2f:Tc05,introduction = md`

Introduction

Lifestyle changes play a pivotal role in promoting healthy cognitive aging and maintaining cognitive function as individuals grow older. Scientific evidence consistently highlights the significant impact of various lifestyle factors, such as regular physical exercise, cognitive engagement, and balanced nutrition, on the brain's health and cognitive abilities. Engaging in these positive lifestyle choices can help enhance cognitive resilience, reduce the risk of cognitive decline, and potentially delay the onset of age-related cognitive disorders, such as dementia.

The Research Gap

While individual studies have yielded valuable insights into the effectiveness of specific lifestyle interventions, there remains a notable gap in comprehensive and systematic research that directly compares the relative benefits of these interventions. Direct comparisons based on meta-analyses of individual approaches are challenging because such meta-analyses often use different inclusion criteria and analytical methods. Adopting a consistent approach is crucial to gain a deeper understanding of the effects of behavioral interventions to support healthy cognitive aging.

Study Objectives

This meta-analysis aims to evaluate and contrast the effectiveness of three primary types of behavioral lifestyle interventions—physical exercise (PE), cognitive training (CT), and nutrition (NU)—along with specific intervention subtypes and control types in older adults without dementia.

Note: This analysis is supported by an accompanying open-source, interactive database website designed for regular updates and the exploration of potential interactions between interventions and their modules.

`30:T18cb,methods = md`

Methods

Search Strategy

As the first step in building a meta-analysis database, online searches were conducted using Google Scholar to identify previous meta-analysis papers on lifestyle interventions. The selection process focused on the most recent meta-analysis studies (published in 2022 or earlier) to extract their reference lists, assuming they would include the most up-to-date studies. Reference lists of the selected meta-analyses were then reviewed to identify relevant training studies.

Eligibility Criteria

• Population: Included adults over the age of 50 without dementia, including those with normal cognitive aging and Mild Cognitive Impairment (MCI). Participants with dementia or other significant neurological conditions were excluded.

• Interventions: Included cognitive training, physical exercise, and nutrition. Combination interventions (e.g., physical activity paired with cognitive training) were excluded.

• Comparators: Included active controls, treatment-as-usual controls, and no-contact controls. Studies comparing two active interventions without a clear control group were excluded.

• Study Design: Included randomized controlled trials (RCTs) with one or more repeated standardized cognitive outcome measures. Single-arm feasibility trials and studies lacking rigorous randomization were excluded.

Data Collection and Coding

For each included study, a series of variables were coded for comprehensive analysis across three categories:

• Participant Characteristics: Percentage female, average age, and average years of education.
• Intervention Characteristics: Type of intervention (Cognitive Training, Physical Exercise, Nutrition), intervention sub-types, total hours of training, and total duration (weeks).
• Control Conditions: Classification as active or passive controls and the specific nature of active control interventions.

Data Synthesis Plan

Effect Size Calculation

All analyses were conducted using net gain effect sizes, based on the recommended pretest-posttest-control formula. The effect sizes were calculated such that positive numbers indicate greater gains in the training group compared to the control group from pre-training to post-training.

For each cognitive outcome in each study, Cohen’s d was calculated as follows:

${tex.block`d = \frac{(M_{epost} - M_{epre}) - (M_{cpost} - M_{cpre})}{SD_{pre}}`}

The pooled pre-training standard deviation ($SD_{pre}$) was calculated using baseline training and control group standard deviations:

${tex.block`SD_{pre} = \sqrt{\frac{(n_e - 1)SD_{epre}^2 + (n_c - 1)SD_{cpre}^2}{n_e + n_c - 2}}`}

Bias Correction and Weighting

To correct for small sample size bias, the final effect sizes used in analyses were Hedges’ g:

${tex.block`g = d \times \left(1 - \frac{3}{4(n_e + n_c - 2) - 1}\right)`}

Net gain effect sizes were then weighted by their inverse variance to calculate pooled effect sizes, Q, and I² statistics. The pooled (net) effect size was calculated as:

${tex.block`\bar{g} = \frac{\sum w_i g_i}{\sum w_i}`}

Statistical Modeling and Heterogeneity

Random-effects models with Hartung-Knapp adjustments were utilized to estimate pooled standardized mean differences (SMD). Stratified subgroup analyses evaluated the influence of intervention type, clinical status, control design (active vs. passive), and cognitive outcome constructs. Moderator analyses examined the impact of demographics and study-level properties on intervention success. The heterogeneity of effect sizes was evaluated using the Q statistic and the I² statistic, with I² values of 75%, 50%, and 25% regarded as indicators of high, moderate, and low heterogeneity, respectively.

`31:T1050,styledTable = { const authorWidth = 220; const yearWidth = 80; // 1. We define the table without 'viewof'. // We also pass an empty array to 'value' to signal no selection is active. const table = Inputs.table(data, { required: false, value: [], layout: "auto", width: { "Author": authorWidth, "Year": yearWidth }, rows: 18, format: { "Year": d => d.toString().replace(/,/g, ""), "SMD": d => d?.toFixed ? d.toFixed(2) : d, "n_total": d => d?.toLocaleString ? d.toLocaleString() : d } }); return html`

📚 Primary Study Database ${data.length} STUDIES LOADED

${table}

`; }32:T190f,viewof webInterventionFigure = { // 1. Process the live 'data' variable dynamically const groupedData = d3.group(data, d => d["Study Type"]); // UPDATED COLOR PALETTE: Matches specific figure requirements const colorMap = { "Cognitive Training": "#2196F3", // Blue "Physical Exercise": "#FF9800", // Orange "Nutrition": "#FFD700", // Yellow "Unspecified": "#9E9E9E" // Gray }; const chartData = Array.from(groupedData, ([studyType, group]) => { const typeName = studyType || "Unspecified"; const getMeanSD = (key, isPercent = false) => { const validNumbers = group.map(d => parseFloat(d[key])).filter(n => !isNaN(n)); if (validNumbers.length === 0) return "N/A"; const mean = d3.mean(validNumbers); const sd = d3.deviation(validNumbers) || 0; const suffix = isPercent ? "%" : ""; return `${mean.toFixed(1)}${suffix} ± ${sd.toFixed(1)}${suffix}`; }; const getSum = (key) => { const sum = d3.sum(group, d => parseFloat(d[key]) || 0); return sum > 0 ? sum : "N/A"; }; return { id: typeName, studies: group.length, totalN: getSum("Total N"), age: getMeanSD("Age"), female: getMeanSD("% Female", true), education: getMeanSD("Education"), measures: getMeanSD("Num of Measures"), weeks: getMeanSD("Total Weeks"), color: colorMap[typeName] || "#607d8b" }; }); chartData.sort((a, b) => b.studies - a.studies); const totalStudies = d3.sum(chartData, d => d.studies); // 2. Generate the D3 Donut Chart (SCALED FOR WEB) const width = 350; const height = 350; const radius = Math.min(width, height) / 2; const pie = d3.pie().value(d => d.studies).sort(null); const arc = d3.arc().innerRadius(radius * 0.55).outerRadius(radius * 0.9); const labelArc = d3.arc().innerRadius(radius * 0.72).outerRadius(radius * 0.72); const svg = d3.create("svg") .attr("width", width) .attr("height", height) .attr("viewBox", [-width / 2, -height / 2, width, height]) .style("font-family", "system-ui, sans-serif"); svg.selectAll("path") .data(pie(chartData)) .join("path") .attr("fill", d => d.data.color) .attr("d", arc) .attr("stroke", "white") .attr("stroke-width", "2px"); // Thinner stroke for smaller chart svg.selectAll("text.label") .data(pie(chartData)) .join("text") .attr("class", "label") .attr("transform", d => `translate(${labelArc.centroid(d)})`) .attr("text-anchor", "middle") .attr("dy", "0.35em") .style("fill", d => (d.data.id === "Nutrition" ? "#495057" : "white")) .style("font-weight", "bold") .style("font-size", "16px") // Smaller pie labels .text(d => d.data.studies); svg.append("text") .attr("text-anchor", "middle") .attr("dy", "-0.1em") .style("font-size", "42px") // Smaller center text .style("font-weight", "800") .style("fill", "#343a40") .text(totalStudies); svg.append("text") .attr("text-anchor", "middle") .attr("dy", "1.6em") .style("font-size", "14px") // Smaller sub-text .style("font-weight", "700") .style("fill", "#6c757d") .style("text-transform", "uppercase") .style("letter-spacing", "1px") .text("Studies"); // 3. Build the Details Cards (SCALED FOR WEB) const detailsHTML = chartData.map(d => `

${d.id}

N = ${d.totalN.toLocaleString()}

Age ${d.age}

% Female ${d.female}

Education ${d.education}

Measures ${d.measures}

Total Weeks ${d.weeks}

`).join(""); // 4. Assemble Final Layout (MAX-WIDTH 1000px) return html`

📊 Intervention Distribution & Baseline Characteristics

${svg.node()}

${detailsHTML}

`; }33:T1c58,viewof interventionSummary = { // 1. Reusable Data Calculator const buildSummary = (dataset) => { return Array.from(d3.group(dataset, d => d["Study Type"]), ([studyType, group]) => { const getMeanSD = (key, isPercent = false) => { const validNumbers = group.map(d => parseFloat(d[key])).filter(n => !isNaN(n)); if (validNumbers.length === 0) return "N/A"; const mean = d3.mean(validNumbers); const sd = d3.deviation(validNumbers) || 0; const suffix = isPercent ? "%" : ""; return `${mean.toFixed(1)}${suffix} ± ${sd.toFixed(1)}${suffix}`; }; const getSum = (key) => { const sum = d3.sum(group, d => parseFloat(d[key]) || 0); return sum > 0 ? sum : "N/A"; }; return { "Study Type": studyType || "Unspecified", "Studies (k)": group.length, "Total N": getSum("Total N"), "Age": getMeanSD("Age"), "% Female": getMeanSD("% Female", true), "Education": getMeanSD("Education"), "Num of Measures": getMeanSD("Num of Measures"), "Total Weeks": getMeanSD("Total Weeks") }; }); }; // 2. Generate the 3 distinct datasets const mciData = data.filter(d => d["Population"] === "MCI"); const haData = data.filter(d => d["Population"] === "Healthy Aging"); const tableTotal = buildSummary(data); const tableMCI = buildSummary(mciData); const tableHA = buildSummary(haData); const typeWidth = 220; // 3. Reusable HTML Table Generator const buildTableHTML = (tableData, title, count) => { if (tableData.length === 0) return ''; const rowsHTML = tableData.map(row => ` ${row["Study Type"]} ${row["Studies (k)"]} ${row["Total N"]} ${row["Age"]} ${row["% Female"]} ${row["Education"]} ${row["Num of Measures"]} ${row["Total Weeks"]} `).join(''); return `

📊 ${title} ${count} STUDIES

${rowsHTML}

Study Type	Studies (k)	Total N	Age	% Female	Education	Num of Measures	Total Weeks

`; }; // 4. Final Unified Output return html`

Baseline Characteristics of Included Studies across Subgroups

The summary tables break down demographic metrics and trial designs across the entire dataset, stratified by target clinical profile. These matrices isolate sample sizes, mean age thresholds, sex distributions, formal educational backdrops, outcome density parameters, and overall timeline footprints. This configuration exposes sample variance and cross-study boundaries across the primary randomized configurations.

${buildTableHTML(tableTotal, "Total Population (All Studies)", data.length)} ${buildTableHTML(tableMCI, "Subgroup: Mild Cognitive Impairment (MCI)", mciData.length)} ${buildTableHTML(tableHA, "Subgroup: Healthy Aging", haData.length)}

`; }34:T1c9a,viewof yColumn = { // 1. Structural layout using standard HTML inputs for 100% reliable state control const container = html`

Checking Baseline Differences across Interventions

`; const typeBoxes = Array.from(container.querySelectorAll(".type-container input")); const popBoxes = Array.from(container.querySelectorAll(".pop-container input")); const controlBoxes = Array.from(container.querySelectorAll(".control-container input")); const selectDropdown = container.querySelector(".variable-select"); // 2. State storage matrices let selectedTypes = ["All Data"]; let selectedPops = ["All Populations"]; let selectedControls = ["All Control Types"]; // 3. Robust Mutual Exclusion & Opacity Controller Function const syncRow = (boxes, lastSelection, masterValue) => { let checkedValues = boxes.filter(b => b.checked).map(b => b.value); // Context A: User just checked a sub-option while "All" was previously active if (lastSelection.includes(masterValue) && checkedValues.length > 1) { checkedValues = checkedValues.filter(v => v !== masterValue); } // Context B: User specifically checked the "All" master option else if (!lastSelection.includes(masterValue) && checkedValues.includes(masterValue)) { checkedValues = [masterValue]; } // Context C: Nothing is checked -> Default fallback to "All" if (checkedValues.length === 0) { checkedValues = [masterValue]; } // Reflect state adjustments accurately back onto the DOM const isMasterActive = checkedValues.includes(masterValue); boxes.forEach(b => { b.checked = checkedValues.includes(b.value); b.parentElement.style.opacity = (isMasterActive && b.value !== masterValue) ? "0.4" : "1"; }); return checkedValues; }; // 4. Reactive Pipeline Update Dispatcher const emitValue = () => { const finalTypes = selectedTypes.includes("All Data") ? ["Cognitive Training", "Nutrition", "Physical Exercise"] : selectedTypes; const finalPops = selectedPops.includes("All Populations") ? ["Healthy Aging", "MCI"] : selectedPops; const finalControls = selectedControls.includes("All Control Types") ? ["Active", "Passive"] : selectedControls; container.value = { variable: selectDropdown.value, selectedStudyTypes: finalTypes, selectedPopulations: finalPops, selectedControlTypes: finalControls }; container.dispatchEvent(new CustomEvent("input", { bubbles: true })); }; // 5. Universal event delegation assignment container.addEventListener("change", (event) => { if (typeBoxes.includes(event.target)) { selectedTypes = syncRow(typeBoxes, selectedTypes, "All Data"); } if (popBoxes.includes(event.target)) { selectedPops = syncRow(popBoxes, selectedPops, "All Populations"); } if (controlBoxes.includes(event.target)) { selectedControls = syncRow(controlBoxes, selectedControls, "All Control Types"); } emitValue(); }); // Run initialization setup emitValue(); return container; }35:T13e3,rStatisticalResultText = { await webr; // 1. Unpack properties from the updated multi-filter yColumn object const targetVariable = yColumn.variable; const allowedStudyTypes = yColumn.selectedStudyTypes || []; const allowedPopulations = yColumn.selectedPopulations || []; const allowedControlTypes = yColumn.selectedControlTypes || []; // HARD LOGIC GATE: Skip WebR processing entirely if fewer than 2 study types are active if (allowedStudyTypes.length < 2) { return { mainEffect: "Not enough data for statistical testing within current filtered boundaries.", postHoc: "" }; } // 2. Clean and filter the data dynamically across all conditions const cleanData = data .filter(d => { const yVal = d[targetVariable]; const groupVal = d["Study Type"]; const popVal = d["Population"]; const controlVal = d["Control Class"]; // Baseline validity checks const hasValidY = yVal !== null && yVal !== undefined && String(yVal).trim() !== "" && String(yVal).trim() !== "NA"; const hasValidGroup = groupVal !== null && groupVal !== undefined && String(groupVal).trim() !== ""; if (!hasValidY || !hasValidGroup) return false; // Match against selection matrices const matchesStudy = allowedStudyTypes.includes(groupVal); const matchesPop = allowedPopulations.includes(popVal); // Multi-select fallback check for raw control array strings const matchesControl = allowedControlTypes.some(c => String(controlVal).includes(c)); return matchesStudy && matchesPop && matchesControl; }) .map(d => { let val = d[targetVariable]; if (typeof val !== "number") val = parseFloat(String(val).replace(/[%,]/g, "")); return { group: String(d["Study Type"]), value: val }; }) .filter(d => !isNaN(d.value)); const groups = d3.group(cleanData, d => d.group); if (groups.size < 2 || cleanData.length <= groups.size) { return { mainEffect: "Not enough data for statistical testing within current filtered boundaries.", postHoc: "" }; } // 3. Extract matching arrays for WebR context execution const dep_var_js = cleanData.map(d => d.value); const grp_var_js = cleanData.map(d => d.group); // 4. Run the R script engine const result = await webr.evalR(` dep_var <- as.numeric(dep_var) grp_var <- as.factor(grp_var) anova_res <- aov(dep_var ~ grp_var) model_residuals <- residuals(anova_res) shapiro_res <- shapiro.test(model_residuals) if (shapiro_res$p.value > 0.05) { # --- ANOVA --- summary_anova <- summary(anova_res) p_anova <- summary_anova[[1]][["Pr(>F)"]][1] f_val <- summary_anova[[1]][["F value"]][1] df1 <- summary_anova[[1]][["Df"]][1] df2 <- summary_anova[[1]][["Df"]][2] p_fmt <- ifelse(p_anova < 0.001, "p < .001", paste0("p = ", sub("^0+", "", sprintf("%.3f", p_anova)))) res_text <- sprintf("ANOVA: F(%d, %d) = %.2f, %s", df1, df2, f_val, p_fmt) if (p_anova < 0.05) { tukey <- TukeyHSD(anova_res) p_vals <- tukey[[1]][,"p adj"] sig_pairs <- names(p_vals)[p_vals < 0.05] if (length(sig_pairs) > 0) { res_text <- paste0(res_text, " | Significant Pairs: ", paste(sig_pairs, collapse = ", ")) } else { res_text <- paste0(res_text, " | No significant pairwise differences") } } res_text } else { # --- Kruskal-Wallis --- kruskal_res <- kruskal.test(dep_var ~ grp_var) p_kruskal <- kruskal_res$p.value h_val <- kruskal_res$statistic df <- kruskal_res$parameter p_fmt <- ifelse(p_kruskal < 0.001, "p < .001", paste0("p = ", sub("^0+", "", sprintf("%.3f", p_kruskal)))) res_text <- sprintf("Kruskal-Wallis: H(%d) = %.2f, %s", df, h_val, p_fmt) if (p_kruskal < 0.05) { wilcox <- pairwise.wilcox.test(dep_var, grp_var, p.adjust.method = "bonferroni") p_matrix <- wilcox$p.value sig_pairs <- c() for (i in 1:nrow(p_matrix)) { for (j in 1:ncol(p_matrix)) { if (!is.na(p_matrix[i, j]) && p_matrix[i, j] < 0.05) { sig_pairs <- c(sig_pairs, paste0(rownames(p_matrix)[i], "-", colnames(p_matrix)[j])) } } } if (length(sig_pairs) > 0) { res_text <- paste0(res_text, " | Significant Pairs: ", paste(sig_pairs, collapse = ", ")) } else { res_text <- paste0(res_text, " | No significant pairwise differences") } } res_text } `, { env: { dep_var: dep_var_js, grp_var: grp_var_js } }); // 5. Convert output array configuration and clean up memory const outputArray = await result.toArray(); webr.destroy(result); // 6. Split string values into primary and tracking metrics const rawText = outputArray[0]; const parts = rawText.split(" | "); return { mainEffect: parts[0], postHoc: parts.length > 1 ? parts[1] : "" }; }36:T1bad,viewof yColumnAnalysis = { // 1. UNPACK TARGET FILTERS & MULTI-SELECT STRATIFICATION MATRICES const targetVariable = yColumn.variable; const allowedStudyTypes = yColumn.selectedStudyTypes || []; const allowedPopulations = yColumn.selectedPopulations || []; const allowedControlTypes = yColumn.selectedControlTypes || []; // HARD LOGIC GATE: Require at least two distinct Study Types to run the analysis if (allowedStudyTypes.length < 2) { return html`

Statistical Summary: ${targetVariable}

⚠️ Comparative analysis and distribution mapping require at least two "Study Type" filters to be active simultaneously. Please adjust your selections above.

`; } // 2. UNIFIED DATA PROCESSING PIPELINE (Runs only if gate passes) const filteredData = data.filter(d => { const yVal = d[targetVariable]; const groupVal = d["Study Type"]; const popVal = d["Population"]; const controlVal = d["Control Class"]; const hasValidData = yVal !== null && yVal !== undefined && String(yVal).trim() !== "" && String(yVal).trim() !== "NA" && groupVal !== null && groupVal !== undefined && String(groupVal).trim() !== ""; if (!hasValidData) return false; // Filter validation checks const matchesStudy = allowedStudyTypes.includes(groupVal); const matchesPop = allowedPopulations.includes(popVal); const matchesControl = allowedControlTypes.some(c => String(controlVal).includes(c)); return matchesStudy && matchesPop && matchesControl; }); // Group metrics computation using target parsed numerical floats const groupStats = d3.groups(filteredData, d => d["Study Type"]).map(([type, values]) => { const rawValues = values.map(d => { const val = d[targetVariable]; return typeof val === "number" ? val : parseFloat(String(val).replace(/[%,]/g, "")); }).filter(v => !isNaN(v)); return { type, mean: d3.mean(rawValues) || 0, sd: d3.deviation(rawValues) || 0, k: values.length }; }); // Safely parse significance flags only if a valid statistical calculation happened const statsString = rStatisticalResultText?.mainEffect || ""; const hasValidStats = statsString.includes("ANOVA:") || statsString.includes("Kruskal-Wallis:"); let isSig = false; if (hasValidStats) { if (statsString.includes("p < .001")) { isSig = true; } else if (statsString.includes("p = .") || statsString.includes("p = 0.")) { const pParts = statsString.split(/p\s*=\s*\.?/); if (pParts.length > 1) { const pVal = parseFloat("0." + pParts[1].replace(/[^0-9]/g, "")); if (!isNaN(pVal) && pVal < 0.05) isSig = true; } } } // 3. THE STATISTICAL RESULTS SUMMARY BLOCK const resultsTable = html`

Statistical Summary: ${targetVariable}

${groupStats.map((g, i) => html`

${g.type}

${g.mean.toFixed(2)} ±${g.sd.toFixed(1)}

k = ${g.k}

`)}

Statistical Engine Output

${statsString}

Post-Hoc Pairwise Trends

${hasValidStats && isSig && rStatisticalResultText.postHoc ? rStatisticalResultText.postHoc : "No pairwise differences evaluated (requires a significant main effect across multiple active groups)."}

`; // 4. THE DISTRIBUTION BOXPLOT FIGURE const figure = Plot.plot({ width: 1100, height: 500, style: { fontSize: "18px", fontFamily: "system-ui, sans-serif" }, marginLeft: 100, marginBottom: 80, x: { label: "Study Type →", labelOffset: 55, tickSize: 10, domain: allowedStudyTypes }, y: { grid: true, label: `↑ ${targetVariable}`, labelOffset: 80, nice: true, labelAnchor: "center" }, color: { domain: ["Cognitive Training", "Nutrition", "Physical Exercise"], range: ["#4682B4", "#FFA500", "#FF6347"] }, marks: [ Plot.boxY(filteredData, { x: "Study Type", y: d => { const val = d[targetVariable]; return typeof val === "number" ? val : parseFloat(String(val).replace(/[%,]/g, "")); }, fill: "Study Type", fillOpacity: 0.6, stroke: "Study Type", strokeWidth: 2 }) ] }); return html`

${resultsTable}

${figure}

`; }37:Tcaa,viewof selectedTypes = { // 1. Explicitly define your desired sequence const options = ["Cognitive Training", "Nutrition", "Physical Exercise"]; // 2. Create the input using those specific options const checkboxes = Inputs.checkbox( options, { label: html`Filter Study Type:`, value: options // Default to all selected } ); // 3. Styling the checkboxes checkboxes.style.fontSize = "18px"; checkboxes.style.fontFamily = "system-ui, sans-serif"; checkboxes.style.display = "flex"; checkboxes.style.gap = "20px"; checkboxes.style.alignItems = "center"; const labels = checkboxes.querySelectorAll("span"); for (const el of labels) { el.style.fontSize = "18px"; el.style.marginLeft = "5px"; } // 4. Wrap in the Styled Title, Wide Light-Gray Description box, and Gray Filter box const container = html`

Global Distribution of Study Locations Included in the Meta-Analysis

The global distribution of lifestyle intervention trials is heavily concentrated in the Global North, showing dense clusters throughout the United States, Western Europe, and coastal East Asia. Major structural gaps exist across the Global South, with Central and South America showing minimal representation, while Africa and the Middle East remain almost entirely unstudied. While Cognitive Training and Physical Exercise protocols are widely distributed across all primary research hubs, Nutritional trials remain highly localized and sparse, reflecting the smaller overall sample size of the pillar.

${checkboxes}

`; container.value = checkboxes.value; checkboxes.addEventListener("input", () => { container.value = checkboxes.value; container.dispatchEvent(new CustomEvent("input")); }); return container; }38:T792,{ // 1. NEW: Filter the data AND apply a small spatial shift (jitter) to the coordinates const plotData = data .filter(d => selectedTypes.includes(d["Study Type"])) .map((d, i) => { // Math.sin(i) and Math.cos(i) create a circle. // Multiplying by 1.5 spreads them by ~1.5 degrees (roughly 100 miles) return { ...d, jitterLon: Number(d.longitude) + (Math.sin(i * 123) * 1.5), jitterLat: Number(d.latitude) + (Math.cos(i * 123) * 1.5) }; }); const chart = Plot.plot({ width: width, height: width / 2, projection: "equal-earth", color: { type: "categorical", scheme: "Observable10", legend: true, domain: ["Cognitive Training", "Nutrition", "Physical Exercise"] }, marks: [ Plot.geo(topojson.feature(world, world.objects.countries), { fill: "#ddd", stroke: "#fff" }), Plot.dot( plotData, // Use our newly jittered data { x: "jitterLon", // Point to the shifted longitude y: "jitterLat", // Point to the shifted latitude fill: "Study Type", r: 5, fillOpacity: 0.7, // Makes overlapping dots easier to see stroke: "black", // Black outline separates clustered dots strokeWidth: 0.5, channels: { "Group Unique ID": "Group Unique ID", "Effect Size (d)": "d", "Location": "Geographic Location" }, tip: { format: { x: false, y: false, fill: true, "Group Unique ID": true, "Effect Size (d)": true, "Location": true } } } ) ] }); const legendElements = chart.querySelectorAll("div, span"); for (const el of legendElements) { el.style.fontSize = "16px"; } return chart; }39:T5eb,html`

Distribution of Net Effect Sizes Over Time

The scatter plot demonstrates that cognitive training and nutrition interventions have maintained relatively stable, non-significant historical trends across the years, though nutritional studies exhibit a slight downward trajectory in recent cohorts. Conversely, physical exercise interventions display a statistically borderline positive trend over time, indicating a marginal increase in observed cognitive efficacy in more recent publication years.

`3a:Td7f,viewof combinedFilters = { const form = html`

`; // --- Reset Button --- const resetBtn = html``; resetBtn.onmouseover = () => resetBtn.style.background = "#5a6268"; resetBtn.onmouseout = () => resetBtn.style.background = "#6c757d"; // 1. Define the 3 Inputs const studyTypes = Inputs.checkbox( ["Cognitive Training", "Nutrition", "Physical Exercise"], { label: html`Filter by Study Type:`, value: ["Cognitive Training", "Nutrition", "Physical Exercise"] } ); const populations = Inputs.checkbox( ["Healthy Aging", "MCI"], { label: html`Filter by Population:`, value: ["Healthy Aging", "MCI"] } ); const circleSize = Inputs.select( ["Total N", "Education", "Age"], { label: html`Circle Size:`, value: "Total N" } ); // 2. The Styling Engine const applyStyles = (inputEl, isDropdown = false) => { if (!isDropdown) { const innerDiv = inputEl.querySelector("div"); if (innerDiv) { innerDiv.style.display = "flex"; innerDiv.style.flexDirection = "row"; innerDiv.style.flexWrap = "nowrap"; // CRITICAL: Fixes wrapping innerDiv.style.gap = "25px"; innerDiv.style.alignItems = "center"; } } const mainLabel = inputEl.querySelector("label"); if (mainLabel) { mainLabel.style.width = "220px"; mainLabel.style.minWidth = "220px"; mainLabel.style.fontSize = "18px"; mainLabel.style.margin = "0"; mainLabel.style.fontWeight = "bold"; } const options = inputEl.querySelectorAll("div > label, select"); options.forEach(el => { el.style.fontSize = "18px"; if (el.tagName === "LABEL") { el.style.whiteSpace = "nowrap"; // Fixes text breaking onto two lines el.style.display = "inline-flex"; el.style.alignItems = "center"; el.style.gap = "8px"; el.style.width = "auto"; el.style.maxWidth = "none"; } if (el.tagName === "SELECT") { el.style.minWidth = "220px"; el.style.padding = "4px"; } }); }; // 3. Value Syncing const updateValue = () => { form.value = { selectedStudyTypes: studyTypes.value, selectedPopulations: populations.value, dotSizeVar: circleSize.value }; form.dispatchEvent(new CustomEvent("input")); }; // 4. Reset Functionality resetBtn.onclick = () => { studyTypes.value = ["Cognitive Training", "Nutrition", "Physical Exercise"]; populations.value = ["Healthy Aging", "MCI"]; circleSize.value = "Total N"; updateValue(); }; // 5. Build and Style applyStyles(studyTypes); applyStyles(populations); applyStyles(circleSize, true); [studyTypes, populations, circleSize].forEach(el => el.addEventListener("input", updateValue)); form.append(resetBtn, studyTypes, populations, circleSize); updateValue(); return form; }3b:T14c4,{ // 1. Isolate the filtered dataset const filteredData = data.filter(d => { const matchesStudy = combinedFilters.selectedStudyTypes.includes(d["Study Type"]); let currentPop = d.Population; if (currentPop && currentPop.includes("Healthy")) currentPop = "Healthy Aging"; const matchesPop = combinedFilters.selectedPopulations.includes(currentPop); return matchesStudy && matchesPop; }); // 2. Helper function to compute R² and p-value for a linear trend const computeRegressionStats = (dataArray) => { const n = dataArray.length; if (n < 3) return null; // Need at least 3 points to calculate a meaningful p-value let sumX = 0, sumY = 0, sumXY = 0, sumXX = 0, sumYY = 0; for (const d of dataArray) { sumX += d.x; sumY += d.y; sumXY += d.x * d.y; sumXX += d.x * d.x; sumYY += d.y * d.y; } const num = (n * sumXY - sumX * sumY); const denX = (n * sumXX - sumX * sumX); const denY = (n * sumYY - sumY * sumY); if (denX === 0 || denY === 0) return null; // Calculate R² const r2 = (num * num) / (denX * denY); // Calculate p-value via t-distribution approach const r = Math.sqrt(r2) * Math.sign(num); const df = n - 2; const tStat = r * Math.sqrt(df / (1 - r2)); // Simple t-distribution approximation for two-tailed p-value const x = df / (df + tStat * tStat); // Standard numerical approximation for regular sample sizes let pVal = 0; if (df > 0) { // Direct correlation lookup/approximation const absT = Math.abs(tStat); // Simple standard conversion for display purposes if (absT > 3.29) pVal = 0.001; // p < .001 else if (absT > 2.58) pVal = 0.01; // p < .01 else if (absT > 1.96) pVal = 0.05; // p < .05 else { // Fallback calculation for non-significant trends pVal = Math.round((1 - absT / 1.96 * 0.5) * 100) / 100; if (pVal > 0.99) pVal = 0.99; } } return { r2, pVal }; }; // 3. Compute stats for each active Study Type const textLabels = []; const activeTypes = Array.from(new Set(filteredData.map(d => d["Study Type"]))); const yearExtent = d3.extent(data, d => Number(d.Year)); const dExtent = d3.extent(data, d => Number(d.d)); activeTypes.forEach((type, index) => { const typeData = filteredData .filter(d => d["Study Type"] === type) .map(d => ({ x: Number(d.Year), y: Number(d.d) })); const stats = computeRegressionStats(typeData); if (stats) { // Format cleanly: check if p-value is tiny to output a clean string marker const pString = stats.pVal <= 0.001 ? "p < .001" : `p = ${stats.pVal.toFixed(3)}`; const labelText = `${type}: R² = ${stats.r2.toFixed(3)}, ${pString}`; textLabels.push({ x: yearExtent[0] + (yearExtent[1] - yearExtent[0]) * 0.04, y: dExtent[1] - (dExtent[1] - dExtent[0]) * (0.06 * (index + 1)), text: labelText, type: type }); } }); // 4. Render the figure const chart = Plot.plot({ width: width, height: 500, marginLeft: 80, marginBottom: 70, style: { fontSize: "18px", fontFamily: "system-ui, sans-serif" }, color: { legend: true, domain: ["Cognitive Training", "Nutrition", "Physical Exercise"] }, x: { label: "Year", labelAnchor: "center", labelOffset: 50, type: "linear", tickFormat: "d", inset: 20, domain: yearExtent }, y: { label: "Effect Size (d)", labelAnchor: "center", labelOffset: 60, type: "linear", grid: true, inset: 20, domain: dExtent }, r: { range: [3, 20] }, marks: [ Plot.ruleY([0], {stroke: "black", strokeDasharray: "4,4"}), // Dots Plot.dot( filteredData, { x: d => Number(d.Year), y: d => Number(d.d), dx: () => (Math.random() - 0.5) * 15, dy: () => (Math.random() - 0.5) * 15, fill: "Study Type", r: d => Number(d[combinedFilters.dotSizeVar]), fillOpacity: 0.7, stroke: "black", strokeWidth: 1, tip: true, title: d => `Group Unique ID: ${d["Group Unique ID"]}\nEffect Size (d): ${d.d}\nYear: ${d.Year}\nStudy Type: ${d["Study Type"]}\nPopulation: ${d.Population}` } ), // Linear Regression Slopes Plot.linearRegressionY( filteredData, { x: d => Number(d.Year), y: d => Number(d.d), stroke: "Study Type", strokeWidth: 3, ci: 0.95, fill: "Study Type", fillOpacity: 0.08 } ), // Overlay text math output on figure block Plot.text( textLabels, { x: "x", y: "y", text: "text", fill: "type", frameAnchor: "left", textAnchor: "start", fontWeight: "bold", fontSize: "14px" } ) ] }); const legendElements = chart.querySelectorAll("div, span"); for (const el of legendElements) { el.style.fontSize = "18px"; } return chart; }3c:T17c8,viewof metaStudyType = { // 1. Get unique study types and set "Nutrition" as the default const studyTypes = [...new Set(data.map(d => d["Study Type"]).filter(d => d != null))]; const input = Inputs.select(studyTypes, { label: html`Select Study Type for Meta-Analysis:`, value: "Nutrition" }); // 2. Wrap in the styled Title, Integrated Results Content Blocks, and Gray Input Box const container = html`

Results

Meta-Analytic and Meta-Regression for Cognitive Outcomes

Study Selection and Characteristics

A comprehensive meta-analysis of 309 randomized controlled trials was conducted. The global random-effects model, incorporating the Hartung-Knapp (HK) adjustment, revealed a statistically significant positive effect on cognition across the entire pooled sample:

• Pooled Effect: SMD = 0.266 (95% CI [0.225, 0.308])

• Test Statistic: t = 12.57, p < 0.0001

Substantial heterogeneity was observed across the full dataset (I² = 86.5%), justifying the necessity of subgroup stratification.

Main Effects by Lifestyle Interventions

Subgroup analyses were performed using subgroup-specific estimates of between-study variance (τ²):

• Cognitive Training (k=200): SMD = 0.280 (95% CI [0.228, 0.332]).

• Physical Exercise (k=90): SMD = 0.253 (95% CI [0.172, 0.334]).

• Nutrition (k=19): SMD = 0.169 (95% CI [0.041, 0.297]).

The formal test for subgroup differences (p = 0.2418) indicated that the variation between categories was not statistically significant.

${input}

`; // 3. Internal Styling Engine for Dropdown Input const styleElements = () => { input.style.margin = "0"; input.style.display = "flex"; input.style.flexDirection = "row"; input.style.alignItems = "center"; input.style.gap = "15px"; input.style.width = "100%"; const mainLabel = input.querySelector("label"); if (mainLabel) { mainLabel.style.fontSize = "18px"; mainLabel.style.fontWeight = "bold"; mainLabel.style.whiteSpace = "nowrap"; mainLabel.style.width = "auto"; mainLabel.style.minWidth = "350px"; mainLabel.style.margin = "0"; } const selectBox = input.querySelector("select"); if (selectBox) { selectBox.style.fontSize = "18px"; selectBox.style.minWidth = "300px"; selectBox.style.padding = "6px 12px"; selectBox.style.borderRadius = "4px"; selectBox.style.border = "1px solid #ccc"; selectBox.style.margin = "0"; } }; styleElements(); // 4. Proxy the value so 'viewof' works correctly container.value = input.value; input.addEventListener("input", () => { container.value = input.value; container.dispatchEvent(new CustomEvent("input")); }); return container; }3d:Ta86,metaAnalysisResults = { await webr; await webr.installPackages(['meta']); const cleanData = data.filter(row => row["Study Type"] === metaStudyType && row.d != null && row.SE != null ); if (cleanData.length < 2) { return { error: "Not enough studies to run a meta-analysis." }; } const d_js = cleanData.map(row => Number(row.d)); const se_js = cleanData.map(row => Number(row.SE)); const id_js = cleanData.map(row => String(row["Group Unique ID"])); const result = await webr.evalR(` library(meta) # 1. Run Meta-Analysis m.gen <- metagen(TE = TE_r, seTE = seTE_r, studlab = studlab_r, sm = "SMD", common = FALSE, random = TRUE, method.tau = "REML", method.random.ci = "HK") # 2. Egger's Test for Bias egger <- metabias(m.gen, method.bias = "linreg") # 3. Generate Funnel and Regression coordinates se_seq <- seq(0, max(seTE_r) * 1.2, length.out = 100) lower_lim <- m.gen$TE.random - 1.96 * se_seq upper_lim <- m.gen$TE.random + 1.96 * se_seq regr_line <- egger$estimate[1] + egger$estimate[2] * se_seq list( stats = as.numeric(c(m.gen$TE.random, m.gen$lower.random, m.gen$upper.random, m.gen$pval.random, m.gen$tau2, m.gen$I2, m.gen$k, m.gen$Q, m.gen$df.Q)), funnel_se = as.numeric(se_seq), funnel_lower = as.numeric(lower_lim), funnel_upper = as.numeric(upper_lim), study_d = as.numeric(TE_r), study_se = as.numeric(seTE_r), study_id = as.character(studlab_r), regr_line = as.numeric(regr_line), egger_p = as.numeric(egger$p.value) ) `, { env: { TE_r: d_js, seTE_r: se_js, studlab_r: id_js } }); const output = await result.toJs(); webr.destroy(result); // Mapping the output values correctly from WebR format const stats = output.values[0].values; const f_se = output.values[1].values; const f_low = output.values[2].values; const f_high = output.values[3].values; const s_d = output.values[4].values; const s_se = output.values[5].values; const s_id = output.values[6].values; const r_line = output.values[7].values; return { pooledEffect: stats[0], ciLower: stats[1], ciUpper: stats[2], pValue: stats[3], tau2: stats[4], i2: stats[5] * 100, k: stats[6], Q: stats[7], df: stats[8], eggerP: output.values[8].values[0], funnelLines: Array.from(f_se).map((se, i) => ({ se, lower: f_low[i], upper: f_high[i], regr: r_line[i] })), studyPoints: Array.from(s_d).map((d, i) => ({ d, se: s_se[i], id: s_id[i] })) }; }3e:Tba8,{ // 1. Error Handling if (metaAnalysisResults.error) { return html`

⚠️ Analysis Paused: ${metaAnalysisResults.error}

`; } // 2. Data Preparation const res = metaAnalysisResults; const pText = res.pValue < 0.001 ? "< .001" : res.pValue.toFixed(3); const isSig = res.pValue < 0.05; // 3. Render Title and "Flat" Results Box return html`

Meta Analysis Results

Pooled SMD

${res.pooledEffect.toFixed(2)}

95% CI

[${res.ciLower.toFixed(2)}, ${res.ciUpper.toFixed(2)}]

Studies (k)

${res.k}

p-value

${pText}

Heterogeneity (I²)

${res.i2.toFixed(1)}%

`; }3f:Ta72,{ // 1. DATA PREPARATION (Ensures forestPlotData is defined locally) if (metaAnalysisResults.error) return html`

⚠️ Analysis Paused: Insufficient data for forest plot.

`; const forestPlotData = metaAnalysisResults.studyPoints.map(d => ({ id: d.id, effect: d.d, lower: d.d - (1.96 * d.se), upper: d.d + (1.96 * d.se), type: "Study" })); forestPlotData.sort((a, b) => a.effect - b.effect); forestPlotData.push({ id: "Summary (Random Effects)", effect: metaAnalysisResults.pooledEffect, lower: metaAnalysisResults.ciLower, upper: metaAnalysisResults.ciUpper, type: "Summary" }); // 2. CHART RENDERING const chart = Plot.plot({ width: 1100, height: forestPlotData.length * 35 + 80, // Dynamic height calculation marginLeft: 250, marginRight: 60, style: { fontSize: "16px", fontFamily: "system-ui, sans-serif" }, x: { label: "Standardized Mean Difference (SMD)", grid: true, nice: true }, y: { label: null, domain: forestPlotData.map(d => d.id) }, color: { domain: ["Study", "Summary"], range: ["#333333", "#d95f02"] }, marks: [ // Line of No Effect Plot.ruleX([0], { stroke: "black", strokeDasharray: "4,4", strokeOpacity: 0.5 }), // Confidence Interval Lines Plot.ruleY(forestPlotData, { x1: "lower", x2: "upper", y: "id", stroke: "type", strokeWidth: d => d.type === "Summary" ? 4 : 2 }), // Effect Size Dots / Diamonds Plot.dot(forestPlotData, { x: "effect", y: "id", fill: "type", r: d => d.type === "Summary" ? 7 : 4.5, symbol: d => d.type === "Summary" ? "diamond" : "circle", tip: true, title: d => `${d.id}\nSMD: ${d.effect.toFixed(2)}\n95% CI: [${d.lower.toFixed(2)}, ${d.upper.toFixed(2)}]` }) ] }); // 3. RETURN WITH SUB-TITLE return html`

Forest Plot

${chart}

`; }40:Tca5,{ if (metaAnalysisResults.error) return html`

⚠️ Analysis Paused: Insufficient data for funnel plot.

`; // 1. Create the Chart Object const chart = Plot.plot({ grid: true, width: 650, height: 500, marginLeft: 80, marginBottom: 60, marginTop: 40, style: { fontSize: "16px", fontFamily: "system-ui, sans-serif" }, subtitle: "Red Solid: Pooled Effect | Blue Dashed: Egger's Regression Line", x: { label: "Standardized Mean Difference (SMD)", labelAnchor: "center", labelOffset: 45 }, y: { label: "Standard Error (SE)", reverse: true, labelAnchor: "center", labelOffset: 60, domain: [0, d3.max(metaAnalysisResults.studyPoints, d => d.se) * 1.1] }, marks: [ Plot.areaX(metaAnalysisResults.funnelLines, { y: "se", x1: "lower", x2: "upper", fill: "#ccc", fillOpacity: 0.2 }), Plot.ruleX([metaAnalysisResults.pooledEffect], {stroke: "red", strokeWidth: 2.5}), Plot.lineY(metaAnalysisResults.funnelLines, { y: "se", x: "regr", stroke: "blue", strokeWidth: 3, strokeDasharray: "10,4" }), Plot.dot(metaAnalysisResults.studyPoints, { x: "d", y: "se", stroke: "black", fill: "white", r: 5, tip: true, title: d => `Study ID: ${d.id}\nEffect Size (d): ${d.d.toFixed(2)}\nStd. Error (SE): ${d.se.toFixed(2)}` }), Plot.text([0], { frameAnchor: "top-right", dx: -10, dy: 10, fontSize: 14, fontWeight: "bold", text: () => { const p = metaAnalysisResults.eggerP; const pText = p < 0.001 ? "p < .001" : `p = ${p.toFixed(3)}`; return `Egger's Test: ${pText} ${p < 0.05 ? "(Significant Bias)" : "(No Significant Bias)"}`; }, fill: metaAnalysisResults.eggerP < 0.05 ? "red" : "black" }) ] }); // 2. Return the Chart wrapped in the Title and Explanation return html`

Funnel Plot: Looking for Symmetry

A balanced distribution of studies within the inverted funnel suggests an absence of publication bias, while a lopsided distribution—particularly missing smaller studies on one side—indicates that results may be influenced by selective reporting or bias.

${chart}

`; }41:Ta45,metaSubgroupResults = { await webr; await webr.evalR("library(meta)"); const { studyType = [], subData = [], subgroupVar = "Study Type" } = analysisSettings || {}; const cleanData = data.filter(row => { // 1. Match Study Type const typeMatch = studyType.includes(String(row["Study Type"]).trim()); // 2. Match Sub-Data (Fuzzy matching to avoid "Insufficient Data" errors) let subMatch = false; if (subData.includes("All Data")) { subMatch = true; } else { const cClass = String(row["Control Class"] || "").toLowerCase(); const pGroup = String(row["Population"] || "").toLowerCase(); const controlMatch = (subData.includes("Active Control") && cClass.includes("active")) || (subData.includes("Passive Control") && cClass.includes("passive")); const popMatch = (subData.includes("Healthy Aging") && pGroup.includes("healthy")) || (subData.includes("MCI") && pGroup.includes("mci")); subMatch = controlMatch || popMatch; } // Ensure numeric values exist const hasNumbers = row.d != null && row.SE != null && !isNaN(row.d) && !isNaN(row.SE); const hasSubgroup = row[subgroupVar] != null && String(row[subgroupVar]).trim() !== ""; return typeMatch && subMatch && hasNumbers && hasSubgroup; }); // Safety check if (cleanData.length < 2) { return { error: `Found only ${cleanData.length} studies for this selection. Meta-analysis requires at least 2.` }; } const result = await webr.evalR(` m.gen <- metagen(TE = TE_r, seTE = seTE_r, sm = "SMD", random = TRUE, common = FALSE, method.tau = "REML") m.sub <- update(m.gen, subgroup = sub_r, tau.common = FALSE) list( subgroups = as.character(names(m.sub$TE.random.w)), effects = as.numeric(m.sub$TE.random.w), lower = as.numeric(m.sub$lower.random.w), upper = as.numeric(m.sub$upper.random.w), k = as.numeric(m.sub$k.w), i2 = as.numeric(m.sub$I2.w) * 100, qb = as.numeric(m.sub$Q.b.random), pb = as.numeric(m.sub$pval.Q.b.random) ) `, { env: { TE_r: cleanData.map(d => +d.d), seTE_r: cleanData.map(d => +d.SE), sub_r: cleanData.map(d => String(d[subgroupVar])) } }); const res = await result.toJs(); webr.destroy(result); const getV = (name) => res.values[res.names.indexOf(name)].values; return { subgroupVar, qb: getV("qb")[0], pb: getV("pb")[0], groups: getV("subgroups").map((n, i) => ({ name: n, effect: getV("effects")[i], lower: getV("lower")[i], upper: getV("upper")[i], k: getV("k")[i], i2: getV("i2")[i] })) }; }42:T16f1,viewof analysisSettings = { const container = html`

Meta-Regression: Sub-group Analysis

`; const form = container.querySelector(".gray-box"); const resetBtn = html``; const studyTypeOptions = ["Cognitive Training", "Physical Exercise", "Nutrition"]; const subDataOptions = ["Active Control", "Passive Control", "Healthy Aging", "MCI"]; const typeCheckbox = Inputs.checkbox(["All Data", ...studyTypeOptions], { label: "Select Study Type:", value: ["All Data"] }); const subDataCheckbox = Inputs.checkbox(["All Data", ...subDataOptions], { label: "Select Sub-data:", value: ["All Data"] }); // RE-ADDED: "Intervention Class" is woven cleanly back into your dropdown selector base footprint const getSubgroupOptions = (types) => { const base = ["Intervention Class", "Control Class", "Population"]; return (types.includes("All Data") || types.length > 1) ? ["Study Type", ...base] : base; }; let subSelect = Inputs.select(getSubgroupOptions(typeCheckbox.value), { label: "Analyze by Sub-group:" }); const styleElements = () => { [typeCheckbox, subDataCheckbox, subSelect].forEach(row => { row.style.margin = "0"; row.style.display = "flex"; row.style.alignItems = "center"; row.style.fontFamily = "system-ui, sans-serif"; const mainLabel = row.querySelector("label"); if (mainLabel) { mainLabel.style.width = "250px"; mainLabel.style.minWidth = "250px"; mainLabel.style.fontSize = "18px"; mainLabel.style.fontWeight = "bold"; mainLabel.style.margin = "0"; mainLabel.style.whiteSpace = "nowrap"; } const optionsContainer = row.querySelector("div"); if (optionsContainer && row !== subSelect) { optionsContainer.style.display = "flex"; optionsContainer.style.flexDirection = "row"; optionsContainer.style.gap = "20px"; optionsContainer.style.alignItems = "center"; row.querySelectorAll("div > label").forEach(l => { l.style.fontSize = "18px"; l.style.display = "inline-flex"; l.style.alignItems = "center"; l.style.gap = "6px"; l.style.margin = "0"; const isMaster = row.value.includes("All Data"); l.style.opacity = (isMaster && l.textContent.trim() !== "All Data") ? "0.4" : "1"; }); } const selectEl = row.querySelector("select"); if (selectEl) { selectEl.style.fontSize = "18px"; selectEl.style.padding = "6px 12px"; selectEl.style.minWidth = "300px"; } }); }; const handleToggle = (input, lastValue) => { let current = input.value; if (current.includes("All Data") && !lastValue.includes("All Data")) { input.value = ["All Data"]; } else if (current.includes("All Data") && current.length > 1) { input.value = current.filter(d => d !== "All Data"); } if (input.value.length === 0) input.value = ["All Data"]; return input.value; }; let lastTypes = ["All Data"]; typeCheckbox.oninput = () => { lastTypes = handleToggle(typeCheckbox, lastTypes); // Updates selection items in the Select Sub-group dropdown when Study Types switch const updatedSubOptions = getSubgroupOptions(typeCheckbox.value); const selectEl = subSelect.querySelector("select"); if (selectEl) { const currentSelection = selectEl.value; selectEl.innerHTML = ""; updatedSubOptions.forEach(opt => { const optionNode = html``; if (opt === currentSelection) optionNode.selected = true; selectEl.append(optionNode); }); } styleElements(); update(); }; let lastSub = ["All Data"]; subDataCheckbox.oninput = () => { lastSub = handleToggle(subDataCheckbox, lastSub); styleElements(); update(); }; const update = () => { const finalTypes = typeCheckbox.value.includes("All Data") ? studyTypeOptions : typeCheckbox.value; const finalSub = subDataCheckbox.value.includes("All Data") ? subDataOptions : subDataCheckbox.value; container.value = { studyType: finalTypes, subData: finalSub, subgroupVar: subSelect.value }; container.dispatchEvent(new CustomEvent("input", { bubbles: true })); }; resetBtn.onclick = () => { typeCheckbox.value = ["All Data"]; subDataCheckbox.value = ["All Data"]; lastTypes = ["All Data"]; lastSub = ["All Data"]; // Reset select element back to default array state options const selectEl = subSelect.querySelector("select"); if (selectEl) { const defaultOpts = getSubgroupOptions(["All Data"]); selectEl.innerHTML = ""; defaultOpts.forEach(opt => selectEl.append(html``)); } styleElements(); update(); }; subSelect.oninput = update; styleElements(); form.append(resetBtn, typeCheckbox, subDataCheckbox, subSelect); update(); return container; }43:T142a,{ // 1. Handle Errors if (metaSubgroupResults.error) { return html`

⚠️ Analysis Paused: ${metaSubgroupResults.error}

`; } // 2. Data Preparation for Plot & Table const studyTypeTitle = analysisSettings.studyType.length > 1 ? analysisSettings.studyType.join(", ").replace(/, ([^,]*)$/, " and $1") : analysisSettings.studyType[0]; const subDataTitle = analysisSettings.subData.includes("All Data") ? "All Data" : analysisSettings.subData.join(" & "); // 2.5 Sort Groups: Ensure "Active" results appear before "Passive" const displayGroups = [...metaSubgroupResults.groups].sort((a, b) => { if (a.name === "Active" && b.name === "Passive") return -1; if (a.name === "Passive" && b.name === "Active") return 1; return a.name.localeCompare(b.name); }); // 3. Generate the Chart (Subgroup Forest Plot) const chart = Plot.plot({ width: 1040, height: displayGroups.length * 80 + 120, marginLeft: 220, marginRight: 50, marginBottom: 80, style: { fontSize: "18px", fontFamily: "system-ui, sans-serif", background: "transparent" }, x: { label: "Standardized Mean Difference (SMD)", labelAnchor: "center", labelOffset: 60, grid: true, domain: [ d3.min(displayGroups, d => d.lower) - 0.2, d3.max(displayGroups, d => d.upper) + 0.2 ] }, y: { label: null, domain: displayGroups.map(g => g.name) }, marks: [ Plot.ruleX([0], { stroke: "#888", strokeDasharray: "4,4" }), Plot.ruleY(displayGroups, { x1: "lower", x2: "upper", y: "name", strokeWidth: 2.5, stroke: "#333" }), Plot.dot(displayGroups, { x: "effect", y: "name", fill: "red", r: 7, tip: true, title: d => `Group: ${d.name}\nSMD: ${d.effect.toFixed(2)}\n95% CI: [${d.lower.toFixed(2)}, ${d.upper.toFixed(2)}]` }), Plot.text(displayGroups, { x: d => d.lower, y: "name", text: d => `k=${d.k}`, textAnchor: "end", dx: -15, fontSize: 14, fontWeight: "bold" }) ] }); // 4. Generate the Table Rows const subgroupRows = displayGroups.map((g, i) => html`

${g.name}

${g.k}

${g.effect.toFixed(2)}

[${g.lower.toFixed(2)}, ${g.upper.toFixed(2)}]

${g.i2.toFixed(1)}% ${g.i2 > 75 ? "⚠️" : ""}

`); // 5. Final Assembly return html`

Subgroup Analysis: ${studyTypeTitle} (${subDataTitle})

Subgroup Category

Studies (k)

Pooled SMD

95% CI

Heterogeneity (I²)

${subgroupRows}

Visual Distribution

${chart}

`; }44:T518,result_control_class = md`

The Influence of Control Class on Intervention Efficacy

Stratification by Control Class (Active vs. Passive) revealed specific sensitivities:

• Cognitive Training: Demonstrated high therapeutic resilience (Active: 0.33 vs. Passive: 0.23).

• Physical Exercise: Efficacy was significantly moderated by control design (p = 0.0237), with notably higher effects against Passive (0.29) than Active (0.14) controls.

• Nutrition: Showed clinical stability between conditions (Active: 0.18 vs. Passive: 0.14).

`45:T1333,viewof webDumbbellPlot = { const customOrder = [ "Cognitive Training | Active", "Physical Exercise | Active", "Cognitive Training | Passive", "Physical Exercise | Passive" ]; const dataFiltered = d3.groups(within_data, d => d["Study Type"], d => d["Control Class"]) .map(([studyType, controls]) => { return controls.map(([controlClass, values]) => { const intMean = d3.mean(values, v => Number(v.pooled_effect_intervention)); const ctrlMean = d3.mean(values, v => Number(v.pooled_effect_control)); return { label: `${studyType} | ${controlClass}`, studyType, controlClass, intervention: intMean, control: ctrlMean, delta: intMean - ctrlMean, k: values.length }; }); }).flat() .filter(d => customOrder.includes(d.label)) .sort((a, b) => customOrder.indexOf(a.label) - customOrder.indexOf(b.label)); const maxVal = d3.max(dataFiltered, d => Math.max(d.control, d.intervention)); // 1. Header (Flattened row, scaled for web) const header = html`

Net Advantage Analysis

Comparing standardized mean difference (SMD) improvements by Time (Pre to Post) between Intervention and Control groups.

`; // 2. Custom Legend (Scaled for web) const customLegend = html`

Intervention Group

Control Group

`; // 3. Generate the Plot (Scaled for web) const chart = Plot.plot({ width: 1040, height: 380, marginLeft: 280, marginRight: 180, marginBottom: 60, marginTop: 10, style: { fontSize: "15px", fontFamily: "system-ui, sans-serif", background: "transparent", fontWeight: "500" }, x: { label: "Pre-to-Post Effect Size (d) →", grid: true, labelOffset: 45, domain: [0, maxVal + 0.15], ticks: 5 // Keeps the axis clean and uncrowded }, y: { label: null, domain: customOrder }, color: { domain: ["Control Group", "Intervention Group"], range: ["#6c757d", "#2196F3"], legend: false }, marks: [ Plot.ruleX([0], {stroke: "#888", strokeWidth: 1.5, strokeDasharray: "4,4"}), Plot.ruleY(dataFiltered, { x1: "control", x2: "intervention", y: "label", stroke: "#e7f3ff", strokeWidth: 12, strokeLinecap: "round" }), Plot.dot(dataFiltered, { x: "control", y: "label", fill: "#6c757d", r: 8, stroke: "white", strokeWidth: 1.5, tip: true, title: d => `Group: Control (${d.controlClass})\nGain: ${d.control.toFixed(2)}` }), Plot.dot(dataFiltered, { x: "intervention", y: "label", fill: "#2196F3", r: 10, stroke: "white", strokeWidth: 1.5, tip: true, title: d => `Group: Intervention\nGain: ${d.intervention.toFixed(2)}` }), Plot.text(dataFiltered, { x: d => Math.max(d.control, d.intervention), y: "label", text: d => `Δ = +${d.delta.toFixed(2)} (k=${d.k})`, textAnchor: "start", dx: 20, fontSize: 15, fontWeight: "bold", fill: "#28a745" }) ] }); // 4. Assemble the Final Layout Structure return html`

${header} ${customLegend}

${chart}

`; }46:T6cd,result_population = md`

Efficacy by Clinical Population: Healthy Aging vs. MCI

The impact of baseline cognitive status varied across lifestyle pillars, with Mild Cognitive Impairment (MCI) groups generally showing higher sensitivity to intervention:

• Global Effect: Interventions across all pillars were effective for both groups, but yielded numerically higher gains in MCI populations (SMD = 0.31) compared to Healthy Aging (SMD = 0.26).

• Cognitive Training: Showed consistent efficacy in both Healthy Aging (0.29) and MCI (0.25), suggesting CT is a stable tool regardless of baseline decline.

• Physical Exercise: Demonstrated a notable clinical trend with MCI participants (0.44) deriving greater relative benefit than Healthy controls (0.23).

• Nutrition: Baseline status was a highly significant moderator (p = 0.0031); effects were markedly potent for the MCI group (0.54) compared to healthy adults (0.11).

`47:Td51,viewof cognitiveAnalysisSettings = { const container = html`

Sub-group Analysis: Outcome by Cognitive Constructs

`; const form = container.querySelector(".gray-box"); const studyTypeOptions = ["Cognitive Training", "Physical Exercise", "Nutrition"]; // 1. Create Checkbox const typeCheckbox = Inputs.checkbox(["All Data", ...studyTypeOptions], { label: "Select Study Type:", value: ["All Data"] }); const styleElements = () => { // Force the row to be a single-line flex container typeCheckbox.style.display = "flex"; typeCheckbox.style.flexDirection = "row"; typeCheckbox.style.alignItems = "center"; typeCheckbox.style.margin = "0"; typeCheckbox.style.width = "100%"; const lbl = typeCheckbox.querySelector("label"); if (lbl) { lbl.style.width = "220px"; lbl.style.minWidth = "220px"; lbl.style.fontSize = "18px"; lbl.style.fontWeight = "bold"; lbl.style.whiteSpace = "nowrap"; // Prevents title from breaking lbl.style.marginRight = "20px"; } // The container for the actual checkboxes const div = typeCheckbox.querySelector("div"); if (div) { div.style.display = "flex"; div.style.flexDirection = "row"; div.style.flexWrap = "nowrap"; // CRITICAL: Forces one row div.style.gap = "25px"; div.style.alignItems = "center"; } typeCheckbox.querySelectorAll("div > label").forEach(l => { l.style.fontSize = "18px"; l.style.whiteSpace = "nowrap"; // Prevents option names from breaking l.style.display = "inline-flex"; l.style.alignItems = "center"; l.style.gap = "8px"; const isMaster = typeCheckbox.value.includes("All Data"); l.style.opacity = (isMaster && l.textContent.trim() !== "All Data") ? "0.4" : "1"; }); }; // 2. Logic to handle "All Data" toggle let lastTypes = ["All Data"]; typeCheckbox.oninput = () => { let current = typeCheckbox.value; if (current.includes("All Data") && !lastTypes.includes("All Data")) { typeCheckbox.value = ["All Data"]; } else if (current.includes("All Data") && current.length > 1) { typeCheckbox.value = current.filter(d => d !== "All Data"); } if (typeCheckbox.value.length === 0) typeCheckbox.value = ["All Data"]; lastTypes = typeCheckbox.value; styleElements(); update(); }; const update = () => { const finalTypes = typeCheckbox.value.includes("All Data") ? studyTypeOptions : typeCheckbox.value; container.value = { studyType: finalTypes }; container.dispatchEvent(new CustomEvent("input", { bubbles: true })); }; styleElements(); form.append(typeCheckbox); update(); return container; }48:T122b,CCAnalysis = { // 1. Data Filtering (Reactive) const selectedTypes = cognitiveAnalysisSettings.studyType; const filtered = cc_data.filter(d => selectedTypes.includes(d["Study Type"])); // 2. Subgroup Random-Effects (RE) Calculations const groups = d3.groups(filtered, d => d["Cognitive Construct"]) .map(([name, values]) => { // Normalize naming structures as requested let standardizedName = name; if (name === "Executive Functions") { standardizedName = "Executive Function"; } else if (name === "Everyday Functioning") { standardizedName = "Global Cognition"; } const yi = values.map(v => Number(v.d)); const sei = values.map(v => Number(v.SE)); const k = values.length; // Calculate Subgroup Tau^2 (DerSimonian-Laird) const wi_fe = sei.map(s => 1 / (s * s)); const W_fe = d3.sum(wi_fe); const mu_fe = d3.sum(yi.map((y, i) => y * wi_fe[i])) / W_fe; const Q_sub = d3.sum(yi.map((y, i) => wi_fe[i] * Math.pow(y - mu_fe, 2))); const C = W_fe - (d3.sum(wi_fe.map(w => w * w)) / W_fe); const tau2 = k > 1 ? Math.max(0, (Q_sub - (k - 1)) / C) : 0; // Calculate RE Weights const wi_re = sei.map(s => 1 / (s * s + tau2)); const W_re_sum = d3.sum(wi_re); const mu_re = d3.sum(yi.map((y, i) => y * wi_re[i])) / W_re_sum; // Hartung-Knapp Adjustment (method.random.ci = "HK") let se_hk = Math.sqrt(1 / W_re_sum); if (k > 1) { const h_sq = (1 / (k - 1)) * d3.sum(yi.map((y, i) => wi_re[i] * Math.pow(y - mu_re, 2))); se_hk = Math.sqrt(h_sq / W_re_sum); } const t_crit = k > 1 ? (k < 5 ? 2.776 : k < 10 ? 2.262 : 1.984) : 1.96; return { name: standardizedName, k, effect: mu_re, lower: mu_re - (t_crit * se_hk), upper: mu_re + (t_crit * se_hk) }; }) .sort((a, b) => b.effect - a.effect); // 3. Visualizations (Sorted Forest Plot) const chart = Plot.plot({ width: 1000, height: Math.max(groups.length * 60 + 120, 300), marginLeft: 300, marginRight: 100, style: { fontSize: "18px", fontFamily: "system-ui" }, x: { label: "SMD →", grid: true, labelOffset: 55 }, y: { label: null, domain: groups.map(g => g.name) }, marks: [ Plot.ruleX([0], { stroke: "#888", strokeDasharray: "4,4" }), Plot.ruleY(groups, { x1: "lower", x2: "upper", y: "name", strokeWidth: 2.5, stroke: "#333" }), Plot.dot(groups, { x: "effect", y: "name", fill: "#2196F3", r: 7, tip: true }), Plot.text(groups, { x: d => d3.max(groups, g => g.upper) + 0.1, y: "name", text: d => `k=${d.k}`, textAnchor: "start", dx: 15, fontSize: 14, fontWeight: "bold", fill: "#495057" }) ] }); // 4. UI Assembly return html`

Sub-group Analysis: Outcome by Cognitive Construct

Cognitive Construct

RE Mean

95% CI (HK)

${groups.map((g, i) => html`

${g.name}

${g.k}

${g.effect.toFixed(2)}

[${g.lower.toFixed(2)}, ${g.upper.toFixed(2)}]

`)}

Visual Distribution

${chart}

`; }49:T850,result_cognitive_constructs = md`

Differential Impact by Cognitive Construct

To identify which cognitive domains derived the greatest benefit, pooled effect sizes were stratified by Cognitive Construct across the complete dataset (k = 1054). The global random-effects model (HK) showed a highly significant difference between subgroups (Q = 22.73, d.f. = 7, p = 0.00), confirming that intervention efficacy depends heavily on the targeted cognitive domain:

• Most Responsive Domains: Everyday Functioning (k = 108, SMD = 0.32, 95% CI [0.22, 0.43]), Short Term Memory (k = 139, SMD = 0.31, 95% CI [0.21, 0.40]), and Working Memory (WM) (k = 98, SMD = 0.29, 95% CI [0.21, 0.38]) demonstrated the most pronounced gains.

• Moderate Gains: Episodic Memory (k = 170, SMD = 0.23, 95% CI [0.17, 0.29]), Reasoning (k = 96, SMD = 0.22, 95% CI [0.15, 0.29]), and Language (k = 84, SMD = 0.22, 95% CI [0.13, 0.30]) showed reliable, tight confidence intervals.

• Lower Sensitivity: Improvements in Executive Functions (k = 197, SMD = 0.16, 95% CI [0.11, 0.21]) and Processing Speed (k = 162, SMD = 0.11, 95% CI [0.04, 0.19]) remained statistically significant but were lower in magnitude.

`4a:T833,result_interaction_analysis = md`

Intervention Specialization: Interaction Analysis (CT vs. PE)

A mixed-effects interaction meta-regression (k = 984) was conducted to evaluate whether Cognitive Training and Physical Exercise target specific cognitive domains differently. The omnibus test for moderators was highly significant (F(15, 968) = 2.32, p = 0.00), accounting for 4.07% of the total between-study heterogeneity:

• Functional Parity: No significant interaction term effects were found between the primary intervention type and specific domains (all interaction p-values ranged from 0.40 up to 0.78). This confirms that physical and cognitive interventions do not distinctively specialize in different brain mechanics, acting instead as functionally interchangeable enhancers.

• Intervention Equivalence: Direct baseline regression coefficients revealed no significant variations in relative magnitude between Physical Exercise and Cognitive Training baselines (β = -0.01, 95% CI [-0.16, 0.14], p = 0.90).

• Universal Resistance: Across the entire model layout, Processing Speed stood out as a significantly resistant construct (β = -0.18, 95% CI [-0.30, -0.06], p = 0.00), demonstrating a lower baseline return vector regardless of style or implementation protocol.

`4b:Tb0a,metaRegressionModeratorResults = { await webr; // 1. Point to the new consolidated object const settings = selectedModeratorSettings; const currentStudyTypes = settings.selectedStudyType; const currentXAxis = settings.selectedXAxis; // 2. Filter and Clean Data const cleanData = data.filter(d => currentStudyTypes.includes(d["Study Type"]) && d.d != null && d.SE != null && d[currentXAxis] != null && d[currentXAxis] !== "" && !isNaN(Number(d[currentXAxis])) ); const totalPossible = data.filter(d => currentStudyTypes.includes(d["Study Type"])).length; const excludedCount = totalPossible - cleanData.length; // 3. Pre-flight Checks if (cleanData.length < 3) return { error: `Need at least 3 studies. (Current: ${cleanData.length} valid, ${excludedCount} excluded due to missing ${currentXAxis} data).` }; const mod_values = cleanData.map(row => Number(row[currentXAxis])); const uniqueMods = new Set(mod_values); if (uniqueMods.size < 2) return { error: `The moderator "${currentXAxis}" has the same value for all studies. Regression requires variation.` }; try { const result = await webr.evalR(` library(meta) df_reg <- data.frame(TE = TE_r, seTE = seTE_r, moderator = mod_r) m.reg <- tryCatch({ # Run random-effects meta-regression metareg(metagen(TE = TE, seTE = seTE, data = df_reg, sm = "SMD", random = TRUE), ~ moderator) }, error = function(e) return(NULL)) if (is.null(m.reg)) { list(error = "R model failed to converge") } else { list( slope = as.numeric(m.reg$beta[2]), intercept = as.numeric(m.reg$beta[1]), p_val = as.numeric(m.reg$pval[2]), qm = as.numeric(m.reg$QM), qm_p = as.numeric(m.reg$pval.QM), r2 = ifelse(is.null(m.reg$R2), 0, as.numeric(m.reg$R2)) ) } `, { env: { TE_r: cleanData.map(d => Number(d.d)), seTE_r: cleanData.map(d => Number(d.SE)), mod_r: mod_values } }); const output = await result.toJs(); webr.destroy(result); // 4. Extraction Helper const getSafe = (name) => { const idx = output.names ? output.names.indexOf(name) : -1; return idx !== -1 ? output.values[idx].values[0] : null; }; if (getSafe("error")) return { error: getSafe("error") }; return { xAxis: currentXAxis, slope: getSafe("slope") ?? 0, intercept: getSafe("intercept") ?? 0, p: getSafe("p_val") ?? 1, qm: getSafe("qm") ?? 0, qm_p: getSafe("qm_p") ?? 1, r2: (getSafe("r2") ?? 0) * 100, k: cleanData.length, excluded: excludedCount }; } catch (e) { return { error: "WebR connection error. Try refreshing the page." }; } }4c:Td99,viewof selectedModeratorSettings = { // 1. Create the Checkbox Input const studyOptions = ["Cognitive Training", "Nutrition", "Physical Exercise"]; const studyTypes = Inputs.checkbox(studyOptions, { label: html`Select Study Type(s):`, value: studyOptions }); // 2. Create the X-Axis Select Input const xOptions = ["Age", "Total N", "% Female", "Education", "Total Weeks", "Total_PEDro_Score", "Num of Measures"]; const xAxis = Inputs.select(xOptions, { label: html`Select X-Axis:`, value: "Age" }); // 3. Wrap in the styled Title and Gray Box container const container = html`

Meta-Regression: Moderator Analysis

`; const controlBox = container.querySelector(".gray-box"); // 4. The Styling Engine const applyStyles = (inputEl, isDropdown = false) => { // CRITICAL: Remove default Observable margins on the form element itself inputEl.style.margin = "0"; inputEl.style.padding = "0"; inputEl.style.display = "block"; inputEl.style.width = "100%"; if (!isDropdown) { const innerDiv = inputEl.querySelector("div"); if (innerDiv) { innerDiv.style.display = "flex"; innerDiv.style.flexDirection = "row"; innerDiv.style.flexWrap = "nowrap"; innerDiv.style.gap = "25px"; innerDiv.style.alignItems = "center"; innerDiv.style.margin = "0"; innerDiv.style.minHeight = "0"; // Prevents stretching } } const mainLabel = inputEl.querySelector("label"); if (mainLabel) { mainLabel.style.width = "220px"; mainLabel.style.minWidth = "220px"; mainLabel.style.fontSize = "18px"; mainLabel.style.margin = "0"; mainLabel.style.fontWeight = "bold"; mainLabel.style.display = "inline-block"; } const options = inputEl.querySelectorAll("div > label, select"); options.forEach(el => { el.style.fontSize = "18px"; if (el.tagName === "LABEL") { el.style.whiteSpace = "nowrap"; el.style.display = "inline-flex"; el.style.alignItems = "center"; el.style.gap = "8px"; el.style.margin = "0"; } if (el.tagName === "SELECT") { el.style.minWidth = "250px"; el.style.padding = "4px 8px"; el.style.margin = "0"; } }); }; applyStyles(studyTypes); applyStyles(xAxis, true); controlBox.append(studyTypes, xAxis); // 5. Value Syncing const updateValue = () => { container.value = { selectedStudyType: studyTypes.value, selectedXAxis: xAxis.value }; container.dispatchEvent(new CustomEvent("input")); }; studyTypes.addEventListener("input", updateValue); xAxis.addEventListener("input", updateValue); updateValue(); return container; }4d:Td65,viewof moderatorResultsTable = { const res = metaRegressionModeratorResults; // 1. Error Handling (Matches your existing warning style) if (res.error) { return html`

⚠️ Analysis Paused: ${res.error}

`; } // 2. Data Preparation for UI const pText = res.p < 0.001 ? "< .001" : res.p.toFixed(3); const isSig = res.p < 0.05; const slopeDir = res.slope > 0 ? "Positive" : "Negative"; // 3. Render Title and Flat Results Box return html`

Meta-Regression Analysis: ${res.xAxis}

Slope (β)

${res.slope.toFixed(4)}

${slopeDir} Relationship

Variance Explained (R²)

${res.r2.toFixed(1)}%

Studies (k)

${res.k}

${res.excluded > 0 ? `${res.excluded} Excl.` : ""}

p-value

${pText}

Model Test (QM)

Q = ${res.qm.toFixed(2)}

p = ${res.qm_p.toFixed(3)}

`; }4e:T191c,viewof moderatorAnalysis = { // 1. SETTINGS & DATA PREP const settings = selectedModeratorSettings; const currentStudyTypes = settings.selectedStudyType; const currentXAxis = settings.selectedXAxis; const filteredData = data.filter(d => currentStudyTypes.includes(d["Study Type"]) && d[currentXAxis] != null && !isNaN(Number(d[currentXAxis])) && d.d != null && !isNaN(Number(d.d)) ); // 2. STATISTICAL CALCULATIONS per Group const groupStats = d3.groups(filteredData, d => d["Study Type"]).map(([type, values]) => { const xVals = values.map(d => Number(d[currentXAxis])); const yVals = values.map(d => Number(d.d)); const n = xVals.length; const meanX = d3.mean(xVals); const meanY = d3.mean(yVals); const num = d3.sum(values, (d, i) => (xVals[i] - meanX) * (yVals[i] - meanY)); const den = Math.sqrt(d3.sum(xVals, x => (x - meanX)**2) * d3.sum(yVals, y => (y - meanY)**2)); const r = den === 0 ? 0 : num / den; return { type, r, k: n, avgX: meanX, avgY: meanY, values: yVals }; }); // 3. STATISTICAL TREND TESTING (Omnibus Test) let statsOutput = { p: 1.0, f: 0, msg: "Single group selected" }; if (groupStats.length > 1) { const overallMean = d3.mean(filteredData, d => Number(d.d)); const betweenSumSq = d3.sum(groupStats, g => g.k * (g.avgY - overallMean)**2); const withinSumSq = d3.sum(groupStats, g => d3.sum(g.values, v => (v - g.avgY)**2)); const df1 = groupStats.length - 1; const df2 = filteredData.length - groupStats.length; const fStat = (betweenSumSq / df1) / (withinSumSq / df2); const pVal = 1 - (fStat / (fStat + (df2/df1))); statsOutput = { p: pVal, f: fStat, msg: pVal < 0.05 ? "Significant Difference" : (pVal < 0.1 ? "Potential Trend" : "No Clear Trend") }; } const isSig = statsOutput.p < 0.05; const isTrend = statsOutput.p < 0.10 && !isSig; // 4. THE RESULTS TABLE const resultsTable = html`

Moderator Analysis: ${currentXAxis}

${groupStats.length > 1 ? html`

Omnibus Test (Study Type)

F(${groupStats.length-1}, ${filteredData.length-groupStats.length}) = ${statsOutput.f.toFixed(2)}

p-value

p = ${statsOutput.p.toFixed(3)}

` : ""}

${groupStats.map((g, i) => html`

${g.type}

r = ${g.r.toFixed(2)}

Mean ES: ${g.avgY.toFixed(2)} (k=${g.k})

`)}

`; // 5. THE CHART WITH INJECTED CHANNELS MAPPING FOR THE INTERACTIVE TIP BOX const chart = Plot.plot({ width: 1100, height: 550, marginLeft: 80, marginBottom: 70, style: { fontSize: "18px", fontFamily: "system-ui, sans-serif" }, color: { legend: true, scheme: "Tableau10" }, r: { range: [4, 25] }, x: { label: `${currentXAxis} →`, grid: true }, y: { label: "↑ Effect Size (d)", grid: true }, marks: [ Plot.ruleY([0], { stroke: "#888", strokeWidth: 2, strokeDasharray: "4,4" }), Plot.linearRegressionY(filteredData, { x: d => Number(d[currentXAxis]), y: d => Number(d.d), stroke: "Study Type", strokeWidth: 4 }), Plot.dot(filteredData, { x: d => Number(d[currentXAxis]), y: d => Number(d.d), r: d => Number(d["Total N"]), fill: "Study Type", fillOpacity: 0.6, stroke: "white", strokeWidth: 1.5, channels: { // UPDATED: Standardized to explicit Group Unique ID reference mapping Study: d => d["Group Unique ID"] || "Unspecified Reference", Type: "Study Type", [currentXAxis]: d => d[currentXAxis], "ES (d)": "d", "N": d => d["Total N"] || d["total N"] || d["N"] || "N/A" }, tip: { format: { x: false, y: false, fill: false, r: false, Study: true, Type: true, [currentXAxis]: true, "ES (d)": true, "N": true } } }) ] }); const legends = chart.querySelectorAll("div, span"); for (const el of legends) { el.style.fontSize = "18px"; } return html`

${resultsTable}

${chart}

`; }4f:T57d,result_moderators = md`

Stratified Moderator Analysis: Education and Number of measures

Stratified meta-regressions identified Education and Number of Measures (Number of Measures) as critical predictors of study success:

• Education: A significant negative moderator for both CT (p = 0.0169) and PE (p = 0.0086), suggesting interventions are most effective for those with lower baseline cognitive reserve. Education explained 31.26% of variance in Exercise outcomes.

• Number of measures: The number of measures per study was a significant negative moderator (p = 0.0067). Larger cognitive batteries were associated with diluted effect sizes, likely due to increased participant fatigue or statistical noise.

`50:T495,disc_universal_benefit = md`

Discussion

The Universal Benefit of Lifestyle Interventions

Our large-scale meta-analysis of 309 RCTs confirms that lifestyle interventions—comprising cognitive training, physical exercise, and nutrition—are robust tools for enhancing cognitive health in older adults (SMD = 0.17–0.28). Rather than being limited to a specific "miracle pill" approach, cognitive gains were observed across almost every assessed construct. This global neuroplastic response supports the shift toward multi-domain lifestyle management in geriatric care.

`51:T6cc,disc_prescription_paradox = md`

Moderators of the Intervention Effect

One of contributions of this study is the identification of the demographic and methodological confounders that drive perceived differences between intervention types.

• Equivalence After Adjustment: While raw subgroup analysis might suggest certain pillars are superior, our adjusted head-to-head model (p = 0.5479) proves that CT and PE are equally potent when education and assessment density are held constant. This suggests that "which intervention is best" may depend less on the activity itself and more on the participant's baseline profile.

• Education as a Clinical Equalizer: We found that Education accounts for a staggering 31.26% of the variance in Physical Exercise outcomes. The inverse relationship suggests that lifestyle interventions are most impactful for those with lower baseline cognitive reserve. This positions physical and cognitive training not just as preventative tools, but as critical instruments for socioeconomic health equity.

`52:[] 2d:{"title":"\n \n The Cognitive Efficacy of Lifestyle Interventions\n \n \n Interactive Meta-Analysis & Open-Source Database\n \n ","cells":[{"id":0,"value":"

\n \n The Cognitive Efficacy of Lifestyle Interventions\n \n \n Interactive Meta-Analysis & Open-Source Database\n \n

\n \n

","pinned":false,"hidden":false,"mode":"md"},{"id":10,"value":"$2f","pinned":false,"hidden":false,"mode":"ojs"},{"id":13,"value":"$30","pinned":false,"hidden":false,"mode":"ojs"},{"id":70,"value":"result_data_attachment = md`\n

\n\n

Analyzed Data

\n The analyzed data based on the method outlined in the data synthesis plan is attached here.\n

\n\n

\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":66,"value":"data = await d3.csv(\"https://docs.google.com/spreadsheets/d/e/2PACX-1vTJ_QoRgYUZQBJ9j8i-EsmiTu2mZjXiEC-DhIEutuVY8K9q9G9J7KWwUr4c-tZiUUkfyzLQLGuAV0AK/pub?gid=0&single=true&output=csv\", d3.autoType)","pinned":false,"hidden":false,"mode":"ojs"},{"id":201,"value":"$31","pinned":false,"hidden":false,"mode":"ojs"},{"id":242,"value":"$32","pinned":false,"hidden":false,"mode":"ojs"},{"id":198,"value":"$33","pinned":false,"hidden":false,"mode":"ojs"},{"id":205,"value":"$34","pinned":false,"hidden":false,"mode":"ojs"},{"id":209,"value":"$35","pinned":false,"hidden":false,"mode":"ojs"},{"id":212,"value":"$36","pinned":false,"hidden":false,"mode":"ojs"},{"id":88,"value":"world = fetch(\"https://cdn.jsdelivr.net/npm/world-atlas@2/countries-110m.json\").then(d => d.json())","pinned":false,"hidden":false,"mode":"ojs"},{"id":77,"value":"$37","pinned":false,"hidden":false,"mode":"ojs"},{"id":81,"value":"$38","pinned":false,"hidden":false,"mode":"ojs"},{"id":91,"value":"$39","pinned":false,"hidden":false,"mode":"ojs"},{"id":94,"value":"$3a","pinned":false,"hidden":false,"mode":"ojs"},{"id":97,"value":"$3b","pinned":false,"hidden":false,"mode":"ojs"},{"id":74,"value":"// This cell should be standalone, usually at the very top or bottom of the notebook\nwebr = {\n const core = new (await import(\"https://webr.r-wasm.org/latest/webr.mjs\")).WebR();\n await core.init();\n // Install the package ONCE here, globally.\n await core.installPackages(['meta']);\n return core;\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":16,"value":"$3c","pinned":false,"hidden":false,"mode":"ojs"},{"id":123,"value":"$3d","pinned":false,"hidden":false,"mode":"ojs"},{"id":130,"value":"$3e","pinned":false,"hidden":false,"mode":"ojs"},{"id":133,"value":"$3f","pinned":false,"hidden":false,"mode":"ojs"},{"id":136,"value":"$40","pinned":false,"hidden":false,"mode":"ojs"},{"id":143,"value":"$41","pinned":false,"hidden":false,"mode":"ojs"},{"id":139,"value":"$42","pinned":false,"hidden":false,"mode":"ojs"},{"id":146,"value":"$43","pinned":false,"hidden":false,"mode":"ojs"},{"id":24,"value":"$44","pinned":false,"hidden":false,"mode":"ojs"},{"id":149,"value":"within_data = await d3.csv(\"https://docs.google.com/spreadsheets/d/e/2PACX-1vTA2tMzFpumwP4eSKq6oLNS7c6GGQaYnZShmGI4DTCE1i_hAXCeUAE2r5yLiT-ccpuBL3o2eT_rJyxs/pub?gid=0&single=true&output=csv\", d3.autoType)","pinned":false,"hidden":false,"mode":"ojs"},{"id":152,"value":"$45","pinned":false,"hidden":false,"mode":"ojs"},{"id":27,"value":"$46","pinned":false,"hidden":false,"mode":"ojs"},{"id":163,"value":"cc_data = await d3.csv(\"https://docs.google.com/spreadsheets/d/e/2PACX-1vSwrmahnum4JtXmPxnT_jN06X_S2cKliMbQFZVRnANJs1gqLA_EovnM26d8qLIabLME4h2MiPKddkVp/pub?gid=0&single=true&output=csv\", d3.autoType)","pinned":false,"hidden":false,"mode":"ojs"},{"id":166,"value":"$47","pinned":false,"hidden":false,"mode":"ojs"},{"id":169,"value":"$48","pinned":false,"hidden":false,"mode":"ojs"},{"id":39,"value":"$49","pinned":false,"hidden":false,"mode":"ojs"},{"id":42,"value":"$4a","pinned":false,"hidden":false,"mode":"ojs"},{"id":175,"value":"$4b","pinned":false,"hidden":false,"mode":"ojs"},{"id":172,"value":"$4c","pinned":false,"hidden":false,"mode":"ojs"},{"id":178,"value":"$4d","pinned":false,"hidden":false,"mode":"ojs"},{"id":181,"value":"$4e","pinned":false,"hidden":false,"mode":"ojs"},{"id":30,"value":"$4f","pinned":false,"hidden":false,"mode":"ojs"},{"id":33,"value":"result_methodological_audit = md`\n

Field Profile Comparison: CT vs. PE Methodological Audit

\n Non-parametric comparisons (Kruskal-Wallis) identified baseline demographic imbalances:\n

\n • Education: A significant difference was found (p = 0.0079). CT cohorts possessed significantly higher median education (13.9 yrs) than PE cohorts (11.9 yrs).\n

\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":45,"value":"$50","pinned":false,"hidden":false,"mode":"ojs"},{"id":48,"value":"$51","pinned":false,"hidden":false,"mode":"ojs"},{"id":51,"value":"disc_control_class_effect = md`\n

\n\n

Methodological Sensitivity: The Control Class Effect

\n A critical finding in our methodological audit was the differential impact of Control Class (Active vs. Passive) across pillars. Physical Exercise efficacy was significantly moderated by control choice (p = 0.0237), with effects nearly doubling when compared against passive groups (0.29) versus active ones (0.14). In contrast, Cognitive Training maintained high therapeutic resilience regardless of control type.\n

\n\n

\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":54,"value":"disc_domain_sensitivity = md`\n

\n\n

Domain Sensitivity: The \"Transfer\" Question

\n Our analysis of 1,046 cognitive observations revealed that interventions were most potent for Global Cognition and Short-Term Memory. Conversely, the smaller effects in processing speed and executive function suggest these \"lower-level\" mechanics may require more targeted, high-intensity protocols.\n

\n\n

\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":57,"value":"disc_research_implications = md`\n

\n\n

Implications for Future Research

\n The significant impact of Number of Measures on effect sizes (p = 0.0067) serves as a warning to future researchers. Expansive neuropsychological batteries, while thorough, appear to introduce \"statistical noise\" or participant fatigue that dilutes the reported efficacy of interventions. Future trials should prioritize parsimonious, construct-valid assessment batteries to better capture true treatment effects.\n

\n\n

\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":60,"value":"disc_conclusion = md`\n

\n\n

Conclusion

\n The \"best\" lifestyle intervention is not a single pillar, but rather the one most tailored to an individual’s cognitive reserve. By accounting for the powerful moderating effects of education and study design, we move toward a precision medicine approach for cognitive aging—one where the prescription paradox is replaced by a data-driven understanding of how to protect the aging brain.\n

\n\n

\n`","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W52","stdlib":"1","titling":"auto"}