Meta-Analysis Manuscript Assembly

Complete systematic review and meta-analysis manuscripts for journal submission by creating publication-ready tables, figures, and references.

When to Use

Completing meta-analysis manuscript after analyses are done
Creating tables from meta-analysis results
Assembling multi-panel figures from forest/funnel plots
Generating BibTeX references for systematic reviews
Formatting manuscripts for high-impact journals (Lancet, JAMA, NEJM)

Prerequisites

Before using this skill, ensure you have:

Completed meta-analyses with results tables (CSV format)
Generated individual figures (PNG at 300 DPI)
Manuscript text sections written (Abstract, Introduction, Methods, Results, Discussion)
List of all citations needed

Workflow

Phase 1: Tables Creation

Create comprehensive tables from analysis results:

Main Text Tables

Table 1: Trial Characteristics
- Extract from extraction.csv or similar
- Include: NCT number, first author, year, design, sample sizes, intervention details, follow-up
- Format as markdown with abbreviations section
Table 2: Efficacy Outcomes Summary
- Combine results from all meta-analyses (pCR, survival outcomes)
- Include: effect estimates, 95% CI, p-values, I², absolute benefits, NNT
- Add GRADE certainty ratings
Table 3: Safety Outcomes Summary
- From safety meta-analysis results
- Include: adverse events, rates, RR/OR, NNH
- Add clinical management guidance

Supplementary Tables

Risk of Bias Assessment
- Use RoB 2 or ROBINS-I tool format
- Domain-by-domain assessment for each trial
- Overall risk rating with justifications
GRADE Evidence Profile
- Summary of Findings table format
- All outcomes with certainty ratings
- Domain-specific justifications (bias, inconsistency, indirectness, imprecision)
Detailed Results Tables
- Individual trial results
- Subgroup analyses
- Sensitivity analyses

Phase 2: Figure Assembly

Create multi-panel publication-ready figures:

Tool: Python Script with PIL/Pillow

# Create assemble_figures.py
from PIL import Image, ImageDraw, ImageFont
from pathlib import Path

def add_panel_label(img, label, position='top-left', font_size=80, offset=(40, 40)):
    """Add A, B, C labels to panels"""
    draw = ImageDraw.Draw(img)

    # Try to use system font
    try:
        font = ImageFont.truetype("/System/Library/Fonts/Helvetica.ttc", font_size)
    except:
        font = ImageFont.load_default()

    x, y = offset

    # Draw white background box for visibility
    bbox = draw.textbbox((x, y), label, font=font)
    padding = 10
    draw.rectangle(
        [bbox[0] - padding, bbox[1] - padding,
         bbox[2] + padding, bbox[3] + padding],
        fill='white',
        outline='black',
        width=2
    )

    draw.text((x, y), label, fill='black', font=font)
    return img

def create_multi_panel_figure(images_list, output_path, labels=['A', 'B', 'C'], spacing=40):
    """Combine multiple images vertically with labels"""
    # Add labels to images
    labeled_images = [add_panel_label(img, label) for img, label in zip(images_list, labels)]

    # Calculate dimensions
    max_width = max(img.width for img in labeled_images)
    total_height = sum(img.height for img in labeled_images) + spacing * (len(labeled_images) - 1)

    # Create combined image
    combined = Image.new('RGB', (max_width, total_height), 'white')

    # Paste images
    y_offset = 0
    for img in labeled_images:
        combined.paste(img, (0, y_offset))
        y_offset += img.height + spacing

    # Save at 300 DPI
    combined.save(output_path, dpi=(300, 300))
    return output_path

Typical Figure Structure

Main Text:

Figure 1: Multi-panel efficacy (pCR, EFS, OS forest plots)
Figure 2: Subgroup analysis (e.g., by biomarker status)
Figure 3: Safety + Publication bias (SAE forest plot, funnel plot)

Supplementary:

Supp Figure 1: Sensitivity analyses (leave-one-out plots)
Supp Figure 2: Publication bias (funnel plots for all outcomes)

Phase 3: References Management

Create comprehensive BibTeX file:

Steps:

Extract all citations from manuscript using grep

grep -E "¹|²|³|⁴|⁵|⁶|⁷|⁸|⁹|⁰|\[\d+\]" manuscript_sections.md

Create BibTeX entries for each reference
- Include DOI for all entries
- Use standardized journal abbreviations (Index Medicus)
- Format author names correctly
Create mapping document
- Map superscripts (¹, ², ³) to BibTeX keys
- Document citation locations in manuscript
Create usage guide
- Pandoc conversion instructions
- Zotero import instructions
- Manual formatting examples (Lancet, JAMA style)

Phase 4: Figure Legends

Write comprehensive legends for all figures:

Legend Structure:

**Panel A. Outcome Name**
Description of what the panel shows. Forest plot showing [effect measure] for [outcome]
across [N] trials ([total participants]). [Statistical method used]. [Key result].
Horizontal lines represent 95% confidence intervals; diamond represents pooled effect.
Vertical line at [null value] indicates no treatment effect.

**Abbreviations**: List all abbreviations used.

Include:

Statistical methods (random-effects, Hartung-Knapp adjustment)
Heterogeneity measures (I², Cochran's Q)
Clinical interpretations
Abbreviations definitions

Phase 5: Quality Assurance

Before submission, verify:

Tables

All data matches analysis results exactly
Abbreviations defined
Footnotes explain all symbols
Column/row headers clear
Statistical notation consistent

Figures

All figures at 300 DPI minimum
Panel labels (A, B, C) visible and not obscuring data
Legends match figures exactly
Font sizes readable (≥8pt for final print size)
Color schemes work in grayscale

References

All citations have corresponding references
Reference numbers sequential
DOIs correct and working
Journal abbreviations standardized
Author names match original publications

Output Structure

07_manuscript/
├── tables/
│   ├── Table1_Trial_Characteristics.md
│   ├── Table2_Efficacy_Summary.md
│   ├── Table3_Safety_Summary.md
│   ├── SupplementaryTable1_RiskOfBias.md
│   ├── SupplementaryTable2_GRADE_Profile.md
│   └── ...
├── figures/
│   ├── Figure1_Efficacy.png (300 DPI)
│   ├── Figure2_Subgroup.png (300 DPI)
│   ├── Figure3_Safety.png (300 DPI)
│   ├── SupplementaryFigure1_Sensitivity.png
│   └── ...
├── references.bib
├── FIGURE_LEGENDS.md
├── CITATION_MAPPING.md
└── REFERENCES_USAGE_GUIDE.md

Time Estimates

Tables creation: 2-3 hours
Figure assembly: 1-2 hours
References: 1-2 hours
Legends: 1 hour
QA: 1 hour
Total: 6-9 hours

Journal-Specific Formatting

Lancet Oncology

Word limit: 4000-5000 words
Tables: 3-4 main text, unlimited supplementary
Figures: 3-4 main text, unlimited supplementary
References: Vancouver style, 30-40 typical
Resolution: 300 DPI minimum

JAMA

Word limit: 3500 words
Tables: 4 max
Figures: 4 max
References: 40 max
Resolution: 300-600 DPI

New England Journal of Medicine

Word limit: 3000 words
Tables: 3 max
Figures: 3 max
References: 40 max
Resolution: 300 DPI minimum

Common Pitfalls to Avoid

Tables: Don't mix effect measures (RR vs OR vs HR) without clear labeling
Figures: Don't compress below 300 DPI
References: Don't use auto-generated citations without verification
Legends: Don't omit statistical methods or abbreviations
Overall: Don't submit without independent verification of all numbers

Related Skills

/meta-analysis - Perform the statistical analyses
/prisma-flow - Create PRISMA flow diagram
/grade-assessment - Complete GRADE evidence profiles
/risk-of-bias - Assess trial quality with RoB 2 tool

Example Invocation

/meta-manuscript-assembly

Or with specific phase:

/meta-manuscript-assembly tables
/meta-manuscript-assembly figures
/meta-manuscript-assembly references

Success Criteria

✅ All tables publication-ready with comprehensive notes
✅ All figures 300 DPI with professional panel labels
✅ Complete BibTeX file with all 30-40 references
✅ Comprehensive figure legends
✅ All numbers verified against original analyses
✅ Manuscript follows target journal guidelines
✅ Ready for co-author review and submission