approach—leveraging their accessibility, cost-effectiveness, and user-friendly design to deliver behavioral counseling and motivate sustained lifestyle modification effectively [17]. Digital health interventions are defined as all healthcare services delivered using digital technologies for the prevention, management, or treatment of disease [18]. Digital health interventions utilize devices like phones, apps, and web platforms to support patient education and management. By providing interactive features like reminders and peer support, these interventions enhance treatment adherence and exert synergistic benefits on health outcomes [19, 20]. A previous meta-analysis suggested the potential benefits of Internet-based continuity of care for patients with PCOS, reporting improvements in follicle-stimulating hormone(FSH)levels, medication adherence, and pregnancy rates compared to traditional face-to-face care [21]. However, the generalizability and robustness of these findings are subject to notable limitations. Specifically, the included literature was exclusively in Chinese, potentially introducing language and regional bias, and the small number of available studies limited the statistical power. More critically, the significant heterogeneity observed among the included studies compromises the reliability and consistency of the pooled results, making it difficult to draw definitive conclusions. To address these gaps in the existing evidence, we conducted an updated and more comprehensive systematic review and meta-analysis of RCTs. This study aims to rigorously compare the efficacy of digital health interventions with conventional treatment for PCOS management, thereby providing a more robust clinical rationale for long-term management strategies. Methods This systematic review and meta-analysis was designed and conducted following a predetermined protocol according to the Cochrane Handbook’s recommendations [22]. We reported the results using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), a systematic review guideline [23]. Before data extraction, the systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) database (CRD 420251002149). Search strategies We conducted a systematic literature search in PubMed, Embase, Medline, Cochrane Library, China National Knowledge Infrastructure (CNKI), and Wanfang Database from inception until November 2025. The search strategy combined keywords and controlled vocabulary terms (MeSH/Emtree) for the core concepts of “polycystic ovary syndrome” (including “PCOS”) and “digital health” (including “mHealth”, “telemedicine”, and “mobile applications”), with no restrictions on language or publication date. The complete search syntax for each database is available in Supplementary File 1. Additionally, we manually screened the reference lists of included articles and relevant reviews, along with grey literature, to identify further eligible studies. Study selection was performed independently by two reviewers, with any disagreements resolved through consensus or by consulting the corresponding author. Eligibility criteria The following inclusion criteria were applied in the course of the systematic review: Population(P): Adult women aged ≥ 18 years with a clinically confirmed diagnosis of PCOS; Intervention(I): Any type of digital health intervention; Comparator(C): Conventional therapy or standard care; Outcomes (O): All reported clinical, behavioral, and metabolic outcomes were considered; Study design (S): Randomized controlled trials. The following exclusion criteria were applied: conference abstracts, animal experiments, cohort studies, retrospective studies, non-randomized controlled studies; literature with non-compliant interventions; literature with non-compliant outcome metrics; literature with inaccessible full text and duplicated publications; literature with unanalysable data; and literature not in English or Chinese. Study selection criteria We conducted the initial search, removed duplicate records. The study selection was conducted independently by two authors (MX and XQ), who screened titles/abstracts and subsequently assessed full-text articles against the eligibility criteria. Disagreements were resolved through consensus or by consulting a third senior author (JZ). To ensure consistency and quantify inter-rater agreement, we calculated Cohen’s Kappa (κ) statistic based on both authors’ independent screening of records. Data extraction Data extraction was conducted independently by two authors, who meticulously reviewed each included document. Thereafter, the data were meticulously cross-checked to minimize potential errors, and any disagreements were resolved through discussion with the corresponding author. The Cochrane guidelines were used to extract the characteristics of the included studies. The following information was extracted from the included studies: author name, country, year of publication, sample size, description of intervention, description of control, duration, and primary and secondary outcomes. For continuous variables, the mean and standard deviation (SD) for the post-intervention values were extracted. In instances where a specified mean difference (MD) was not available, this was derived from either the standard error, the interquartile range, or the 95% confidence interval. The integrity of the data was maintained throughout the process, and any missing data was obtained through email or telephone communication with the authors. In the event of non-response, the study will be excluded. Data synthesis and analysis All relevant statistical analyses were performed using the software Review Manager 5.4. Continuous variables were assessed using standardized mean difference (SMD) and 95% confidence interval (CI). Dichotomous variables were assessed using the ratio of ratios (OR) and its 95% confidence interval (CI) as an effect size indicator. Comparative pre- and post-treatment analyses were performed in all included studies; however, only four studies reported results for multiple outcome detection time points. Consequently, data from the final time point were utilized for analysis. A P-value of less than 0.05 was used to represent a statistically significant difference. The heterogeneity between studies was estimated using I2. In each analysis, heterogeneity was classified as low (I2 ≤ 40%), moderate (40% ≤ I2 ≤ 70%), or high (I2 > 70%). When I2 was less than 50%, a fixed-effects model was employed; when I2 was greater than 50%, a random-effects model was selected for the analysis. The presence of a time-cumulative effect of the digital health intervention was identified, and consequently, we conducted a series of subgroup analyses. These analyses were stratified according to the duration of the study intervention (≥ 6 months or < 6 months, unspecified, or 3 months), and the type of anxiety and depression symptoms assessment scales (Anxiety and Depression Self-Assessment Scale, Hamilton Anxiety and Depression Scale, Trait Anxiety and Depression Scale, Self-Assessment Scale, or Hospital Anxiety and Depression Scale, etc.). The forest plots were used to represent the results of the statistical analyses, and the stability of the meta-analysis results was verified by employing sensitivity analyses. Finally, publication bias was assessed for outcomes with 10 or more studies using funnel plots, Egger’s linear regression test (p < 0.05 indicating significance), and the non-parametric Trim and Fill method. For outcomes with fewer studies, these tests were deemed unreliable, and reliance was placed on the comprehensive search to minimize bias. Assessment of risk of bias and evidence quality The quality assessment was conducted by two independent researchers using the Cochrane Collaboration’s risk of bias tool for RCTs. The analysis was subjected to meticulous scrutiny, and any discrepancies were resolved through consultation with a third researcher. The Cochrane Risk Assessment Tool has been utilized to evaluate the quality at the study level. The assessment encompassed various domains, including random sequence generation, allocation concealment, blinding methods, incomplete outcome data, selective outcome reporting, and other potential sources of bias. Each domain was meticulously assessed and assigned a verdict of “yes,” “no,” or “unclear” based on The Cochrane Collaboration’s established standards. The risk-of-bias judgments were then assigned to one of three categories: “unclear,” “low,” or “high” risk of bias. The Grading of Recommendations, Assessment, Development, and Evaluation, a novel approach, was used to evaluate the quality level of evidence for important clinical outcomes for five domains (risk of bias, inconsistency, imprecision, indirectness, and publication bias). The quality level of evidence was downgraded by one or two levels depending on seriousness of the limitations. Evaluation of evidence quality (high, moderate, low, or very low) was determined by two review authors independently, with differences resolved by discussion [24]. Results Studies selection and the flow chart The systematic search yielded an initial corpus of 1,775 records from six electronic databases: PubMed (n = 83), Embase (n = 183), Cochrane Library (n = 485), Medline (n = 156), CNKI (n = 263), and Wanfang (n = 605), supplemented by one manually identified reference. Following the removal of 347 duplicates, 1,428 unique records were screened at the title/abstract level, resulting in the exclusion of 1,358 studies. Of the 70 articles subjected to full-text review, 45 were excluded due to non-randomized design (n = 18), protocol deviations (n = 7), inaccessibility of the full text (n = 8), or incomplete outcome reporting (n = 12). The final synthesis included 25 RCTs. The entire selection workflow, presented in Fig. 1, was conducted in accordance with the PRISMA statement. PRISMA flow diagram of the study process. PRISMA, Preferred Reporting Items for Systematic Review and Meta-analysis The agreement between the two independent authors during the study selection process was excellent. The Cohen’s Kappa statistic was κ = 0.88 for the title/abstract screening phase and κ = 0.92 for the full-text assessment phase, indicating a high level of consistency in the application of the inclusion and exclusion criteria. Characteristics of included studies Table 1 summarizes the characteristics of the 25 included studies. All were RCTs published between 2016 and 2025. Geographically, 22 studies were conducted in China [25–46], two in Iran [47, 48], and one in Korea [49]. The analysis pooled data from 2746 PCOS patients (aged 18–50 years), with 1430 assigned to digital health interventions and 1416 to control groups receiving routine care. Digital interventions were delivered through two primary modalities: mobile applications (17 studies), which featuring self-monitoring (diet, physical activity, mood), educational content, and personalized goal-setting; and web platforms (8 studies), which provided structured educational programs with integrated peer support forums. Intervention frequency ranged from daily (e.g., app prompts) to weekly (e.g., educational modules, coaching sessions). Control groups uniformly received conventional care consisting of standard face-to-face consultations and general lifestyle advice, while all participants continued their baseline PCOS treatments. All studies reported outcomes measured at baseline and post-treatment, with four studies [28, 30, 47, 48] also including mid-intervention assessments. The intervention duration was explicitly reported in 19 studies, ranging from 3 to 24 months, but was not specified in the remaining six studies [27, 32–34, 39, 40, 43]. The primary outcome for the systematic review and meta-analysis was treatment compliance. Secondary outcomes included compliance behavior, lifestyle modifications, body composition (BMI, waist circumference), sex hormones (testosterone, FSH, LH), anxiety and depressive symptoms, and quality of life. Table 1 provides a comprehensive summary of these outcomes. Table 1Characteristics of the included studiesStudyCountryPublish yearSample sizeInterventionDuration(month)OutcomesIntervention groupControl groupIntervention groupControl groupJavanbakht et al. [47]Iran20243030online telehealth of mind-body Interventionsface-to-face counseling3⑫Hamzehgardeshi et al. [48]Iran20243030social network-based motivational interviewing on health behaviors and routine clinical treatmentsroutine clinical treatments and counseling2⑫Dilimulati et al. [25]China20244040WeChat-based digital intervention + lifestyle interventionMetformin + lifestyle intervention3⑤⑥⑦⑧⑨⑩⑪Wu et al. [26]China20245151WeChat-based health education + routine medicationRoutine care + routine medication3⑩⑪Lee et al. [49]Korea20231414Lifestyle changes facilitated by mobile appsRoutine health education3⑦⑧⑨⑪Feng et al. [27]China20234343Hospital-home-based health information + conventional medicationsRoutine care + routine medicationNR⑩⑪Wang et al. [28]China20226161Transtheoretical model‑based mobile health application + lifestyle interventionRoutine health education + lifestyle intervention12⑤⑥⑩⑪Guo et al. [30]China20224040Transtheoretical model‑based mobile health application + lifestyle interventionRoutine health education + lifestyle intervention6⑤⑥Liu et al. [29]China20224141online and offline health management intervention + Diane-35Routine health education + Diane-353⑦⑧⑨Zhang et al. [31]China2022213213WeChat + case management + regular medicationCase management + conventional medication3②You [32]China20223030Internet + collaborative multidisciplinary management + conventional drug therapyRoutine management + routine medicationNR②Shi [33]China20224848WeChat platform management + conventional medicationRoutine care + routine medicationNR⑨Chen [34]China20225252Web-based interactive platform management + conventional medicationRoutine care + routine medicationNR②⑩⑪Zeng et al. [35]China20225050Online lifestyle management based on the WeChat platform + conventional drug therapyRoutine care + routine medication6⑤⑥⑩⑪Zhu [36]China20213535WeChat Management + Conventional Medicationroutine medication24⑩⑪Wu et al. [37]China20215353WeChat Applet Health Education + Conventional MedicationRoutine health education + routine medication6②⑤⑥⑩⑪Zhou [38]China20215050WeChat Peer Education + Conventional MedicationRoutine health education + routine medication3①③Zhang et al. [39]China20204545Digital health education + conventional medicationRoutine health education + routine medicationNR⑩⑪Yang et al. [40]China2020237226Remote Motivational Interviewing based on WeChat Platform + Conventional MedicationRoutine health education + Routine medicationNR⑦⑧⑨⑩Wu et al. [41]China20203031WeChat-based platform management + conventional medicationRoutine management + routine medication12②④⑩⑪Guo et al. [42]China20208585WeChat platform management + conventional medicationRoutine management + routine medication3⑦⑧⑨Chen [43]China20194444Web-based platform peer education + conventional medicationRoutine medicationNR②⑩⑪Li [46]China20183836WeChat-based platform management + conventional medicationRoutine management + routine medication3①③⑤⑦⑧⑨⑩Ke et al. [44]China20173939WeChat-based platform management + conventional medicationRoutine health education + routine medication3①③⑤⑥⑦⑨⑩Meng [45]China20164141WeChat-based platform peer education health education + conventional drug therapyRoutine health education + routine medication3①③④⑤⑦⑧⑨Outcomes: ①Treatment compliance; ②Compliance behavior; ③lifestyle modifications; ④follow-up compliance rates; ⑤BMI body mass index; ⑥Waist circumference; ⑦T testosterone; ⑧LH luteinizing hormone; ⑨FSH follicle-stimulating hormone; ⑩Anxiety; ⑪Depression; ⑫Quality of life; NR: Not reported Characteristics of the included studies Outcomes: ①Treatment compliance; ②Compliance behavior; ③lifestyle modifications; ④follow-up compliance rates; ⑤BMI body mass index; ⑥Waist circumference; ⑦T testosterone; ⑧LH luteinizing hormone; ⑨FSH follicle-stimulating hormone; ⑩Anxiety; ⑪Depression; ⑫Quality of life; NR: Not reported Risk bias in included studies All 25 included RCTs were assessed for potential sources of bias using the Cochrane tool. The results were summarized following the risk categories proposed by the tool, as illustrated in Fig. 2A and B. All studies adequately described appropriate random sequence generation methods and maintained strict protocol adherence during follow-up. Six studies [25, 28, 30, 47–49] provided detailed allocation concealment descriptions, demonstrating a low risk of bias in allocation concealment; the remaining 19 studies lacked allocation concealment reporting, resulting in unclear bias risk. One study [49] implemented participant and personnel blinding (low risk); three studies [29, 33, 42] did not blind participants and personnel but utilized objective outcomes (low risk); two studies [25, 48] failed to blind participants and personnel (high risk). 19 studies [26–28, 30–32, 34–41, 43–47] did not report blinding status (unclear risk). Four studies [28, 30, 48, 49] blinded outcome assessors (low risk); three studies [29, 33, 42] used unblinded outcome assessors for objective measures (low risk); One study [25] employed unblinded outcome assessors for subjective outcomes (high risk); Seventeen studies [26, 27, 31, 32, 34–41, 43–47] did not report outcome assessor blinding (unclear risk). 19 studies [25–27, 32–39, 41, 43, 44, 46–50] demonstrated complete outcome data without attrition (low risk); six studies [28–31, 42, 45]had missing data with comparable dropout rates between intervention/control groups (low risk). All studies reported predefined outcomes without selective reporting (low reporting bias risk). Four studies disclosed funding sources without detailed descriptions (unclear other bias risk), while the remaining 21 studies exhibited balanced baseline characteristics and no funding conflicts (low other bias risk). (A) Risk of bias summary for randomized controlled trials. (B) Risk of bias summary for randomized controlled trials Treatment compliance Four RCTs involving 325 PCOS patients [38, 44–46] evaluated the effect of digital health interventions on treatment compliance over a 3-month period. All interventions spanned 3 months. A random-effects meta-analysis demonstrated that digital interventions significantly improved treatment compliance compared to control groups (SMD 1.37, 95% CI 0.37 to 2.37, P = 0.007, I2 = 94%; Fig. 3). The high heterogeneity was substantially reduced (I²=0%) after excluding one study that employed a WeChat-based peer education intervention, suggesting this particular approach as a potential source of variation. Sensitivity analyses were performed to understand the stability of the results. Each time a study was removed from the analysis, the results showed that the digital health intervention significantly improved treatment compliance (Supplementary File 2: Fig. 1). Forest plot of the effect of digital health interventions on treatment compliance in patients with PCOS According to our assessment, the evidence for the effect of digital health interventions on treatment compliance is of moderate quality. (Table 2). Table 2Results of a subgroup analysis of the effects of a digital health intervention on anxiety and depression in patients with PCOSOutcomesSubgroup factorsStudy numberHeterogeneityEffect modelMeta-analysisP valueI2(%)Relative effect (95%CI)P value Anxiety Assessment Scale Self-Rating Anxiety Scale7< 0.0000189Random−1.29(−1.82, −0.76)< 0.00001Hamilton Anxiety Scale3< 0.000191Random−1.06(−1.90, −0.23)0.01Trait Anxiety Scale20.980Fixed−0.90(−1.07, −0.72)< 0.00001self-rating scale1NANANA−1.58(−2.12, −1.04)< 0.00001Hospital Anxiety Depression -Anxiety Scale1NANANA0.07(−0.37, 0.50)0.77 Study duration NR5< 0.0000192Random−1.22(−1.82, −0.63)< 0.00013 months4< 0.0000192Random−0.85(−1.68, −0.02)0.04≥ 6 months50.00474Random−1.16(−1.57, −0.75)< 0.00001 Depression Assessment Scale Self-Depression Assessment6< 0.0000185Random−1.31(−1.80, −0.82)< 0.00001Hamilton Depression Rating Scale3< 0.000191Random−1.15(−1.98, −0.32)0.007Self-Assessment Scale1NANANA−0.82(−1.32, −0.34)0.0009Epidemiological Studies Depression Scale1NANANA−0.17(−0.57, 0.91)0.65Hospital Anxiety Depression-Depression Scale1NANANA0.02(−0.42, 0.46)0.92 Study duration NR4< 0.0000192Random−1.33(−2.15, −0.51)0.0013 months3< 0.0000196Random−0.63(−2.11, 0.85)0.40≥ 6 months50.470Fixed−0.98(−1.18, −0.78)< 0.00001 Results of a subgroup analysis of the effects of a digital health intervention on anxiety and depression in patients with PCOS Compliance behavior Six trials involving 845 PCOS patients [31, 32, 34, 37, 41, 43] compared the efficacy of digital health interventions (n = 422) versus controls (n = 423) on compliance behavior. The Meta-analysis demonstrated that, compared to the control group, digital health interventions significantly improved the following compliance behavior: medication compliance (OR 5.29, 95% CI 1.97 to 14.21, P = 0.001, I²=51%; Fig. 4A), balanced diet adherence (OR 0.20, 95% CI 0.15 to 0.26, P < 0.00001, I²=0%; Fig. 4B), regular sleep schedules (OR 2.24, 95% CI 1.56 to 3.22, P < 0.0001, I²=0%; Fig. 4C), physical activity engagement (OR 4.92, 95% CI 3.42 to 7.09, P < 0.00001, I2 = 0%; Fig. 4D), and self-measurement of body mass (OR 5.10, 95% CI 2.11 to 12.30, P = 0.0003, I2 = 0%; Fig. 4E). (A) Forest plot of the effect of a digital health intervention on compliance behavior-medication compliance in patients with PCOS. (B) Forest plot of the effect of a digital health intervention on compliance behavior-balanced diet adherence in patients with PCOS. (C) Forest plot of the effect of a digital health intervention on compliance behavior-regular sleep schedules in patients with PCOS. (D) Forest plot of the effect of a digital health intervention on compliance behavior-physical activity engagement in patients with PCOS. (E) Forest plot of the effect of a digital health intervention on compliance behavior-self-measurement of body mass in patients with PCOS. (F) Forest plot of the effect of a digital health intervention on follow-up compliance rates in patients with PCOS Separately, two studies [41, 45] evaluated the efficacy of digital health interventions in improving follow-up compliance among PCOS patients. A pooled analysis showed a statistically significant increase in follow-up compliance rates with digital interventions compared to controls (OR 4.04, 95% CI 1.41 to 11.58, P = 0.009, I2 = 24%; Fig. 4F). Lifestyle modifications Four studies participated by a total of 325 PCOS patients [38, 44–46] evaluated the efficacy of digital health interventions on lifestyle modifications. Meta-analysis demonstrated significantly greater improvements in the digital health interventions group versus controls across multiple domains: Dietary rhythm(SMD − 2.31, 95% CI − 2.92 to − 1.70, P < 0.00001, I²=0%; Fig. 5A), Sports consciousness (SMD 2.33, 95% CI 1.54 to 3.13, P < 0.00001, I²=55%; Fig. 5B), Content of the diet (SMD − 1.57, 95% CI − 2.27 to − 0.88, P < 0.00001, I²=68%; Fig. 5C), and rhythm of life(SMD − 0.83, 95% CI − 1.55 to − 0.11, P = 0.02, I²=72%; Fig. 5D). In contrast, no statistically significant effect was observed on body awareness (SMD − 0.24, 95% CI − 0.65 to 0.16, P = 0.24, I²=9%; Fig. 5E). (A) Forest plot of the effect of a digital health intervention on lifestyle modifications-Dietary rhythm in patients with PCOS. (B) Forest plot of the effect of a digital health intervention on lifestyle modifications-Sports consciousness in patients with PCOS. (C) Forest plot of the effect of a digital health intervention on lifestyle modifications-Content of the diet in patients with PCOS. (D) Forest plot of the effect of a digital health intervention on lifestyle modifications-Rhythm of life in patients with PCOS. (E) Forest plot of the effect of a digital health intervention on lifestyle modifications-Body awareness in patients with PCOS. (F) Forest plot of the effect of a digital health intervention on health-promoting lifestyle scores in patients with PCOS Three studies [30, 37, 48] assessed the impact of digital health interventions on health-promoting lifestyle scores in PCOS patients. Meta-analysis demonstrated significantly greater improvements in lifestyle scores with digital health interventions compared to controls (SMD 1.61, 95% CI 0.60 to 2.62, P = 0.002, I2 = 90%; Fig. 5F). Notably, this analysis revealed substantial heterogeneity. Due to the limited number of studies, we could not explore potential sources through subgroup analysis. Sensitivity analysis confirmed the robustness of this finding, as the results remained statistically significant with the sequential exclusion of each study (Supplementary File 2: Fig. 2). Body composition A meta-analysis of nine studies involving 1140 PCOS patients [25, 28, 30, 35, 37, 40, 44–46] evaluated the impact of digital health interventions on BMI. Using a random-effects model, the analysis demonstrated a significantly greater reduction in BMI in the intervention group compared to controls (SMD − 0.43, 95% CI −0.78 to −0.07, P = 0.02, I2 = 87%; Fig. 6A). However, a high degree of inter-study heterogeneity was observed, and subgroup analyses were conducted by examining the duration of the intervention. These analyses revealed that the digital health intervention group significantly reduced BMI in patients with PCOS when the intervention duration was ≥ 6 months(SMD − 0.94, 95% CI −1.15 to −0.72, P < 0.00001, I2 = 0%; Fig. 6B), and when the duration of the intervention was less than 6 months, the results showed no significant difference between the two groups (SMD − 0.07, 95% CI −0.40 to 0.26, P = 0.68, I2 = 52%; Fig. 6C). one study [40] did not specify the exact duration of the intervention. The reduction of heterogeneity to zero in the ≥ 6 months subgroup suggests that intervention duration was a major source of the observed heterogeneity. The certainty of evidence for this outcome was rated as moderate (Table 2). (A) Forest plot of the effect of a digital health intervention on BMI in patients with PCOS. (B) Forest plot of the effect of a digital health intervention on BMI in patients with PCOS (duration was ≥6 months). (C) Forest plot of the effect of a digital health intervention on BMI in patients with PCOS (duration was <6 months). (D) Forest plot of the effect of a digital health intervention on waist circumference in patients with PCOS Six studies [25, 28, 30, 35, 37, 44] analyzed the waist circumference of 520 patients with PCOS, and the meta-analysis showed a significantly greater reduction in waist circumference in the digital health intervention group compared to controls (SMD − 0.53, 95% CI −0.71 to −0.36, P < 0.00001, I2 = 21%; Fig. 6D). Sex hormones Eight studies [25, 29, 40, 42, 44–46, 49] evaluated the effect of digital health interventions on testosterone levels in PCOS patients. A random-effects meta-analysis revealed that digital health interventions significantly reduced testosterone compared to the control group (SMD − 0.37, 95% CI −0.71 to −0.04, P = 0.03, I2 = 83%; Fig. 7A), although substantial heterogeneity was observed. Among these studies, one did not specify the intervention duration while the remainder employed a 3-month intervention period. Subgroup analysis based on duration demonstrated that a 3-month digital health intervention significantly reduced testosterone levels (SMD − 0.25, 95% CI −0.42 to −0.09, P = 0.002, I2 = 0%; Fig. 7B). The certainty of evidence for this outcome was rated as moderate (Table 2). (A) Forest plot of the effect of a digital health intervention on testosterone in patients with PCOS. (B) Forest plot of the effect of a digital health intervention on testosterone in patients with PCOS (duration was 3 months). (C) Forest plot of the effect of a digital health intervention on LH in patients with PCOS. (D) Forest plot of the effect of a digital health intervention on FSH in patients with PCOS Seven studies [25, 29, 40, 42, 45, 46, 49] compared luteinizing hormone (LH) levels between groups. Fixed-effects analysis showed no significant difference (SMD − 0.03, 95% CI −0.10 to −0.15, P = 0.68, I2 = 13%; Fig. 7C). Similarly, a meta-analysis of nine studies [25, 29, 33, 40, 42, 44–46, 49] found no significant difference in FSH levels between groups (SMD 0.00, 95% CI −0.22 to 0.23, P = 0.97, I2 = 67%; Fig. 7D). Anxiety, depression, quality of life Fourteen studies [25–28, 34–37, 39–41, 43, 44, 46] assessed the effect of digital health interventions on anxiety using a random effects model. The meta-analysis demonstrated that digital interventions significantly reduced anxiety in PCOS patients compared to controls (SMD − 1.11, 95% CI −1.41 to −0.78, P < 0.00001, I2 = 88%; Fig. 8A). Heterogeneity between studies was evident and sources of heterogeneity were explored by subgroup analyses. Seven studies [27, 28, 35, 37, 39, 41, 46] used the Self-Rating Anxiety Scale (SAS), three studies [26, 34, 43] used the Hamilton Anxiety Scale (HAMA), two studies [40, 44] used the Trait Anxiety Scale, one study [36] used a self-rating scale and the last study [25] used the Hospital Anxiety and Depression Scale-Anxiety Subscale (HADS-A). Subgroup analysis by assessment tool revealed that digital interventions did not significantly improve anxiety when measured by HADS-A, but showed significant improvement with all other scales. Heterogeneity was eliminated only in the subgroup using the Trait Anxiety Scale (I² = 0%). Subgroup analysis by intervention duration (unspecified, 3 months, or ≥ 6 months) did not resolve heterogeneity. Both the type of anxiety scale and intervention duration may contribute to the observed heterogeneity (Table 3). (A) Forest plot of the effect of a digital health intervention on anxiety in patients with PCOS. (B) Forest plot of the effect of a digital health intervention on depression in patients with PCOS (duration was 3 months). (C) Forest plot of the effect of a digital health intervention on quality of life in patients with PCOS Table 3The evidence quality of digital health interventions for PCOSPatient or population: patients with PCOSSettings: patients with PCOSIntervention: digital health interventionsComparison: usual treatmentOutcomesIllustrative comparative risks* (95% CI)Relative effect(95% CI)No of Participants(studies)Quality of the evidence(GRADE)CommentsAssumed riskCorresponding riskUsual treatmentDigital health interventionsTreatment complianceThe mean treatment compliance in the intervention groups was1.37 standard deviations higher(0.37 to 2.37 higher)325(4 studies)⊕⊕⊕Θmoderate 1SMD 1.37 (0.37 to 2.37)Body mass indexThe mean body mass index in the intervention groups was0.27 standard deviations lower(0.39 to 0.15 lower)1140(9 studies)⊕⊕⊕Θmoderate 1SMD − 0.27 (−0.39 to −0.15)TestosteroneThe mean testosterone in the intervention groups was0.25 standard deviations lower(0.42 to 0.09 lower)585(7 studies)⊕⊕⊕ΘmoderateSMD − 0.25 (−0.42 to −0.09)AnxietyThe mean anxiety in the intervention groups was1.1 standard deviations lower(1.41 to 0.78 lower)1602(14 studies)⊕⊕ΘΘlow 1SMD − 1.1 (−1.41 to −0.78)DepressionThe mean depression in the intervention groups was1.1 standard deviations lower(1.41 to 0.78 lower)1602(14 studies)⊕⊕ΘΘlow 1SMD − 1.1 (−1.41 to −0.78)*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI)CI: Confidence interval;GRADE Working Group grades of evidenceHigh quality: Further research is very unlikely to change our confidence in the estimate of effectModerate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimateLow quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimateVery low quality: We are very uncertain about the estimateCI: indicates Confidence Interval; SMD: standardized mean differences; RR: risk ratios The evidence quality of digital health interventions for PCOS The mean treatment compliance in the intervention groups was 1.37 standard deviations higher (0.37 to 2.37 higher) 325 (4 studies) ⊕⊕⊕Θ moderate 1 The mean body mass index in the intervention groups was 0.27 standard deviations lower (0.39 to 0.15 lower) 1140 (9 studies) ⊕⊕⊕Θ moderate 1 The mean testosterone in the intervention groups was 0.25 standard deviations lower (0.42 to 0.09 lower) 585 (7 studies) ⊕⊕⊕Θ moderate The mean anxiety in the intervention groups was 1.1 standard deviations lower (1.41 to 0.78 lower) 1602 (14 studies) ⊕⊕ΘΘ low 1 The mean depression in the intervention groups was 1.1 standard deviations lower (1.41 to 0.78 lower) 1602 (14 studies) ⊕⊕ΘΘ low 1 *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) CI: Confidence interval; GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate Very low quality: We are very uncertain about the estimate CI: indicates Confidence Interval; SMD: standardized mean differences; RR: risk ratios Twelve studies [25–28, 34–37, 39, 41, 43, 49] including 1015 patients with PCOS reported the effect of digital health interventions on depressed symptoms using a random effects model. The analysis showed that digital interventions significantly improved depressive symptoms compared to controls (SMD − 1.01, 95% CI −1.40 to −0.62, P < 0.00001, I2 = 88%; Fig. 8B). Heterogeneity between trials was evident, with subgroup analyses based on different depressive symptoms scales and duration of intervention. Six studies [27, 28, 35, 37, 39, 41] used the Self-Depression Assessment (SDA) scale, three studies [26, 34, 43] used the Hamilton Depression Rating Scale (HAMD), and the remaining three studies used the Self-Assessment Scale [36], the Epidemiological Studies Depression Scale [49], and the Hospital Anxiety Depression-Depression Scale (HADS-D) [25]. Subgroup analysis revealed no significant improvement when depression was measured with HADS-D or the Epidemiological Studies Depression Scale, whereas significant improvement was observed with all other scales. Heterogeneity persisted across most rating scale subgroups but was eliminated (I² = 0%) in the subgroup of interventions lasting ≥ 6 months, suggesting intervention duration as a potential source of heterogeneity (Table 3). The certainty of evidence for the effects of digital health interventions on both anxiety and depression was rated as low (Table 2). Two studies of 120 patients with PCOS [47, 48] assessed the impact of digital health interventions on quality of life. Meta-analysis showed that digital health interventions significantly improved the quality of life compared to controls (SMD 0.63, 95% CI 0.02 to 1.24, P = 0.04, I2 = 63%; Fig. 8C). Publication bias For anxiety and depression, visual inspection of the funnel plots suggested possible asymmetry (Fig. 9A and B), although Egger’s test did not provide statistically significant evidence of publication bias (p > 0.05, Fig. 9C and D). To further evaluate the potential impact, we conducted Trim and Fill analysis. The method imputed 4 potentially missing studies for anxiety and 3 for depression. The adjusted effect estimates (Anxiety: SMD − 1.33, 95% CI −1.73 to −0.93; Depression: −1.28, 95% CI −1.75 to −0.57) were somewhat attenuated compared to the original estimates but maintained consistent directionality and statistical significance. (A) Funnel plot of publication bias for digital health interventions for anxiety in PCOS patients. (B) Funnel plot of publication bias for digital health interventions for depression in PCOS patients. (C) Egger's test plot for anxiety in PCOS patients with digital health interventions. (D) Egger's test plot for depression in PCOS patients with digital health interventions Sensitivity analysis Substantial heterogeneity was observed for several outcomes, including treatment compliance, health-promoting lifestyle scores, body mass index, anxiety, and depression (Supplementary File 2: Figures 1, 2, 3, 4 and 5). To assess the robustness of these pooled estimates, a leave-one-out sensitivity analysis was performed. This involved sequentially excluding each study and recalculating the summary effect size. The results demonstrated that no single study exerted a disproportionate influence on the overall findings. For all outcomes, the recalculated standardized mean difference (SMD) remained statistically significant and within the confidence interval of the original pooled estimate, confirming the resilience of the meta-analysis results. Discussion This systematic review of 25 randomized controlled trials evaluated the effects of digital health interventions on multiple outcomes in PCOS patients, including treatment adherence, compliance behavior (regulated medication, sensible diet, routine, increased exercise, self-measurement of body mass, follow-up), lifestyle modifications (dietary rhythms, sports consciousness, content of the diet, body awareness, rhythm of life, Healthy Lifestyle Promotion Score), body composition (BMI, waist circumference), sex hormones (testosterone, LH, FSH), anxiety and depression, and quality of life. Meta-analysis showed that digital health interventions significantly improved treatment compliance, compliance behavior, quality of life, health-promoting lifestyles, and reduced anxiety, depression, and serum testosterone levels in patients with PCOS compared to the control group. However, no significant effects were observed on LH or FSH levels, or on body awareness of lifestyle changes. Subgroup analyses indicated that intervention duration moderated certain outcomes: BMI improvement was more pronounced after at least six months of intervention, and depression scores improved significantly only at the six-month time point, whereas anxiety relief remained consistent across durations. Furthermore, subgroup analyses revealed that heterogeneity was influenced by the type of anxiety and depression assessment tools as well as intervention duration, suggesting these factors as potential sources of variability across studies. Unhealthy lifestyles are prevalent in patients with PCOS, for which lifestyle interventions serve as the primary treatment. However, long - term adherence remains challenging due to high dropout rates over time [51]. Digital health interventions offer a promising alternative by providing individualized, convenient, and accessible support to promote behavioral changes in diet, exercise, and treatment adherence [52]. In line with previous meta-analyses [21], our findings confirm that digital health interventions significantly improve treatment adherence in PCOS patients. Nevertheless, a three-component lifestyle intervention (diet, exercise, and cognitive behavioral therapy) incorporating SMS messaging, as implemented by Jiskoot et al. [53, 54], improved eating behaviors and promoted weight loss but did not significantly enhance adherence to the lifestyle intervention compared to the control group. This lack of significant improvement in adherence may be attributed to the absence of interactive doctor-patient communication in the SMS-based approach. These results suggest that the efficacy of digital health interventions may depend on specific implementation features, such as interactivity, modality, frequency, and duration. Obesity represents a significant risk factor for PCOS and exacerbates its clinical manifestations [8, 55]. Consequently, weight management has become a central objective in PCOS treatment. Digital health interventions offer a novel strategy to facilitate behavioral modifications—such as dietary adjustments and increased physical activity—by combining ongoing supervision with personalized support [56]. Previous studies have demonstrated that web-based lifestyle interventions can lead to significant improvements in anthropometric parameters, including body weight and waist circumference [57]. Our meta-analysis further confirms that digital health interventions significantly enhance anthropometric measures, such as BMI and waist circumference, in patients with PCOS. This finding contrasts with a previous meta-analysis [21], which reported no significant improvement in BMI following internet-based extended care compared to control conditions. The discrepancy may be attributed to the shorter intervention durations in earlier studies, where only one trial lasted 6 months and the remainder were limited to 3 months. To more accurately assess the long-term efficacy of digital health interventions on anthropometric outcomes, future trials should incorporate extended intervention periods. Patients diagnosed with PCOS face an elevated risk of psychiatric comorbidities, including clinically significant anxiety and depression, as well as impaired health-related quality of life [58]. These conditions often show poor rates of spontaneous remission, underscoring the need for sustained intervention. Digital health interventions offer a promising approach to addressing these challenges by providing accessible emotional support and disease-related information, thereby potentially alleviating psychological symptoms in this population [59]. While a previous meta-analysis [21] reported no significant improvement in anxiety with internet-based extended care, its conclusion was limited by the inclusion of only three studies exhibiting considerable heterogeneity, which may compromise the reliability of the findings [21] In contrast, a growing body of evidence from recent studies across various clinical contexts has demonstrated the efficacy of digital interventions in improving psychological outcomes such as depression, anxiety, and quality of life [60–62]. Our meta-analysis further supports these benefits specifically in PCOS patients, showing statistically significant improvements in depressive symptoms. However, subgroup analyses revealed important nuances. Treatment effects varied substantially depending on the assessment tool used; no significant improvements were detected when outcomes were measured with the HADS-D or the Epidemiological Studies Depression Scale. Furthermore, intervention duration emerged as a critical factor-consistent symptom reduction with low heterogeneity was observed only in interventions lasting ≥ 6 months. This finding is consistent with a 20-week lifestyle intervention study in which improvements in depressive symptoms among PCOS patients were sustained for only 10 weeks post-intervention and difficult to maintain long-term [63]. Collectively, these results indicate that both the choice of measurement instrument and intervention duration are key moderators of observed efficacy, highlighting the need for standardized assessment methods and sustained engagement strategies in digital mental health interventions for PCOS. A previous systematic review showed that digital health interventions (mobile apps) can enhance self-management and alleviate symptoms in cancer patients, thereby significantly improving the quality of life in those with breast cancer [64]. Similarly, our study found that digital health interventions significantly improved quality of life in PCOS patients. However, this conclusion is based on only two studies included in the meta-analysis, underscoring the need for future research to include more original studies to clarify the benefits of digital health interventions on quality of life in this population. Furthermore, digital health interventions may improve both treatment compliance and quality of life in PCOS patients, which may consequently contribute to hormonal improvements. Our meta-analysis indicated that these interventions help reduce testosterone levels, a finding consistent with that reported by Tian et al. [65]. Notably, however, no significant differences were observed in LH or FSH levels. This absence of effect may be attributed to the relatively short duration of the interventions in the included trials, which generally ranged from 3 to 6 months. Shorter intervention periods may be insufficient to induce detectable changes in these hormone levels. Therefore, future studies with longer intervention durations are warranted to accurately assess the effects of digital health interventions on sex hormone profiles in PCOS patients. Digital health interventions improve the management of PCOS through multiple synergistic mechanisms. First, by providing 24/7 access to educational resources, monitoring tools, and professional guidance, they overcome the spatiotemporal limitations of conventional healthcare and establish a continuous support system for disease management. Second, through features such as self-monitoring, goal setting, and real-time feedback, these interventions translate behavioral change theories into practical tools, enabling patients to decompose complex lifestyle modifications into manageable daily actions-thereby improving self-efficacy and long-term adherence. Furthermore, by delivering low-threshold psychological support (e.g., cognitive behavioral therapy modules and peer support forums), they effectively alleviate anxiety and depression while enhancing motivation for self-management. Ultimately, these sustained behavioral and psychological improvements help ameliorate core metabolic disturbances in PCOS, such as IR and HA, which in turn reduces ovarian androgen production and promotes the restoration of ovulatory function. Thus, digital health interventions serve not merely as reminder tools, but as integrated management systems capable of systematically influencing the fundamental pathological processes of PCOS. This systematic review supports the efficacy of digital health interventions for a range of outcomes in PCOS, with moderate-certainty evidence for improvements in treatment compliance, BMI, and testosterone, and low-certainty evidence for benefits in anxiety and depression. However, it is important to acknowledge the limitations inherent to this study. Primarily, the duration of the intervention in the majority of the included trials was relatively brief, ranging from three to six months, with only three trials spanning one to two years. Furthermore, the intervention duration was not reported in seven trials, which hinders the ability to conclusively determine the effectiveness of long-term digital health interventions. Secondly, there was considerable heterogeneity between the studies, which may be attributable to the specific intervention modality, the duration of the intervention, and the method of assessing outcome indicators. The relatively small sample size in some analyses could increase the risk of Type I and Type II statistical errors. Finally, as all included studies were conducted in Asian populations, the applicability of our findings to other ethnic and cultural settings requires further validation. Digital health interventions have been shown to offer significant advantages in the management of chronic diseases. However, a variety of specific modalities are available, with the most common applications including WeChat platforms, text messaging, mobile health apps, and Internet platforms. These different modalities may cater to the diverse preferences of PCOS patients in terms of the use of electronic devices, thereby providing them with a wider range of options. Consequently, future research in the domain of digital health interventions for PCOS may encompass diverse modalities as interventions, to identify additional therapeutic digital health interventions for PCOS. Conclusion This systematic review and meta-analysis demonstrated that digital health interventions significantly improved treatment compliance, health-promoting behaviors, lifestyle, quality of life, body composition, psychological symptoms, and serum testosterone levels in patients with PCOS. However, no significant improvements were observed in LH or FSH levels, potentially due to the shorter intervention durations of the included studies. Given the substantial heterogeneity among the included trials, future research should include larger, multicenter, high-quality studies with longer intervention periods. Supplementary Information Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Acknowledgements Thanks to all the authors for their contributions to the manuscript. Abbreviations Polycystic ovary syndrome China national knowledge infrastructure Randomized controlled trial Hyperandrogenemia Polycystic ovarian morphology Insulin resistance Standard deviation Mean difference I-squared Standardized mean differences Ratio of ratios Confidence interval Luteinizing hormone Follicle-stimulating hormone The self-rating anxiety scale The self-depression assessment The hamilton anxiety scale The hamilton depression rating scale The hospital anxiety and depression scale Author contributions MX and JZ put forward the ideas and data analysis of the article, MX and XQ were responsible for formulating the inclusion and exclusion criteria of the article and conducting literature search and screening, HH and BX were responsible for the final inspection of the included literature, MX carried out the data analysis and the first draft of the article was written, JZ made revisions to the article, and all authors provided final review and confirmation of the manuscript. Funding This research was supported by Chengdu technological innovation research and development project (2024-YF05-00258-SN). Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate Not applicable. Consent for publication All authors have reviewed the manuscript and consented for publication. 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. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Data Availability Statement No datasets were generated or analysed during the current study.