PEPRKD-depression: A knowledge database supporting evidence-based personalized exercise prescription recommendations in depression

2.1 Study design

This study is designed as a knowledge base construction study aiming to systematically compile exercise intervention protocols for depression. Key reporting elements were structured with reference to the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Scoping Reviews) framework, including search strategy, eligibility criteria, study selection, and data extraction. The study selection process is illustrated in Figure 1. The corresponding PRISMA-ScR checklist is provided in Supplemental Material S1.

OPEN IN VIEWER

Figure 1.

The flowchart for the collection of exercise intervention protocols for depression.

2.2 Data source

Original research articles published between 1960 and 2023 were primarily retrieved from PubMed. PubMed, a literature abstract and citation database managed by the National Institutes of Health (NIH), is one of the world’s largest and most authoritative biomedical literature databases. All articles indexed in PubMed undergo rigorous peer review, earning it widespread recognition in the field. To systematically retrieve literature on exercise interventions for depression, we used “depress*”, “exercise”, “physical”, and “training” as search keywords in PubMed, with filters applied to include only titles containing both “depress*” and exercise-related terms. Due to the concept-specific nature of the topic, the PubMed search was limited to title fields to enhance retrieval specificity. Preliminary searches showed that title-and-abstract searches substantially increased records with peripheral relevance without materially improving the identification of directly eligible studies. Given that standardized assessment scales for depression were first developed in the 1960s, we set the literature inclusion timeframe as January 1, 1960, to December 31, 2023. Specifically, the search strategy was formulated as “(depress*[ti]) AND (exercise[ti] OR physical[ti] OR training[ti] OR aerobic[ti] OR sport[ti] OR athletic*[ti] OR yoga[ti]) AND 1960/01/01:2023/12/31[dp]”.

2.3 Inclusion and exclusion criteria

The knowledge database mainly includes original intervention research, such as randomized controlled trials, quasi-experimental studies, and self-controlled pre-post cohort studies. The scope of depression-related conditions is broad, encompassing depressive disorder/depression, subthreshold depression, depressive symptoms, and perinatal depression. The database excludes reviews, meta-analyses, letters, case reports, editorials, comments, book chapters, animal studies, and non-English publications. Additionally, studies without specific exercise protocols, such as qualitative research, mechanistic studies, and correlation analyses, were excluded.

2.4 Data extraction

Study selection was performed independently by two reviewers to minimize bias. A standardized data extraction form with predefined extraction items was developed prior to formal data collection to ensure consistency across studies. Before data extraction, the reviewers discussed and unified the extraction criteria. Data extraction was conducted by one reviewer and independently verified by a second reviewer for accuracy and completeness. Any disagreements were resolved through discussion, and if consensus could not be reached, a third reviewer made the final decision. As illustrated in Figure 2, the entity relationships within PEPRKD-Depression were constructed in accordance with the American College of Sports Medicine’s (ACSM) guidelines ¹³ for exercise testing and prescription, as well as recommendations from psychiatrists. Detailed information on exercise interventions for depression was extracted based on the ACSM FITT-VP framework (Frequency, Intensity, Time, Type, Volume, Progression). By integrating data from the included literature, PEPRKD-Depression encompasses information on depression-related details, baseline characteristics, physical fitness assessments, exercise risk evaluations, exercise intervention protocols, adverse event monitoring and prevention, and exercise efficacy evaluations. The specific content of exercise interventions is presented in Table 1.

OPEN IN VIEWER

Figure 2.

Entity-relationship diagram of the Personalized Exercise Prescription Recommendation Knowledge Database for Depression (PEPRKD-Depression). The diagram illustrates the overall architecture of the knowledge base and how different components are logically connected. At the center is the core knowledge framework of PEPRKD-Depression, which integrates five functional modules. These modules collectively represent the key elements required to build a knowledge base for personalized exercise prescriptions for depression. Building outward from the center, the figure systematically presents the detailed contents of five modules. Among them, the exercise/physical activity module (shown on the left side of the figure) is further expanded into four sub-tables, which provide a comprehensive description of exercise protocol details, as well as exercise-related risk assessment, effect evaluation, and adverse events. Different colors are used to distinguish modules and substructures, helping to clarify the hierarchical organization and relationships within the knowledge base.

OPEN IN VIEWER

Table 1.

The exercise intervention content and description of Personalized Exercise Prescription Recommendation Knowledge Database for Depression.

Website page	Content	Description
Exercise risk assessment	PMID, Medical clearance, PAR-Q/PARmed-X, Exercise test, CVD risk factors evaluation, Physical examination, Method	/
Physical activity intervention protocol	PMID, Protocol number, Protocol name	/
	Classification	Aerobic exercise, Resistance exercise, Flexibility exercise, Balance exercise, Anaerobic exercise, Yoga, Tai Chi, Qigong, Coordination exercise, Dancing, Taekwondo, Other
	Frequency	Times or days per week
	Intensity	Very light, Light, Moderate, Vigorous, Near-maximal to maximal, Other
	Intensity calculation indicator	RM, RPE, METs, HRmax, HRpeak, HRR, VO2peak, VO2max, Lactic acid workload
	Time	Minutes per session or day
	Type	Mode of physical activity
	Volume	Minutes per week
	Periods of time	Days without dialysis, During hemodialysis treatment, In the morning/afternoon/evening, Alternate days, 72 h after each chemotherapy course, Specific time period
	Warm-up/Cool-down	Time, Mode
	Number of sets per session/Time of each set/Repetitions per set	Specific to resistance exercise
	Progression	YES/NA
	Intertrial interval	Exercise rest intervals
	Duration	Total duration of exercise intervention
	Medical supervision	YES/NA
	Exercise contraindications	/
	Combination therapy	Antidepressant medication, Cognitive behavioral therapy, Electroconvulsive therapy, Music, Other
	Protocol	Original full intervention program
Adverse events evaluation	PMID, Protocol name, Adverse events, Adverse events definition/content, Prevention protocol, Conclusion	/
	Adverse events classification	Cardiovascular negative events, Musculoskeletal injury, Adverse respiratory events, Attempted suicide, Sleep quality, Gastrointestinal adverse events, Hematologic adverse events, Urinary adverse events, Maternal adverse events, Sports fatigue, Autonomic nerves system, Central nerves system, Mental health, Glucose metabolism disorder, Reproductive system adverse events, Microvascular disease
Physical activity effect evaluation	PMID, Protocol number, Protocol name, Intervention time, Follow-up period, Classification of effect, Indicators, P value, Control, Results, Conclusion	/

PMID, PubMed Unique Identifier; PAR-Q, Physical activity readiness questionnaire; PARmed-X, Physical Activity Readiness Medical Examination; CVD, cardiovascular disease; RM, repetition maximum; RPE, rating of perceived exertion; METs, metabolic equivalents; HR, heart rate; HRR, heart rate reserve; VO2peak, peak oxygen consumption; VO2max, maximum oxygen uptake.

2.5 Evidence classification

Included studies were categorized according to the Oxford Centre for Evidence-Based Medicine (OCEBM) Levels of Evidence (2009) ³¹ to provide a general indication of study design hierarchy. Because the primary aim of this study was to construct a structured knowledge database of exercise intervention protocols rather than to synthesize intervention effectiveness, a formal risk-of-bias assessment was not conducted. This classification did not influence study selection or the overall exercise recommendation, but was considered as one of several factors in protocol ranking rather than the sole determinant.

2.6 Knowledge database implementation

PEPRKD-Depression employs Vue3 (https://cn.vuejs.org/) as the web development framework and Nginx as the web server, with data stored in a MySQL database. Additionally, several JavaScript plugins—including ECharts.js (https://echarts.apache.org), Sigma.js (https://www.sigmajs.org/), and Gephi (https://gephi.org/)—were utilized to generate visual data charts.

3.1 The landscape of PEPRKD-Depression

PEPRKD-Depression contains a total of 567 studies featuring 769 exercise intervention protocols. These physical activity intervention protocols primarily include data on frequency (997 records), intensity (592 records), time (915 records), type (979 records), volume (941 records), progression (334 records), periods of time (169 records), warm-up (464 records), cool-down (406 records), and combination therapy (463 records). Additionally, exercise-related assessments encompass exercise risk evaluation (208 records), adverse event evaluation (103 records), and physical activity efficacy evaluation (1,745 records). Several classification characteristics of PEPRKD-Depression are presented in Table 2.

OPEN IN VIEWER

Table 2.

Characteristics of personalized exercise prescription recommendation knowledge database for depression.

Item	Classification	n (%)
Publication	/	567(100)
Exercise intervention protocols	/	769(100)
Age	<18 years	33(3.61)
​	18-65 years	635(69.40)
	>65 years	247(26.99)
Sex	Female; Male	419(73.90)
​	Female	131(23.10)
	Male	17(3.00)
Depression	Depressive symptoms	395(59.67)
​	Depressive disorder/Depression	228(34.44)
	Perinatal depression	36(5.44)
	Subthreshold depression	3(0.45)
Comorbidity (mental disease)	Anxiety	183(30.70)
​	Other mental disease	27(4.53)
Underlying diseases	Neurological disorders	31(5.34)
​	Any type of cancer	28(4.82)
	Cardiovascular disease	22(3.79)
	Stroke	16(2.75)
	End-stage renal disease	12(2.07)
	Diabetes	12(2.07)
	Chronic pain	11(1.89)
Region	United States	141(24.87)
​	China	49(8.64)
	Germany	38(6.70)
	India	32(5.64)
	Brazil	31(5.47)
	Iran	26(4.59)
	Republic of Korea	26(4.59)

Furthermore, additional statistical information from the knowledge database is available on our website (https://dpa.bioinf.org.cn/), including regional distribution maps of all studies, statistics on comorbidities, and statistical charts illustrating improvements in efficacy evaluation indicators following exercise. The relationships between exercise effects and corresponding exercise types, frequencies, intensities, and volumes are visualized using Sankey diagrams. Trends in key exercise variables, data completeness, and annual publication distribution are presented in Supplementary Table S2.

3.2 Applications of PEPRKD-Depression

3.2.1 A preliminary decision-support tool for exercise recommendation

PEPRKD-Depression provides basic functions such as knowledge browsing, information retrieval, and original study information downloading. In addition, it serves as a preliminary decision-support tool for exercise recommendation. When users provide baseline characteristics, clinical information related to depression and comorbidities, suicide risk assessment results, responses to exercise preparticipation health screening questionnaires, and desired exercise goals, the tool applies the established matching logic to identify potentially relevant exercise protocols (Figure 3 and Supplemental Table S3). Users may review the recommended exercise protocols and select preferred options, while clinicians may refer to the information when considering exercise interventions for individual patients.

OPEN IN VIEWER

Figure 3.

Physical activity protocol recommendation tool of PEPRKD-Depression. The exercise program recommendation process consists of five steps: users provide baseline information and details on depression and comorbidities, complete a suicide risk assessment and a pre-exercise health screening questionnaire, and define exercise goals, after which the system generates personalized exercise plans for selection. For users at high suicide risk, the system terminates the recommendation process and advises seeking professional help. In Step 4, if any symptoms (e.g., chest discomfort with exertion) or medical conditions (e.g., heart attack) are reported, the system also discontinues recommendations for safety reasons and advises obtaining medical clearance before exercise. Users who already engage in regular physical activity and have medical conditions may continue light- to moderate-intensity exercise without clearance, but vigorous-intensity exercise requires prior approval from a healthcare professional.

The exercise protocols displayed on the website are ranked based on the level of literature evidence, the completeness of the exercise intervention protocol, and whether an exercise risk assessment is included. In the scoring system, evidence level contributed 20% of the total score, whereas protocol completeness accounted for the largest proportion (70%), followed by risk assessment (10%). Detailed scoring criteria are provided in Supplemental Table S4. Greater weight was assigned to the completeness of the exercise protocol, as an effective exercise prescription requires clear and comprehensive reporting of its core elements. Therefore, protocols with lower levels of evidence could rank higher if they provided more detailed and implementable exercise information, while high-quality studies could rank lower if key elements were insufficiently reported.

3.3 The extension of the FITT-VP principle

In line with ACSM guidelines, PEPRKD-Depression not only compiles exercise intervention protocols but also incorporates pre-exercise fitness and risk assessments, exercise-related adverse events, and evaluations of intervention efficacy. The intervention protocols are structured according to the “FITT-VP” principle, including exercise classification, weekly frequency, intensity, session duration, exercise modality, weekly volume, progression during the intervention (e.g., changes in intensity or duration), total intervention duration, availability of medical supervision, combination therapies, and detailed program descriptions.

Furthermore, we found that “Periods of time” (i.e., specific periods for exercise) are also critical in designing exercise programs. Currently, 15.61% of the exercise protocols in the database include descriptions of “Periods of time”. Consequently, we innovatively expanded the “FITT-VP” principle to “FITT-VPP” by incorporating this additional dimension of timing. Based on the data collected so far, we have made a preliminary definition and classification of “Periods of time”.

Definition

The specific time periods during which exercise is restricted, as opposed to exercise that can occur at any time or during any phase of a disease.

Classification:

(a) By different stages of the disease: before dialysis, during dialysis, non-dialysis days, 72 hours after chemotherapy.

4.1 Potential applications of PEPRKD-Depression

While extensive literature highlights the benefits of exercise for depression,^5,34–36 and several guidelines recommend exercise as a standalone or adjunct treatment for depression,^13–15 no existing systematic research or guidelines provide individualized exercise guidance for patients with varying comorbidities or depression subtypes. By synthesizing exercise intervention protocols from PubMed-indexed studies, PEPRKD-Depression provides diverse, literature-derived exercise program options that consider patient characteristics and treatment heterogeneity, thereby serving as a structured knowledge resource and preliminary decision-support framework for future exercise recommendation research. Table 3 presents a multidimensional comparison with relevant guidelines. These guidelines primarily provide a broad, integrated framework and theoretical support for the treatment of depression, but they typically do not offer specific structured exercise intervention recommendations. To address this gap, we developed a comprehensive exercise intervention knowledge database for depression that systematically organizes existing evidence into structured and accessible exercise program options. However, it is important to note that further validation is needed to assess its clinical effectiveness.

OPEN IN VIEWER

Table 3.

The comparison with current guidelines related to depression.

​	PEPRKD-depression	APA	CANMAT	NICE	WFSBP and ASLM taskforce
Aims	Recommend exercise protocols	Provide guidelines and principles for treatment	Provide guidelines and principles for treatment	Provide guidelines and principles for treatment	Provide guidelines and principles for treatment
Exercise protocol principle	FITT-VPP	/	/	/	/
Exclusion of contraindications for exercise	The contraindications are clearly listed in the details of the exercise protocol.	It is pointed out that the prerequisite for exercise is the absence of contraindications, but the specific content is not clearly defined.	No	No	No
Personalized recommendation	Comprehensively consider the conditions of gender, age, BMI, type of depression, comorbidities and underlying diseases. And generate explainable exercise programs.	No	No	No	No

4.2 The importance of “periods of time”

Exercise prescriptions are typically developed following the “FITT-VP” principle, ¹³ and we have added “periods of time” as an additional dimension, expanding it to the “FITT-VPP” principle. The specific limitation of exercise time is directly related to the safety and effectiveness of exercise for patients. Next, we will discuss this in detail according to different classifications.

(a) “Periods of time” restrictions at different stages of disease may involve exercise safety considerations. For example, for breast cancer patients with depressive symptoms, exercise is recommended to begin 72 hours after chemotherapy to ensure safety. ³⁷ Chemotherapy can cause acute reactions such as fatigue, nausea, bone marrow suppression, and blood pressure fluctuations. Exercising too soon may increase the risk of infection or induce adverse events. For patients with chronic kidney failure who require long-term hemodialysis, exercise can be arranged during dialysis or on non-dialysis days.^38–41 However, during dialysis, many types of exercise are difficult to perform due to limitations in position, vascular access, and potential hypotension, posing safety risks. In these cases, exercise is usually limited to lower limb resistance training or stationary bicycle. On non-dialysis days, patients are generally more stable and can engage in a wider variety of exercises.

4.3 Ethical considerations

The use of automated suicide risk screening tools (C-SSRS) in digital environments raises important ethical considerations. Compared with clinician-administered assessments, automated systems lack contextual judgment and may increase the risk of misinterpretation. Accordingly, the present assessment is intended for preliminary screening rather than clinical diagnosis.

To address safety concerns, the platform incorporates safeguards for high-risk users. When high-risk responses are detected, the platform provides immediate access to crisis support resources (e.g., helplines recommended by the International Association for Suicide Prevention) and encourages users to seek professional help. However, effective referral and follow-up remain challenging in fully automated contexts.

Clinical oversight remains essential, as responsibility for diagnosis and treatment lies with licensed professionals. Institutions adopting the platform should ensure appropriate oversight and risk management. The platform is intended for both clinician-guided and patient self-directed use; however, self-use does not constitute a diagnosis or treatment.

4.4 Limitations

This study has several limitations. First, the literature search was limited to the PubMed database and English-language publications, which may have introduced selection bias and restricted the comprehensiveness of the evidence base. Additionally, the search strategy focused on title terms only, potentially missing relevant studies in which key concepts are described in the abstract but not in the titles. This approach may have also biased the results toward well-represented regions and populations, while neglecting studies from underrepresented regions or different demographic groups. Future studies may consider broader search strategies to minimize potential selection bias.

Second, the baseline characteristics used for matching in the database are limited to variables reported in existing studies. However, other patient characteristics, such as preferences, motivation, psychiatric family history, or biological factors like BDNF, ⁴⁸ may be more predictive of treatment outcomes. To enhance precision, future recommendations for personalized exercise prescriptions should integrate additional personalized or mechanistic factors. Multicenter studies, alongside multidimensional data collection and mechanistic research, will help identify the specific impact of these factors. Additionally, this data-driven matching approach relies entirely on pre-existing study data and lacks adaptive learning capabilities. Future updates could incorporate real-world data and user feedback to refine the matching logic.

Third, although the included studies were categorized according to the OCEBM Levels of Evidence, a formal risk-of-bias assessment was not conducted because the primary aim of this study was to summarize and structure reported exercise intervention protocols rather than to perform comparative effectiveness synthesis. Nevertheless, the absence of standardized quality appraisal may limit interpretation of the relative methodological strength of the included evidence. Future studies may incorporate formal risk of bias tools to support more comprehensive evidence evaluation.

Fourth, this study lacks external validation and has not undergone a formal usability evaluation with clinicians to assess whether the system contains all the necessary information for exercise prescription in real-world clinical practice. Therefore, claims of generalizability and replicability should be considered preliminary until such validation is completed. Future research should incorporate both clinician-based usability testing ⁴⁹ and external validation ⁵⁰ to enhance the system’s clinical validity, usability, and generalizability. This should include clinician benchmarking to assess agreement with expert decision-making, user-centered testing to evaluate usability, and real-world implementation studies to assess its feasibility and clinical applicability.

Finally, the current version of the database primarily serves as a platform for summarizing data and offering exercise program recommendations that are tailored to patient characteristics. However, these recommendations should be considered supportive rather than fully personalized exercise prescriptions. Future iterations of the database will be continuously improved through external validation, algorithm optimization, and clinical testing, and may evolve into a more refined decision-support framework for personalized exercise recommendation. To improve the comprehensiveness and timeliness of the database, we plan to update it every two years using a combination of automated literature searches and manual curation from major databases (e.g., PubMed and Web of Science). This update process will be overseen by the corresponding author’s research team, with support from trained research assistants. In parallel, future implementation efforts may focus on clinician training, academic dissemination, Hospital Information System (HIS) integration, and multi-institutional collaboration.