Search Strategy

The databases used for this review included Web of Science, Scopus, PubMed, PsycInfo (PsycArticles and PsycBooks), and Scielo. The search strategy was developed using keywords derived from the research questions, combined with the Boolean operator “AND.” The main search terms were “Treatment,” “Depression,” and “University Students.”

Additionally, a search in the gray literature was conducted using PubPsych and ClinicalTrials.gov to identify clinical trials that had been completed but not yet published. The same search terms were applied, with platform-specific restrictions, such as temporal filters in PubPsych and the “with results” filter in ClinicalTrials.gov. The primary author conducted the initial search and analysis, with validation by a second reviewer to address any uncertainties.

Study Selection

After retrieving studies from the databases and removing duplicates, an initial screening process was conducted. Two reviewers independently and blindly assessed the titles and abstracts of the studies. This process, performed on the “Rayyan” platform for systematic reviews, followed pre-established inclusion and exclusion criteria.

Eligibility Criteria

Eligibility criteria were defined using the PICOT framework: P (Participants): University students diagnosed with or presenting symptoms of depression; I (Intervention): Mental health interventions aimed at improving or alleviating depression symptoms; C (Comparator): Studies with control groups, randomized or non-randomized; O (Outcome): Primary outcome was a significant reduction in depression symptom scores, assessed with validated quantitative or qualitative instruments (culturally adapted where applicable). Secondary outcomes included evaluations of stress, burnout, and quality of life; and T (Type of study): Studies published in peer-reviewed journals between 2018 and 2023.

The inclusion criteria were: (1) Studies investigating the efficacy of mental health interventions targeting university students with depression; (2) Interventions conducted within universities or online (e.g., via apps), provided they were led by health professionals; (3) Studies with a primary outcome of significant changes in depression scores, using validated instruments; (4) Articles published in Portuguese, English, French, Spanish, or German; and (5) Freely accessible articles or those available through the University of São Paulo’s institutional access.

Exclusion criteria included: (1) Studies with only pre-test and post-test results from the intervention group. (2) Non-peer-reviewed studies, such as conference proceedings or dissertations.

Data Extraction and Classification

The full text of the studies selected in the first screening was analyzed by two reviewers independently and blindly, with 10% of these studies initially analyzed together by pairs to ensure adherence to the criteria before proceeding to independent analysis. Disagreements between reviewers regarding the eligibility of studies were resolved through discussion and consensus, mediated by a third team member.

Data were extracted into a standardized and pre-tested form, which included collecting information related to identification, methodological characteristics, sample characteristics, recruitment method, data treatment, type of control, intervention characteristics, results, level of evidence, and methodological rigor assessment. Data extraction was conducted independently by each reviewer, maintaining the rigor of the systematic review. Similarly, 10% of the included articles had their data extracted by two reviewers, ensuring the quality and accuracy of the process. Discrepancies were resolved following the same procedure as the previous step.

Data Treatment

Primary data analysis focused on a descriptive approach, detailing intervention configurations, therapeutic methods, delivery modes, and participant outcomes before and after interventions. Secondary analysis coded information related to research design, participant characteristics, intervention methods, and results, enabling a comprehensive narrative synthesis.

The narrative synthesis approach was chosen because the included studies varied significantly in methodology, interventions, and populations, precluding statistical meta-analysis. This method allowed for greater flexibility in interpreting the data, ensuring a broader inclusion of studies to address the review’s main questions and their practical and research implications.

Quality Assessment

The risk of bias assessment for each article was conducted using the “JBI Critical Appraisal Tool” (Barker et al., 2023). This instrument includes a checklist tailored to the specific type of research under analysis, covering various study designs. This level of detail was crucial in the tool selection process.

A group of reviewers evaluated each study based on these criteria, classifying the overall risk of bias for each study as low, moderate, or high. The assessment process was similar to the screening process: three articles were jointly analyzed in group meetings to ensure understanding of the tool, while the remaining articles were independently and blindly evaluated by two reviewers. No discrepancies were found between the reviewers’ assessments.

Study and Sample Selection

The review began with a total of 4,625 articles. Of these, 1,651 were removed as duplicates, resulting in 2,948 studies eligible for title and abstract screening. After this selection, 2,872 studies were excluded, leaving 76 articles for full-text review. Among these, one article was inaccessible, and 75 articles were fully reviewed and individually assessed. A group review of all evaluations was conducted to resolve discrepancies, leading to the final selection of 13 scientific articles published between 2018 and 2023 for this systematic review. The screening process described above is illustrated below (Figure 1 - PRISMA).

Figure 1 PRISMA flow diagram 

Characteristics of the Studies and Sample

The included studies comprised 939 participants with depressive symptoms. Of these, 61.5% used a diagnosis of depression confirmed by a health professional or determined through a cutoff on a self-report scale; 23% used self-report scales without specifying a cutoff for inclusion; and 15% did not report the instruments used for participant selection. The entire sample consisted of university students, with only two studies conducted at private institutions (^{Sawamura et al., 2022}; ^{Hamed et al., 2021}).

The study population was 69% female, with participants’ average age ranging from 19.7 ± 2.98 to 24.21 ± 5.2. Only two studies reported the participants’ economic status: ^{Ofoegbu et al. (2020)} noted that 47.8% of the treatment group received some form of scholarship, while ^{Ezegbe et al. (2019)} reported that 67% of participants were classified as middle class.

Eight studies (61.5%) were identified as randomized clinical trials (RCTs), while five (38.5%) were controlled pre- and post-test studies. All articles provided data for T0 and T1, and 61.5% included follow-up measures at T2. It is important to note that T2 measures were not exclusive to RCTs, as shown in Table 1. However, the studies by ^{Ezeudu et al. (2019)} and ^{Zhang et al. (2022)} did not specify the time intervals between tests.

Of the 13 studies, 11 evaluated the efficacy of mental health interventions of interest compared to a control condition, such as a waiting list, support group, or usual treatment. However, three studies employed distinct designs or objectives. ^{Lerardi et al. (2022)} conducted a pre- and post-test study to compare the efficacy of psychodynamic therapies in online and face-to-face modalities. Similarly, ^{Hamed et al. (2021)} investigated the differences between cognitive-behavioral therapy (CBT) and physical exercise, using CBT as the control condition. Lastly, ^{Yamamoto et al. (2018)} conducted an exploratory clinical trial with a crossover design, in which participants received both treatments sequentially, separated by a “washout” period to minimize interference between interventions.

Comparisons between groups were made, though only ^{Lerardi et al. (2022)} did not present a baseline homogeneity analysis. For their study, they utilized SPSS Statistics 27.0 to compare participants who received emergency online counseling during the COVID-19 pandemic (January 2020–July 2021) with those who received on-demand counseling before the pandemic (January 2016–December 2019).

Efficacy of interventions

The analysis of the research revealed that 12 studies (92.3%) demonstrated a significant reduction in depression scale scores for the treatment groups compared to the control groups. This indicates that the interventions used were effective in reducing depression levels, with the reduction being significantly greater than that observed in the control groups.

Table 1 presents the statistical significance (“p”) reported for each study in relation to the improvement in depressive symptoms in the treatment group compared to the control group at T1 and, when applicable, T2 (follow-up). Of the 13 studies, 11 reported p < 0.05 for this comparison, indicating that the interventions were significantly more effective than the controls (^{Ritkumrop et al., 2022}; ^{García & Maldonado, 2022}; ^{Ofoegbu et al., 2020}; ^{Ezegbe et al., 2019}; ^{Demir & Ercan, 2022}; ^{Sawamura et al., 2022}; ^{Ezeudu et al., 2019}; ^{Zhang et al., 2022}; ^{Liang et al., 2021}; ^{Yamamoto et al., 2018}; ^{Hamed et al., 2021}; ^{Fereydouni & Forstmeier, 2022}).

However, the study by ^{Demir and Ercan (2022)} did not find a statistically significant difference between the intervention and control groups in symptom improvement. Nonetheless, a significant improvement in depressive symptoms was observed within the treatment group over time (effect size = 0.970, p < 0.001). Similarly, ^{Lunardi et al. (2022)} reported no significant differences in the effectiveness of face-to-face versus online therapy. However, their study demonstrated the efficacy of both modalities, with face-to-face therapy showing t = 5.12, p = 0.000*, Cohen’s d = 0.57, and online therapy showing t = 2.82, p = 0.008*, Cohen’s d = 0.48.

^{Hamed et al. (2021)} reported comparable results, showing that both cognitive-behavioral therapy (CBT) and physical exercise effectively reduced depressive symptoms compared to the pre-treatment period (p ≤ 0.001). However, the study noted that “there was no significant difference in stress and depression between the groups after the intervention (p ≥ 0.05)” (^{Hamed et al., 2021}).

Characteristics of Interventions

In terms of intervention characterization, 61.5% of the studies investigated therapies based on the cognitive-behavioral approach (^{Demir & Ercan, 2022}; ^{Ezegbe et al., 2019}; ^{Ezeudu et al., 2019}; ^{García & Maldonado, 2022}; ^{Ofoegbu et al., 2020}; ^{Hamed et al., 2021}; ^{Liang et al., 2021}; ^{Ritkumrop et al., 2022}). Among these eight interventions, six (75%) were face-to-face, while two (25%) were conducted online. One online intervention was described as “based on ICBT theory” (^{Ofoegbu et al., 2020}), and the other was a conventional CBT program conducted online (^{Hamed et al., 2021}).

Within the cognitive-behavioral approach, 25% of the studies (or 15% of the total articles) followed third-wave cognitive-behavioral therapy frameworks, specifically Dialectical Behavior Therapy (DBT) (^{Liang et al., 2021}) and Acceptance and Commitment Therapy (ACT) (^{Ritkumrop et al., 2022}). Both interventions were conducted face-to-face.

The remaining five studies adopted different theoretical approaches. These included online psychodynamic therapy (^{Lenardi et al., 2022}), interpersonal therapy (^{Sawamura et al., 2022}; ^{Yamamoto et al., 2018}), life education counseling (^{Zhang et al., 2022}), and spiritually sensitive logotherapy (^{Fereydouni & Forstmeier, 2022}). Among these seven studies, only one (^{Lenardi et al., 2022}) involved online and individual psychotherapy. The other interventions were conducted face-to-face, with half delivered in group settings.

^{Sawamura et al. (2022)} followed the traditional model of individual, face-to-face therapy in 50-minute sessions, using interventions from third-wave cognitive-behavioral therapies (ACT and DBT). However, the study did not adhere to the recommended application protocol. Additionally, ^{Zhang et al. (2022)} noted that participants in their study received the intervention while undergoing medication treatment, with dosage and use consistent across both the intervention and control groups.

Four interventions followed established protocols previously published and tested by other authors. ^{Ezegbe et al. (2019)} used the “Treatment Manual for Cognitive Behavioral Therapy for Depression” by Rosselló and Bernal (2007). ^{Ezeudu et al. (2019)} followed the “Managing Depression Using Rational Emotive Behavior Therapy (REBT)” manual by David et al. (2004). ^{Hamed et al. (2021)} utilized “A Therapist’s Guide to Brief Cognitive Behavioral Therapy” by ^{Cully and Teten (2008)}. ^{Sawamura et al. (2022)} based their intervention on the manual “Interpersonal Counseling (IPC) for Depression in Primary Care” by Weissman et al. (2014), while ^{Yamamoto et al. (2018)} referred to the manual “How to Proceed with Interpersonal Counseling” by Mizushima.

However, two studies did not provide sufficient details about the content and organization of their sessions to enable replication (^{Lenardi et al., 2022}; ^{Ofoegbu et al., 2020}). In contrast, all other studies described the themes and techniques used during the sessions, allowing for the potential implementation of these treatments in future research.

Quality Assessment and Risk of Bias

Twelve of the included articles were classified as having a medium risk of bias, while one was categorized as high risk. The study with high risk employed a crossover design but failed to report the washout period between interventions, potentially leading to result interference. Furthermore, only five participants opted to try the alternative intervention after the initial phase, undermining the validity of the crossover design, the analysis between conditions, and ultimately contributing to the high risk of bias (^{Yamamoto et al., 2018}).

Nevertheless, some level of bias was present in all studies. A primary source of bias was the lack of blinding for participants and researchers. The nature of psychotherapy research makes it challenging to ensure participants are unaware of the intervention they are receiving. Only three studies included blinding in data analysis, where analysts were not informed of the treatment or control groups (^{Ezegbe et al., 2019}; ^{Demir & Ercan, 2022}; ^{Sawamura et al., 2022}). Other significant sources of bias included randomization of groups, choice of instruments, and transparency in reporting. The distribution of studies by risk of bias is shown in Figure 2.

Figure 2 Descriptive Risk of Bias 

Randomization of the population is critical for enhancing the reliability of results, and 76% of the reviewed studies employed this methodology. However, potential selection bias remains a concern, as voluntary participation may affect the representativeness of the sample. Another limitation was the reliance on self-report scales to assess participants’ depressive symptoms, which compromises data reliability. All studies included in this review used self-report scales. Additionally, 76% of the studies relied on a single scale to measure depression, 20% used both a depression scale and a quality-of-life scale, and only 10% employed more than one self-report scale for depressive symptoms (Table 2). Variability in scales across studies complicates comparisons and result interpretation. Commonly used scales included the Beck Depression Inventory I and II (BDI-I, BDI-II), Depression Anxiety Stress Scales-21 (DASS-21), Zung Self-Rating Depression Scale (SDS), Symptom Severity Scale (SSS), Patient Health Questionnaire-9 (PHQ-9), Perceived Stress Scale-10 (PSS-10), and the Symptom Checklist-90-Revised (SCL-90-R). It is important to note that the DASS-21 measures depression as one of its three subscales.

Transparency issues were also identified. Two studies did not report the follow-up period, and five studies failed to fully conduct follow-ups. Attrition rates were unreported in four studies, and for half of these, attrition was manually calculated during this review, as only limited data were available. Similarly, three articles did not report the number of participants who completed the interventions (Table 2).

Limitations

This review has important limitations. It did not restrict inclusion to randomized controlled trials (RCTs), aiming to encompass quantitative and qualitative studies and underrepresented approaches. However, no included study employed qualitative methodologies, limiting the depth and subjectivity of the findings. Other limitations include the small number of included articles, heterogeneity among studies, and the absence of a meta-analysis.

Future Research

Future research should address the feasibility of long-term interventions, including resource availability, funding, and the training of qualified professionals. Additionally, studies should focus on the practical application of intervention protocols, ensuring their rigor and sustainability in real-world settings.