Forest Plot Creation

This skill enables creation and interpretation of forest plots, the standard visualization for meta-analysis results.

Overview

A forest plot displays individual study results alongside the pooled estimate, showing effect sizes, confidence intervals, and study weights at a glance.

When to Use This Skill

Activate this skill when users:

• Ask to "create a forest plot"
• Want to visualize meta-analysis results
• Need to interpret forest plot components
• Ask about study weights or diamonds
• Mention "blobbogram" (alternative name)

Forest Plot Anatomy

Study            Effect [95% CI]        Weight
─────────────────────────────────────────────────
Smith 2020       ──■──                   15.2%
                 0.75 [0.52, 1.08]
                 
Jones 2021         ──■──                 22.1%
                   0.82 [0.65, 1.03]
                   
Lee 2022        ─■─                      8.4%
                0.45 [0.22, 0.92]
─────────────────────────────────────────────────
Overall         ◆                        100%
                0.72 [0.58, 0.89]
                
        ←Favors Treatment | Favors Control→
                    │
                   1.0 (null)

Key Components:

• Square (■): Point estimate for each study
• Horizontal line: 95% confidence interval
• Square size: Proportional to study weight
• Diamond (◆): Pooled effect estimate
• Diamond width: Confidence interval of pooled effect
• Vertical line: Line of no effect (OR=1, SMD=0)

R Code for Forest Plots

Basic Forest Plot

library(metafor)

# Example data: effect sizes and standard errors
dat <- data.frame(
  study = c("Smith 2020", "Jones 2021", "Lee 2022", "Chen 2023"),
  yi = c(-0.29, -0.20, -0.80, -0.35),  # log odds ratios
  sei = c(0.18, 0.12, 0.37, 0.15)       # standard errors
)

# Fit random-effects model
res <- rma(yi = yi, sei = sei, data = dat, method = "REML")

# Create forest plot
forest(res, 
       slab = dat$study,
       header = "Study",
       xlab = "Log Odds Ratio",
       mlab = "Random Effects Model")

Customized Forest Plot

# Enhanced forest plot with more options
forest(res,
       slab = dat$study,
       header = c("Study", "Log OR [95% CI]"),
       xlab = "Log Odds Ratio",
       mlab = "RE Model (DL)",
       xlim = c(-2.5, 1.5),
       alim = c(-1.5, 0.5),
       at = c(-1.5, -1, -0.5, 0, 0.5),
       refline = 0,
       col = "navy",
       border = "navy",
       fonts = 1,
       cex = 0.9)

# Add text annotations
text(-2.5, -1, "Favors Treatment", pos = 4, cex = 0.8)
text(1.5, -1, "Favors Control", pos = 2, cex = 0.8)

Forest Plot with Subgroups

# Add subgroup variable
dat$subgroup <- c("RCT", "RCT", "Observational", "RCT")

# Fit model with moderator
res_sub <- rma(yi = yi, sei = sei, mods = ~ subgroup, data = dat)

# Forest plot with subgroups
forest(res_sub,
       slab = dat$study,
       order = order(dat$subgroup),
       rows = c(1:2, 5:6),
       ylim = c(-1, 9),
       header = TRUE)

# Add subgroup labels
text(-2.5, c(3, 7), c("Observational", "RCT"), font = 2, pos = 4)

Teaching Interpretation

Guide Users Through Reading

1. Start with the overall effect (diamond)

   • "The diamond shows our pooled estimate"
   • "Its position tells us the direction of effect"
   • "Its width shows our uncertainty"

2. Check if it crosses the null line

   • "Does the diamond touch or cross the vertical line?"
   • "If not, the effect is statistically significant"

3. Examine individual studies

   • "Look at how studies cluster or spread"
   • "Wide spread = heterogeneity"
   • "Larger squares = more influential studies"

4. Assess consistency

   • "Do most studies point the same direction?"
   • "Are there outliers?"

Socratic Questions for Interpretation

• "What does the size of each square tell you?"
• "Why might one study have a much wider confidence interval?"
• "If the diamond crosses 1 (or 0), what does that mean?"
• "Looking at this plot, would you say the studies agree with each other?"

Common Issues and Solutions

Issue	Solution
Overlapping labels	Use rows argument to space studies
Axis too wide	Set xlim and alim manually
Studies in wrong order	Use order argument
Need back-transformed values	Use atransf = exp for odds ratios

Assessment Questions

1. Basic: "What does a larger square in a forest plot indicate?"

   • Correct: Greater study weight (usually from larger sample/smaller variance)

2. Intermediate: "The diamond doesn't touch the line at 1. What does this tell us?"

   • Correct: The pooled effect is statistically significant

3. Advanced: "Studies show effects on both sides of the null line. What should we investigate?"

   • Correct: Heterogeneity - examine I² and consider subgroup analysis

Related Skills

• meta-analysis-fundamentals - Understanding effect sizes
• heterogeneity-analysis - When studies disagree
• r-code-generation - Advanced R programming

Adaptation Guidelines

Glass (the teaching agent) MUST adapt this content to the learner:

1. Language Detection: Detect the user's language from their messages and respond naturally in that language
2. Cultural Context: Adapt examples to local healthcare systems and research contexts when relevant
3. Technical Terms: Maintain standard English terms (e.g., "forest plot", "effect size", "I²") but explain them in the user's language
4. Level Adaptation: Adjust complexity based on user's demonstrated knowledge level
5. Socratic Method: Ask guiding questions in the detected language to promote deep understanding
6. Local Examples: When possible, reference studies or guidelines familiar to the user's region

Example Adaptations:

• 🇧🇷 Portuguese: Use Brazilian health system examples (SUS, ANVISA guidelines)
• 🇪🇸 Spanish: Reference PAHO/OPS guidelines for Latin America
• 🇨🇳 Chinese: Include examples from Chinese medical literature