Should I register in PROSPERO before starting my search?

Yes — PROSPERO registration must be completed BEFORE you begin your database searches. Retrospective registration (after searches are done) is not accepted. Register as soon as your protocol is finalized. This is both an ethical requirement and increasingly a journal submission requirement.

The AI said meta-analysis is "not recommended." What do I do?

This means the studies you plan to include are likely too heterogeneous to pool statistically. Follow the SWiM (Synthesis Without Meta-analysis) guidelines for narrative synthesis. This is a valid and publishable approach — do not force a meta-analysis when study designs are incompatible, as it will be detected by reviewers and questioned.

Do I really need a medical librarian for the search strategy?

Strongly recommended. Medical librarians are specialists in database search construction. A poorly constructed search string can miss critical studies, which is the most common reason systematic reviews are rejected or deemed unreliable. Most academic medical centers provide this as a free service to researchers.

What is the minimum number of studies needed for a meta-analysis?

There is no fixed minimum, but most statisticians recommend at least 3–5 studies for pooling. With fewer studies, confidence intervals will be extremely wide, and publication bias assessment (funnel plots) becomes unreliable. Below 5 studies, a narrative synthesis is usually preferable.

Should I include gray literature (conference abstracts, unpublished studies)?

Yes — omitting gray literature introduces publication bias (positive results are published more often). Include a search of ClinicalTrials.gov, WHO ICTRP, relevant conference proceedings, and contact authors of ongoing trials where possible. Include gray literature in primary analysis; run a sensitivity analysis excluding it.

How reliable is the AI output? Should I verify everything before using it?

Yes — always verify before using. AI models can generate plausible-sounding but incorrect citations, statistics, or methodological claims (known as "hallucinations"). This prompt includes rules instructing the AI to mark uncertain references as [needs verification] or [not verified via PubMed], but the AI does not always flag every uncertain claim. Before citing any paper the AI mentions: (1) search PubMed to confirm it exists and details are correct; (2) read the abstract to verify it supports the claim; (3) never cite a paper you have not personally reviewed. Treat the AI's output as a starting point, not a final answer.

Systematic Review & Meta-Analysis Planner -- medPrompt

Usage Guide

Fill in your PICO question and review scope.
Click AI Run — receive a complete systematic review protocol directly in the chat.
Review each component and ask for refinements to search strategy, inclusion criteria, or risk-of-bias tools.

Medical Research Assistant

Fill in variables and run directly with AI

Wiki

What this tool does

A PROSPERO-ready systematic review and meta-analysis protocol generator for clinical researchers. Provide your research question, PICO elements, study types, databases, and whether meta-analysis is planned — the AI builds a complete, submission-quality protocol covering PICO/PICOS formalization, database-specific search string construction, inclusion/exclusion criteria table, PRISMA flow plan, risk-of-bias tool selection, and heterogeneity pre-check. A scope guard is built in to flag over-broad protocols before you commit weeks of work.

Protocol components covered

The generated protocol covers every section required for PROSPERO registration and most journal submission requirements: a structured PICO/PICOS question, Boolean search strings optimized for PubMed/MEDLINE, Embase, and CENTRAL, a full inclusion/exclusion criteria table with rationale, a PRISMA 2020-compliant flow diagram plan, selection of the appropriate risk-of-bias tool (RoB 2 for RCTs, ROBINS-I for non-randomized studies, QUADAS-2 for diagnostic studies, Newcastle-Ottawa for observational studies, SYRCLE for animal studies), and a pre-specified data extraction form.

Meta-analysis feasibility assessment

Before committing to a quantitative synthesis, the AI assesses whether your planned studies are sufficiently similar to pool statistically. Key checks: clinical heterogeneity (do the populations, interventions, and outcomes overlap enough?), methodological heterogeneity (are the study designs compatible?), and statistical heterogeneity (anticipated I² range). If the feasibility check suggests high heterogeneity, the AI recommends narrative synthesis following SWiM guidelines rather than forcing a potentially misleading pooled estimate.

Risk-of-bias tool selection guide

Choosing the wrong tool is a common error that reviewers flag immediately. The AI selects based on study type: RoB 2 for parallel-group and crossover RCTs; ROBINS-I for non-randomized comparative studies; QUADAS-2 for diagnostic test accuracy studies; Newcastle-Ottawa Scale for cohort and case-control studies; SYRCLE for animal experimental studies. For mixed-design reviews including both RCTs and observational studies, both RoB 2 and ROBINS-I are included with guidance on how to report combined assessments.

Scope guard

One of the most common reasons systematic reviews are never completed — or rejected — is an over-broad scope. The AI checks your protocol for scope creep signals: excessively broad population definitions, too many PICO combinations (recommending a split into separate reviews), an outcome list that is too extensive to extract reliably, and time frames that make the search volume unmanageable. Scope issues are flagged with specific, actionable recommendations before you invest in the full search.

Limitations and recommendations

The AI generates a protocol template based on your inputs; it does not conduct the actual database searches or retrieve papers. For complex or high-impact reviews, involving a medical librarian for search string peer review and a second independent screener for study selection are strongly recommended to meet Cochrane and PRISMA standards.

FAQ