> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pav.bio/llms.txt
> Use this file to discover all available pages before exploring further.

# Coverage and sources

> Where each dataset comes from, what it covers, and which fields are partial.

Pav data is built from public sources. Every record links back to where it
came from, so any value can be checked against its source.

## Sources

| Dataset         | Source                                                                                                      | Provenance on each record                                                                     |
| --------------- | ----------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------- |
| Programs        | Pipelines published by the companies themselves                                                             | `source_url` (the company page) and `source_checked_at`                                       |
| Companies       | Company pipelines, ClinicalTrials.gov sponsors, SEC filings and USPTO records, linked to one Pav company id | `last_checked_at` for the pipeline                                                            |
| Clinical trials | ClinicalTrials.gov                                                                                          | `nct_url`                                                                                     |
| Deals           | SEC filings and company press releases                                                                      | `events[].source_url`, `source_documents[]`, and an `evidence` sentence for each stated value |
| Patents         | USPTO applications, grants, prosecution and maintenance events, and assignment records                      | Application and patent numbers on every member and event                                      |
| FDA records     | FDA Orange Book, Purple Book, orphan drug designations, warning letters and recalls                         | `source`, `source_version`, and `source_url` where FDA publishes a page                       |
| Ontology terms  | MONDO (indications), NCIT (targets)                                                                         | `term_id` on each term                                                                        |

## Program field coverage

Some fields are not disclosed by every company. Measured on all 27,296
programs on 28 September 2026:

| Field                                                  | Programs with a value |
| ------------------------------------------------------ | --------------------- |
| Normalized phase (`phase_norm` other than `Unknown`)   | 97%                   |
| At least one linked clinical trial (`clinical_trials`) | 90%                   |
| Indication mapped to MONDO (`indication_terms`)        | 51%                   |
| Normalized modality (`modality_norm`)                  | 29%                   |
| Mechanism of action (`mechanism_of_action`)            | 28%                   |
| Target (`target`)                                      | 26%                   |
| Target mapped to NCIT (`target_terms`)                 | 23%                   |

<Warning>
  A filter on a partial field returns only programs where the field is set.
  `target=KRAS` cannot return a KRAS program whose company publishes no target.
  For complete landscapes, combine filters with relevance search (`q`). See
  [Search and filters](/concepts/search-and-filters).
</Warning>

A small number of programs have `drug` set to `Undisclosed`, where the company
publishes a program without naming the asset.

Check current phase and modality counts at any time with
[`GET /v1/stats`](/api-reference/stats/dataset-summary-statistics).

## Deals

* Many deals disclose no value. `min_value_usd` and `sort=headline_value`
  consider only deals with a stated USD headline value.
* `companies` lists parties linked to Pav companies. A party that is not a
  Pav company is named in the deal text but is not in `companies`.
* Each stated value carries `raw_text` and the `evidence` sentence it was read
  from. Read `raw_text` for qualifiers such as "up to" or "approximately".

## Patents

* Coverage is US families only.
* About a quarter of families are linked to a Pav company. No rows for a
  `company_id` does not mean the company holds no patents; search its owner
  name with `q`.
* `statutory_term_end` is the earliest nonprovisional filing date plus 20
  years, with no term adjustments or extensions.

## FDA records

* Orange Book, Purple Book and orphan-designation records reflect FDA's
  current published edition. Their `record_key` changes with each FDA edition; use
  `application_key`, `product_key`, `patent_number` or `designation_id` as
  stable handles.
* Company links cover about 68% of Orange Book products, 74% of orphan
  designations, 27% of recalls and about 5% of warning letters. Search the FDA
  organization name with `q` as well as filtering by `company_id`.
* Warning letters start in 2021 and recalls in 2017.

## Clinical trials

* Trials are ClinicalTrials.gov studies only.
* `programs` and `company_ids` list only the links Pav has made. A trial with
  no linked program can still belong to a tracked company; check
  `lead_sponsor.company_id`.

## Freshness

| Dataset         | How to see freshness                                                  |
| --------------- | --------------------------------------------------------------------- |
| Programs        | `source_checked_at` on each program; `last_checked_at` on the company |
| Clinical trials | `last_update_post_date` from the registry                             |
| Deals           | `latest_update_at` and `activity_date`                                |
| Patents         | `latest_status_date` on the family and `status_date` on each member   |
| FDA records     | `source_version` (edition) and `document_date`                        |
| Changes         | `detected_at` on each event                                           |

## Report a data issue

Send the record id and the `request_id` of the response to Pav at
[pav.bio](https://pav.bio).