1. Determining Eligibility Criteria

Eligibility criteria ensure only relevant, high-quality randomized controlled trials (RCTs) are included in your meta-analysis of cognitive behavioral therapy (CBT) for depression. These criteria should be predefined in a protocol (e.g., following PRISMA or Cochrane guidelines) to minimize bias. Here's a step-by-step guide to defining them:

Step 1: Establish Inclusion Criteria

• Population: Focus on adults (or specify age groups, e.g., 18+) diagnosed with depression. Use standardized criteria like DSM-5/ICD-11 major depressive disorder, or validated scales (e.g., Hamilton Depression Rating Scale score ≥17). Exclude comorbidities if they confound results (e.g., bipolar disorder unless specified).


• Intervention: Trials where CBT is the primary intervention (individual, group, or online formats). Specify duration (e.g., at least 8 sessions) and compare to control conditions like waitlist, treatment as usual (TAU), or active treatments (e.g., pharmacotherapy).


• Comparison: Active controls, placebo, or no treatment to assess CBT's efficacy.


• Outcomes: Primary: Depression symptom reduction (e.g., via Beck Depression Inventory [BDI] or Patient Health Questionnaire-9 [PHQ-9]). Secondary: Remission rates, quality of life, or relapse prevention.


• Study Design: Only RCTs with random allocation to groups. Quasi-experimental or non-randomized studies are ineligible.


• Publication Status: Include published and unpublished trials (e.g., gray literature) to reduce publication bias. No language restrictions if feasible, but prioritize English for practicality.


• Time Frame: Studies from a specific period (e.g., 1980–present) to capture CBT's evolution.

Step 2: Define Exclusion Criteria

• Non-RCT designs (e.g., cohort studies, case series).


• Samples not primarily depressed (e.g., anxiety disorders as primary).


• CBT combined inseparably with other therapies (e.g., CBT + mindfulness unless you plan a subgroup analysis).


• Short-term follow-up only (e.g., <6 weeks post-treatment) if your focus is long-term effects.


• Duplicate publications or subsets of larger trials.

Step 3: Operationalize and Document

• Use a PICO framework (Population, Intervention, Comparison, Outcome) to structure criteria.


• Create a screening form for title/abstract and full-text review (e.g., via tools like Covidence or Rayyan).


• Involve at least two independent reviewers to resolve disagreements (e.g., via Kappa statistic for inter-rater reliability >0.8).


• Pilot-test criteria on 10–20 studies to refine them.

2. Extracting Effect Sizes from Studies

Effect sizes quantify CBT's impact on depression (e.g., how much it reduces symptoms compared to control). Common metrics for continuous outcomes (depression scores) are standardized mean differences (SMD, e.g., Cohen's d) or Hedges' g (bias-corrected). For binary outcomes (e.g., remission), use odds ratios (OR) or risk ratios (RR). Follow these steps:

Step 1: Data Extraction Protocol

• Team and Tools: Have two reviewers extract data independently into a spreadsheet (e.g., Excel or Google Sheets) or software like RevMan/Comprehensive Meta-Analysis. Reconcile discrepancies.


• Key Data Fields:


• Study details: Authors, year, sample size (n per group), demographics (age, sex, baseline severity).


• Intervention: CBT type, duration, fidelity checks.


• Outcomes: Pre- and post-intervention means, SDs, or SEs for depression scales; event rates for binary outcomes; follow-up times.


• Quality indicators: Allocation concealment, blinding, attrition rates (use tools like Cochrane Risk of Bias 2).

Step 2: Calculate Effect Sizes

• For Continuous Outcomes (e.g., Mean Depression Scores):


• Formula for SMD: \( d = \frac{M1 - M2}{SD{pooled}} \), where \( M1 \) and \( M2 \) are intervention/control means, and \( SD{pooled} = \sqrt{\frac{(n1-1)SD1^2 + (n2-1)SD2^2}{n1 + n2 - 2}} \).


• Use Hedges' g for small samples: \( g = d \times \left(1 - \frac{3}{4(n1 + n2) - 9}\right) \).


• If data is missing (e.g., only p-values), impute using formulas or contact authors.


• Handle pre-post correlations (assume r=0.5 if unreported) for change scores.


• For Binary Outcomes:


• OR = \( \frac{(a/c)}{(b/d)} \), where a=remitted in CBT, b=non-remitted in CBT, c=remitted in control, d=non-remitted in control.


• Convert to SMD if mixing with continuous data (e.g., using Chapman's method).


• Software Recommendations:


• Free: R (metafor package) or RevMan for calculations.


• Paid: Comprehensive Meta-Analysis or Stata for advanced handling.

Step 3: Address Heterogeneity and Issues

• Subgroup Analyses: Extract moderators like CBT format (e.g., online vs. in-person) or depression severity.


• Missing Data: Note and use intention-to-treat (ITT) analyses where possible; sensitivity analyses for imputation.


• Effect Size Interpretation: SMD <0.2=small, 0.5=medium, >0.8=large. Report with 95% CIs.


• Quality Check: Assess for reporting biases (e.g., funnel plots later in meta-analysis).