Quickstart: from research question to report

Before you start searching, set up the question and eligibility criteria that drive every downstream step:

• Pick a framework: PICO for intervention reviews (Population, Intervention, Comparator, Outcomes) or PEO for prevalence and risk-factor reviews (Population, Exposure, Outcomes).
• In the Configure tab, fill in each clause — population, intervention/exposure, comparator (if applicable), and outcomes — or describe your question in chat and let the AI build the structured protocol for you.
• Add screening questions and inclusion/exclusion rules. These flow into AI-assisted abstract screening and full-text triage downstream.

app.axelium.io · Configure

Start by identifying the studies you want to include:

• Use the Search step to add trials by NCT ID and publications by PMID or keyword.
• Alternatively, click Generate strategy (AI) at the top of the Search tab to have the agent build and validate optimized PubMed + ClinicalTrials.gov queries from your PICO/PEO. The agent screens scouted results against your criteria in real time and self-corrects if they fail; choose Broad (recall), Balanced, or Tight (precision) mode. If you haven't configured the analysis yet, an inline form (Population + Intervention + optional Comparator/Outcomes/Year range/Study design) lets you generate a strategy directly from the Search tab.
• Know a few highly-relevant papers already? Paste up to 5 PMIDs into the Seed studies (pearls) card that appears below Generate strategy. Pearls are folded into the plan as extra vocabulary (MeSH + author keywords) and wrapped into the final PubMed term as an OR (PMID[UID]) clause, so the agent can't miss them.
• Trials (with nctId) and publications (with pmid/ doi) are stored as studies.
• ClinicalTrials.gov trials are automatically linked to their publications when PMIDs are present in the registry — each trial becomes a canonical study with one or more reports.
• Already have a library in EndNote, Zotero, Mendeley, PubMed, Embase, Web of Science, or Ovid? Import RIS / BibTeX reads the exported file, previews what will be added, and flags duplicates of studies already in the project. Exporting back out to RIS or BibTeX is one click from the same page.
• Additional databases — Search now spans PubMed and ClinicalTrials.gov plus five free secondary sources (OpenAlex, Europe PMC, MAUDE, ERIC, and a CENTRAL approximation) so you can broaden coverage beyond the primary indexes. See concepts for the full list and routing rules.
• Discover (citation chasing) — Use the Discover tab to find papers your Boolean query missed via Semantic Scholar references, citers, and recommendations. See concepts.

app.axelium.io · Search

Move to the Full Text tab and obtain PDFs and supplementary materials:

• The system tries to auto-download a PDF from open access sources or known publisher links. When automatic resolvers fail, a the system automatically discovers additional sources.
• Supplements are auto-fetched alongside the main PDF — including PDF appendices, Excel tables, and Word documents from PMC and publisher sites. Office formats are automatically parsed and processed.
• Successful downloads are marked as Parsed once the PDF has been processed into text and tables.
• If a PDF cannot be downloaded (e.g. paywalled, not found), the row shows Manual upload. Use Upload PDF manually to attach your own copy, or use Bulk Upload to drag‑and‑drop up to 50 PDFs at once with automatic study matching.
• Publisher-blocked papers — Captcha-walled publishers are auto-detected and stamped fetch_blocked. Before reporting failure, an OA recovery layer (PMC mirrors and NCBI efetch) backfills automatically; anything still missing is surfaced on the Bulk upload missing fulltexts page (/analysis/[id]/fulltext/missing), where you can drag-drop a folder of PDFs and the matcher auto-pairs filenames to studies.

app.axelium.io · Full text

Not every parsed document is suitable for data extraction. The system automatically filters out non-results documents:

• Trials that are still recruiting or have no posted results on ClinicalTrials.gov are not shown in the Extraction tab.
• Protocol-only or feasibility papers are excluded from extraction — they describe planned methods, not actual outcomes.
• Trials become eligible once they have CT.gov posted results, or linked primary_result / secondary_analysis publications with parsed PDFs.

Once eligible studies have parsed full text, extract endpoint data and compute pooled effects:

• Use Auto‑Extract All Outcomes to batch‑extract data across all eligible studies in parallel. The system handles document acquisition, companion paper resolution, and multi‑agent extraction automatically.
• Each extracted value carries a confidence badge (green/amber/red) and a provenance trail linking back to the source snippet in the PDF. Values below confidence thresholds are routed to a Review Queue for human validation.
• The Live Effect Stripe above the data grid shows the derived effect estimate (Est, SE, 95% CI) in real time — green when plausible, amber when warnings are detected.
• The math engine checks and derives missing quantities (e.g. SDs from SEs or CIs), detects arm swaps, and validates consistency before accepting data.
• When sources conflict (e.g., registry vs. PDF values), the conflict side‑by‑side view lets you compare and choose with one click.
• Use the Triage Matrix view (toggle button in the center panel header) for a cross-study overview of extraction progress across all endpoints.
• In the Analysis tab, build a dataset for your endpoint and run fixed-effect or random-effects models.

app.axelium.io · Extraction

app.axelium.io · Analysis

Turn your analysis into a narrative report using the Evidence Board and Reports:

• From the Analysis view, pin important plots and tables as snapshots on the Evidence Board.
• Pin key chat messages (e.g. explanations, decisions) and PRISMA summaries so they are available as evidence.
• In the Reports & Evidence tab, group evidence by section (Introduction, Methods, Results, Discussion).
• Use Generate Draft to create a report from only the evidence you selected. Pick a target audience from seven presets — Academic, Clinical, Regulatory, Internal Stakeholders, HTA, Payer, and Patient (plus Custom) — to tailor the tone and detail level, then edit and export as DOCX or XLSX.
• Need to hand off to Cochrane authors? RevMan 5 export (.rm5) packages the review for RevMan with comparisons and outcomes mapped to the appropriate data type. See concepts for which measures map to DICH_OUTCOME, CONT_OUTCOME, and IV_OUTCOME.

app.axelium.io · Reports

Next: keep the review current

A finished review does not have to be a frozen snapshot. Once your protocol and search queries are locked, you can turn the project into a living review — Axelium then re-runs this whole pipeline on a daily, weekly, or monthly schedule, collects newly-found studies in a Review Inbox, and emails you when the pooled result changes. Set it up from the Living-Review Automation dialog on the Configure sidebar.