ITC Code Review - Expert R Code Analysis and Amendment

Review ITC analysis code for: $ARGUMENTS

Purpose

Comprehensive review of R code for indirect treatment comparison analyses (pairwise MA, NMA, MAIC, STC, ML-NMR) by an expert combining world-class ITC methodology knowledge with rigorous software engineering practices.

Review Options

Specify your preferences:

• review_type: comprehensive (detailed) | summary (bullet points)
• action: review_only | amend (create corrected code in subfolder)
• output_folder: Path for amended files (default: ./amended/)
• focus_areas: Specific concerns (optional)

Review Workflow

<Task> subagent_type: itc-code-reviewer prompt: | Conduct expert review of ITC analysis code: $ARGUMENTS

Review Configuration

• Review type: {review_type} (comprehensive or summary)
• Action: {action} (review_only or amend)
• Output folder: {output_folder}
• Focus areas: {focus_areas}

Step 1: Understand the Analysis

Before reviewing, understand:

• What ITC method is being used?
• What is the research question?
• What data structure is present?
• What packages are being used?

Step 2: Methodological Review

Check for issues in these areas:

Pairwise Meta-Analysis (meta, metafor, bayesmeta)

• Correct effect measure for outcome type
• Appropriate fixed vs random effects choice
• Heterogeneity properly assessed (Q, I², τ², prediction interval)
• Publication bias addressed (if ≥10 studies)
• Sensitivity analyses included
• Forest plot generated
• Knapp-Hartung adjustment used (if applicable)

Network Meta-Analysis (netmeta, gemtc)

• Network connectivity verified
• Transitivity assumption assessed
• Consistency tested (node-splitting, net heat)
• Multi-arm correlations handled
• Rankings reported with uncertainty
• Convergence checked (Bayesian)
• Priors specified and justified (Bayesian)

MAIC (maicplus)

• Covariate selection justified
• ESS calculated and acceptable
• Weight diagnostics checked
• Covariate balance verified
• Anchored vs unanchored appropriate
• Bootstrap CI used
• Comparison with STC (sensitivity)

STC (stc)

• Effect modifiers identified and justified
• Covariates centered on AgD population
• Treatment-covariate interactions included
• Robust SE used
• Model diagnostics checked
• Comparison with MAIC (sensitivity)

ML-NMR (multinma)

• Integration points sufficient
• Convergence diagnostics (R-hat, ESS)
• Priors specified and sensitivity tested
• Node-splitting for consistency
• Target population specified
• Predictions properly marginalized

Step 3: Statistical Review

• Correct variance calculations
• Proper confidence/credible intervals
• Appropriate hypothesis tests
• Multiple testing addressed
• Effect size interpretation correct
• Uncertainty properly propagated

Step 4: Code Quality Review

Tidy Modelling Compliance (TMwR)

• Consistent interfaces
• Recipe-like data preparation
• Results as tibbles
• Workflow structure clear

R Best Practices

• Vectorization over loops
• Appropriate functions used
• Memory efficiency
• Error handling present
• Input validation

Reproducibility

• Seeds set for stochastic operations
• Package versions documented
• Complete session info available
• Data provenance clear

Documentation

• Functions documented
• Code comments explain "why"
• Analysis narrative present
• Results interpretation included

Step 5: Generate Output

If review_type = "comprehensive":

Provide detailed analysis:

• Line-by-line issues with file:line references
• Methodology background for each issue
• Complete corrected code examples
• References to guidelines

If review_type = "summary":

Provide concise output:

• Bullet-point issue list
• Priority categorization (Critical/Major/Minor)
• One-line recommendations

If action = "amend":

Create files in output_folder:

{output_folder}/
├── {filename}_amended.R    # Corrected code
├── REVIEW_NOTES.md         # Change summary
└── backup/
    └── {filename}_original.R

Step 6: REVIEW_NOTES.md Structure

# ITC Code Review - {date}

## Summary
- Total issues: X (Y Critical, Z Major, W Minor)
- Files reviewed: [list]
- Files amended: [list]

## Critical Issues
[Detailed description of each critical issue]

## Major Issues
[Detailed description of each major issue]

## Minor Issues
[Brief description of minor issues]

## Recommendations
- Additional analyses to consider
- Best practice improvements

## Changes Made (if amended)
- [List of specific changes per file]

## Package Versions
[Document all package versions used]

Issue Priority Definitions

Critical: Could invalidate results

• Wrong method for data structure
• Major assumption violation
• Incorrect statistical calculation
• Missing essential diagnostics

Major: Significantly impacts quality

• Missing important sensitivity analysis
• Suboptimal method choice
• Incomplete diagnostics
• Poor reproducibility

Minor: Best practice improvements

• Code style issues
• Documentation gaps
• Efficiency improvements
• Minor formatting issues

Proceed with the review following these guidelines. </Task>

Output Deliverables

Based on configuration, the review will provide:

review_only Mode

• Comprehensive or summary review document
• Issue list with priorities
• Specific recommendations with code examples
• No files modified

amend Mode

All of the above, plus:

• Amended code files in subfolder
• REVIEW_NOTES.md with change summary
• Backup of original files
• Original files NEVER modified

Example Usage

# Comprehensive review only
Review my NMA analysis in analysis/nma_analysis.R
review_type: comprehensive
action: review_only

# Summary review with amendments
Review my MAIC code and fix issues
review_type: summary
action: amend
output_folder: ./reviewed/

# Focused review
Review the convergence diagnostics in my ML-NMR analysis
focus_areas: convergence, priors

Safety Guarantees

• Original files are NEVER modified directly
• All amendments go to specified subfolder
• Backup copies created before any changes
• Clear documentation of all modifications
• Reversible changes only