HARMONi-3

Every edition, the free report included, names who said each statement, as the source shows them. Each named voice's card gives a clinical role only when the NPI registry or the person's own institution's staff page confirms it, and says which; a role only a source states is labelled as the source's. Here 21 of the 41 cited voices have a role confirmed that way. Every edition also comes with MCP access, so your AI agent reads the same evidence with your license key. The verified speaker identities add-on, sold with 12-month plans, adds each verified person's IDs as data (the NPI number, X, Bluesky and LinkedIn accounts, ORCID and institutional pages), so your team can check each attribution itself. It is for checking attribution inside your organization, never for outreach or targeting, and it does not unlock who said what.

Discussed means at least one verified statement addresses the aspect, not that every question in it is covered: gaps inside a discussed aspect are listed under Scope and limits, and the full report gives each aspect its own limitations. No commentary found means none in the sources read. Assessment incomplete marks an aspect with no statement found while known source limits apply.

Screened window

May 3, 2023 → Oct 28, 2026

Cited window

Oct 14, 2023 → Oct 28, 2026

Retrieved

Nov 4, 2026

Title

A Randomized, Double-blind, Multiregional Phase 3 Study of Ivonescimab Combined With Chemotherapy Versus Pembrolizumab Combined With Chemotherapy for the First-line Treatment of Metastatic Non-small Cell Lung Cancer (HARMONi-3)

Registry

NCT05899608

Study IDs

SMT112-3003

Sponsor

Summit Therapeutics

Design

Phase 3 · Interventional, randomized, parallel

Status

Active, not recruiting

Enrollment

1,600 estimated

Registered

2023-06-12

Start

2023-10-26 actual

Primary completion

2028-12-31 estimated

Primary endpoint

Overall Survival (OS)

Interventions

2 registered

This report does not judge efficacy, estimate HCP prevalence or consensus or give treatment advice.

• –Discussion that does not name the trial may be missing from this edition.
• –For 1 recording, a transcript separated by speaker was not available, so quotes come from the original automatic transcript and may lack speaker labels.
• –Not yet captured in this edition: a view on global applicability, patient selection and VEGF-related toxicity in first-line ivonescimab-based treatment.
• –Not yet captured in this edition: a view on immunotherapy plus an ADC as an emerging first-line NSCLC treatment alternative.
• –Not yet captured in this edition: a view on immunotherapy plus an ADC as an emerging first-line treatment alternative in non-small cell lung cancer.
• –1 more clinician view in the sources used in this report is not yet captured as evidence in this edition.
• –131 passages in 90 sources used in this report were not fully assessed for clinician views, so a view there may be missing from this edition.
• –67 statements in the sources reviewed for this edition are about other patients or topics, or only mention the trial, and are not counted.
• –Further source limits: discussion that describes the trial without naming it may be incomplete; some discussion that describes the trial without naming it may be missing; 70 sources about other trials or topics are not included.
• –The sources reviewed for this edition are a selected collection, not a representative sample of HCP opinion. Repeated posts, relays and overlapping recordings do not establish additional independent speakers or consensus.
• –Some potentially relevant passages were not fully assessed. Discussions that did not use recognizable trial names or that were transcribed differently may be missing, so the report cannot claim exhaustive coverage.
• –Several passages, particularly on the September 2024 episode and in a short recap video, cannot be assigned to named speakers. They may belong to participants already discussed and are not counted as separate people.
• –The material mixes clinical discussion with author, organization and investment-account commentary. A verified identity, a source-described credential or a dated affiliation does not establish expertise, independence or the truth of a view.
• –No exact HARMONi-3 results publication, complete protocol or statistical plan, or exact regulatory label was available. Interim thresholds, modeled success probabilities, approval expectations and historical amendments are therefore reported at the level the sources support.
• –Findings from HARMONi-2, HARMONi-6 and other studies are context for the first-line treatment question, not HARMONi-3 results. Cross-trial estimates, subgroup interpretations and different hemorrhage definitions cannot be combined into a single comparative conclusion.
• –Dates attached to posts or published recordings do not necessarily date the underlying session or readout. Different timing statements are kept as contemporaneous expectations, not interpreted as demonstrated changes of mind.
• –Some speakers' identities or clinical roles are not established in the sources used in this report; each statement is attributed only as far as its source shows.

How reports are built and checked: Methods.

Each new edition gets a new version: a new or changed finding raises the middle number (1.0.0 to 1.1.0), added evidence or newly verified speakers alone the last number (1.0.0 to 1.0.1), and a change of report format the first number. A correction fixes the published edition in place without making a new edition, so the version stays the same. Changes are corrections that add, remove or change a statement or finding, or withdraw a source; routine maintenance (people cards, names and spellings, recounts, tags and page wording) changes none. Every correction is listed here with its date. Last checked for new commentary Nov 4, 2026 UTC. No changes. Follow new editions and changes in this report's change feed (JSON Feed), or the catalog's.

Paste this into ChatGPT, Claude, Microsoft Copilot or your own agent. It previews this report through the free Erudio MCP server, tells you whether it answers your question, and orders it if you ask. Some agents ask you to confirm pasted instructions before acting; that is expected.

Help me decide whether to buy the Erudio report "HARMONi-3" (NCT05899608), and buy it for me if I ask. Erudio reports cover what clinicians say in public about one clinical trial, with every verified statement linked to its source.

1. Use the free Erudio MCP server at https://mcp.erudio.com/free; it needs no sign-in. Nothing needs installing: if it is not connected in this app and you can make HTTP requests, call it directly with JSON-RPC POSTs to that URL (initialize, then tools/call), sending "Accept: application/json, text/event-stream". Call get_report_card with slug "harmoni-3-nct05899608". If you cannot make requests, read https://www.erudio.com/reports/harmoni-3-nct05899608.md instead.
2. In a few lines, tell me what the report covers: its findings, its verified statements and sources, its evidence window, what it leaves out, and the price of each product (offer). Then ask what I need it for, and say plainly whether this report answers it.
3. If I want it, ask for my work email and the product: the current edition, or 3, 6 or 12 months of updates. Call create_order with report "NCT05899608", that product and email, then show me the quote and the checkout_url to pay by card. If you have a payment client for HTTP 402 payments and I agree, you can pay the pay_url instead. Nothing is charged until one is paid.
4. Once it is paid, call get_order with the status_url: it returns my license key. Then use the Erudio MCP server at https://mcp.erudio.com/mcp with the header "Authorization: Bearer <key>". Call read_report with report_id NCT05899608 and sections ["all"], and summarize the findings for me with their sources. The full report is large (about 500 KB): save the response to a file and work from it rather than printing it. If you cannot, read sections ["narratives"] first, the executive summary and analysis.

When you tell me what clinicians said, give each view as theirs, with its source. Counts are the sources the report kept, not the share of clinicians who agree. Reports are not medical advice.
Keep the key private to my organization.

curl https://www.erudio.com/reports/harmoni-3-nct05899608.json
curl https://www.erudio.com/reports/harmoni-3-nct05899608.md

# MCP, no sign-in (https://mcp.erudio.com/free)
get_report_card({ "slug": "harmoni-3-nct05899608" })