Datasets:

anamnesis-data
/

clinical_trials_history_sample

-- Sample queries for the ClinicalTrials.gov version-history sample dataset (DuckDB).
--
-- Run them from the directory that holds the sample files (one folder per table, plus
-- setup.sql). In the DuckDB CLI:
--     .read setup.sql
--     .read sample_queries.sql
-- setup.sql creates one view per table, named after its folder ("core", "interventions",
-- "results_outcome_measures", ...).
--
-- Reading the data:
--   * The dataset is a version history: nct_version numbers a trial's versions from 0, and
--     nearly every table has one set of rows per (nct_id, nct_version).
--   * There is no "current" flag. The current record is the highest nct_version held for an
--     nct_id. That is the latest version in this dataset, which can be older than the live
--     registry record if the trial has changed since it was last collected.
--   * Every table ends with is_obsolete / obsolete_by: true when ClinicalTrials.gov now
--     resolves the ID to a different registration.
--
-- Each query starts with "-- name:" and "-- level:" lines; the blocks are independent.

-- name: current_record
-- level: basic
-- The current record of every trial: one row per nct_id, its highest nct_version.
SELECT nct_id, nct_version, overall_status, lead_sponsor_name, start_date,
       primary_completion_date, enrollment_count, last_update_post_date
FROM core
QUALIFY row_number() OVER (PARTITION BY nct_id ORDER BY nct_version DESC) = 1
ORDER BY nct_id;

-- name: versions_per_trial
-- level: basic
-- How many versions each trial has, and the span of dates they were posted.
SELECT nct_id,
       count(*) AS versions,
       min(last_update_post_date) AS first_posted,
       max(last_update_post_date) AS last_posted,
       count(DISTINCT overall_status) AS distinct_statuses
FROM core
GROUP BY nct_id
ORDER BY versions DESC;

-- name: status_changes_for_one_trial
-- level: basic
-- Every version of one trial in which the overall status changed.
WITH v AS (
    SELECT nct_id, nct_version, last_update_post_date, overall_status,
           lag(overall_status) OVER (PARTITION BY nct_id ORDER BY nct_version) AS previous_status
    FROM core
    WHERE nct_id = 'NCT03709680'
)
SELECT nct_version, last_update_post_date, previous_status, overall_status
FROM v
WHERE previous_status IS DISTINCT FROM overall_status
ORDER BY nct_version;

-- name: current_interventions
-- level: basic
-- Interventions of each trial's current version (core joined to interventions on the
-- two-column version key).
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
)
SELECT c.nct_id, c.nct_version, i.intervention_type, i.intervention_name
FROM current AS cur
INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
INNER JOIN interventions AS i ON i.nct_id = c.nct_id AND i.nct_version = c.nct_version
ORDER BY c.nct_id, i.intervention_type, i.intervention_name;

-- name: current_trials_by_phase_and_status
-- level: basic
-- Current trials by phase and overall status. A trial with several phases (for example
-- PHASE2 and PHASE3) counts once per phase.
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
)
SELECT p.phase, c.overall_status, count(*) AS trials
FROM current AS cur
INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
INNER JOIN phases AS p ON p.nct_id = c.nct_id AND p.nct_version = c.nct_version
GROUP BY p.phase, c.overall_status
ORDER BY p.phase, trials DESC;

-- name: trial_profile
-- level: advanced
-- One row per trial combining six tables at the current version. Each child table is
-- aggregated to one row per trial first, so the joins cannot multiply rows.
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
),
phase_agg AS (
    SELECT nct_id, nct_version, string_agg(phase, ', ' ORDER BY phase) AS phases
    FROM phases GROUP BY nct_id, nct_version
),
condition_agg AS (
    SELECT nct_id, nct_version, count(*) AS conditions,
           string_agg(condition, '; ' ORDER BY condition) AS condition_list
    FROM conditions GROUP BY nct_id, nct_version
),
intervention_agg AS (
    SELECT nct_id, nct_version, count(*) AS interventions,
           string_agg(DISTINCT intervention_type, ', ') AS intervention_types
    FROM interventions GROUP BY nct_id, nct_version
),
location_agg AS (
    SELECT nct_id, nct_version, count(*) AS sites, count(DISTINCT country) AS countries
    FROM locations GROUP BY nct_id, nct_version
),
collaborator_agg AS (
    SELECT nct_id, nct_version, count(*) AS collaborators
    FROM collaborators GROUP BY nct_id, nct_version
)
SELECT c.nct_id, c.overall_status, c.lead_sponsor_name, ph.phases, c.enrollment_count,
       cd.conditions, i.interventions, i.intervention_types,
       coalesce(l.sites, 0) AS sites, coalesce(l.countries, 0) AS countries,
       coalesce(co.collaborators, 0) AS collaborators
FROM current AS cur
INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
LEFT JOIN phase_agg AS ph ON ph.nct_id = c.nct_id AND ph.nct_version = c.nct_version
LEFT JOIN condition_agg AS cd ON cd.nct_id = c.nct_id AND cd.nct_version = c.nct_version
LEFT JOIN intervention_agg AS i ON i.nct_id = c.nct_id AND i.nct_version = c.nct_version
LEFT JOIN location_agg AS l ON l.nct_id = c.nct_id AND l.nct_version = c.nct_version
LEFT JOIN collaborator_agg AS co ON co.nct_id = c.nct_id AND co.nct_version = c.nct_version
ORDER BY c.nct_id;

-- name: registry_as_of_date
-- level: advanced
-- What the registry said about each trial on a past date: the latest version posted on or
-- before that date, with its conditions, compared with today's status. Change as_of below.
WITH params AS (
    SELECT DATE '2019-01-01' AS as_of
),
as_of_version AS (
    SELECT c.nct_id, c.nct_version
    FROM core AS c
    CROSS JOIN params AS p
    WHERE c.last_update_post_date <= p.as_of
    QUALIFY row_number() OVER (PARTITION BY c.nct_id ORDER BY c.nct_version DESC) = 1
),
current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
),
condition_agg AS (
    SELECT nct_id, nct_version, string_agg(condition, '; ' ORDER BY condition) AS conditions
    FROM conditions GROUP BY nct_id, nct_version
)
SELECT then_.nct_id, then_.nct_version AS version_then, then_.overall_status AS status_then,
       then_.primary_completion_date AS primary_completion_then, cd.conditions AS conditions_then,
       now_.nct_version AS version_now, now_.overall_status AS status_now,
       now_.primary_completion_date AS primary_completion_now
FROM as_of_version AS a
INNER JOIN core AS then_ ON then_.nct_id = a.nct_id AND then_.nct_version = a.nct_version
LEFT JOIN condition_agg AS cd ON cd.nct_id = then_.nct_id AND cd.nct_version = then_.nct_version
INNER JOIN current AS cur ON cur.nct_id = then_.nct_id
INNER JOIN core AS now_ ON now_.nct_id = cur.nct_id AND now_.nct_version = cur.nct_version
ORDER BY then_.nct_id;

-- name: primary_completion_date_slips_by_phase
-- level: advanced
-- How often sponsors push out a trial's primary completion date, by phase. Each version is
-- compared with the one before it (core), dated from version_history (ct_gov_version equals
-- nct_version) and attributed to the phases of the version that made the change.
-- A trial with several phases counts once per phase. Days are gross: a date moved earlier
-- and later again counts each push.
WITH dated AS (
    SELECT c.nct_id, c.nct_version, c.primary_completion_date,
           lag(c.primary_completion_date) OVER (PARTITION BY c.nct_id ORDER BY c.nct_version)
               AS previous_date,
           h.change_date
    FROM core AS c
    INNER JOIN version_history AS h
        ON h.nct_id = c.nct_id AND h.ct_gov_version = c.nct_version
),
slips AS (
    SELECT *, date_diff('day', previous_date, primary_completion_date) AS slip_days
    FROM dated
    WHERE primary_completion_date > previous_date
)
SELECT p.phase,
       count(DISTINCT s.nct_id) AS trials_pushed,
       count(*) AS pushes,
       round(avg(s.slip_days)) AS avg_days_per_push,
       max(s.slip_days) AS longest_push_days,
       min(s.change_date) AS first_push_posted,
       max(s.change_date) AS last_push_posted
FROM slips AS s
INNER JOIN phases AS p ON p.nct_id = s.nct_id AND p.nct_version = s.nct_version
GROUP BY p.phase
ORDER BY pushes DESC;

-- name: primary_outcome_results
-- level: advanced
-- Reported results for primary outcomes at the current version: outcome, group, category
-- and measured value. Joins five tables through the outcome / class / category ordinals.
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
)
SELECT om.nct_id, om.title AS outcome, om.unit_of_measure,
       g.title AS group_title, cat.title AS category, ms.value, ms.spread
FROM current AS cur
INNER JOIN results_outcome_measures AS om
    ON om.nct_id = cur.nct_id AND om.nct_version = cur.nct_version AND om.type = 'PRIMARY'
INNER JOIN results_outcome_classes AS cl
    ON cl.nct_id = om.nct_id AND cl.nct_version = om.nct_version AND cl.outcome_ord = om.outcome_ord
INNER JOIN results_outcome_categories AS cat
    ON cat.nct_id = cl.nct_id AND cat.nct_version = cl.nct_version
   AND cat.outcome_ord = cl.outcome_ord AND cat.class_ord = cl.class_ord
INNER JOIN results_outcome_measurements AS ms
    ON ms.nct_id = cat.nct_id AND ms.nct_version = cat.nct_version
   AND ms.outcome_ord = cat.outcome_ord AND ms.class_ord = cat.class_ord
   AND ms.category_ord = cat.category_ord
INNER JOIN results_outcome_groups AS g
    ON g.nct_id = ms.nct_id AND g.nct_version = ms.nct_version
   AND g.outcome_ord = ms.outcome_ord AND g.group_id = ms.group_id
ORDER BY om.nct_id, om.outcome_ord, cl.class_ord, cat.category_ord, g.group_id;

-- name: serious_adverse_event_rates_by_group
-- level: advanced
-- Serious adverse events at the current version: for each trial, arm group and event term,
-- how many participants were affected out of those at risk (events, event stats and groups).
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
)
SELECT e.nct_id, g.title AS group_title, e.organ_system, e.term,
       s.num_affected, s.num_at_risk,
       round(100.0 * s.num_affected / nullif(s.num_at_risk, 0), 2) AS pct_affected
FROM current AS cur
INNER JOIN results_adverse_serious_events AS e
    ON e.nct_id = cur.nct_id AND e.nct_version = cur.nct_version
INNER JOIN results_adverse_serious_event_stats AS s
    ON s.nct_id = e.nct_id AND s.nct_version = e.nct_version AND s.event_ord = e.event_ord
INNER JOIN results_adverse_event_groups AS g
    ON g.nct_id = s.nct_id AND g.nct_version = s.nct_version AND g.group_id = s.group_id
WHERE s.num_affected > 0
ORDER BY pct_affected DESC, e.nct_id, e.term
LIMIT 50;

-- name: arms_and_their_interventions
-- level: advanced
-- Which intervention each study arm receives, at the current version (arm_groups joined to
-- the arm/intervention link table and to interventions for the description).
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
)
SELECT a.nct_id, a.arm_label, a.arm_type, i.intervention_type, i.intervention_name
FROM current AS cur
INNER JOIN arm_groups AS a ON a.nct_id = cur.nct_id AND a.nct_version = cur.nct_version
INNER JOIN intervention_arm_links AS l
    ON l.nct_id = a.nct_id AND l.nct_version = a.nct_version AND l.arm_label = a.arm_label
INNER JOIN interventions AS i
    ON i.nct_id = l.nct_id AND i.nct_version = l.nct_version
   AND i.intervention_name = l.intervention_name AND i.intervention_type = l.intervention_type
ORDER BY a.nct_id, a.arm_label, i.intervention_name;

-- name: sponsor_site_footprint
-- level: advanced
-- Each lead sponsor's current trials and where they run: trials, sites and countries
-- (core joined to locations at the current version).
WITH current AS (
    SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
)
SELECT c.lead_sponsor_name,
       count(DISTINCT c.nct_id) AS trials,
       count(*) AS sites,
       count(DISTINCT l.country) AS countries
FROM current AS cur
INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
INNER JOIN locations AS l ON l.nct_id = c.nct_id AND l.nct_version = c.nct_version
GROUP BY c.lead_sponsor_name
ORDER BY sites DESC;

	-- Sample queries for the ClinicalTrials.gov version-history sample dataset (DuckDB).
	--
	-- Run them from the directory that holds the sample files (one folder per table, plus
	-- setup.sql). In the DuckDB CLI:
	-- .read setup.sql
	-- .read sample_queries.sql
	-- setup.sql creates one view per table, named after its folder ("core", "interventions",
	-- "results_outcome_measures", ...).
	--
	-- Reading the data:
	-- * The dataset is a version history: nct_version numbers a trial's versions from 0, and
	-- nearly every table has one set of rows per (nct_id, nct_version).
	-- * There is no "current" flag. The current record is the highest nct_version held for an
	-- nct_id. That is the latest version in this dataset, which can be older than the live
	-- registry record if the trial has changed since it was last collected.
	-- * Every table ends with is_obsolete / obsolete_by: true when ClinicalTrials.gov now
	-- resolves the ID to a different registration.
	--
	-- Each query starts with "-- name:" and "-- level:" lines; the blocks are independent.

	-- name: current_record
	-- level: basic
	-- The current record of every trial: one row per nct_id, its highest nct_version.
	SELECT nct_id, nct_version, overall_status, lead_sponsor_name, start_date,
	primary_completion_date, enrollment_count, last_update_post_date
	FROM core
	QUALIFY row_number() OVER (PARTITION BY nct_id ORDER BY nct_version DESC) = 1
	ORDER BY nct_id;

	-- name: versions_per_trial
	-- level: basic
	-- How many versions each trial has, and the span of dates they were posted.
	SELECT nct_id,
	count(*) AS versions,
	min(last_update_post_date) AS first_posted,
	max(last_update_post_date) AS last_posted,
	count(DISTINCT overall_status) AS distinct_statuses
	FROM core
	GROUP BY nct_id
	ORDER BY versions DESC;

	-- name: status_changes_for_one_trial
	-- level: basic
	-- Every version of one trial in which the overall status changed.
	WITH v AS (
	SELECT nct_id, nct_version, last_update_post_date, overall_status,
	lag(overall_status) OVER (PARTITION BY nct_id ORDER BY nct_version) AS previous_status
	FROM core
	WHERE nct_id = 'NCT03709680'
	)
	SELECT nct_version, last_update_post_date, previous_status, overall_status
	FROM v
	WHERE previous_status IS DISTINCT FROM overall_status
	ORDER BY nct_version;

	-- name: current_interventions
	-- level: basic
	-- Interventions of each trial's current version (core joined to interventions on the
	-- two-column version key).
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	)
	SELECT c.nct_id, c.nct_version, i.intervention_type, i.intervention_name
	FROM current AS cur
	INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
	INNER JOIN interventions AS i ON i.nct_id = c.nct_id AND i.nct_version = c.nct_version
	ORDER BY c.nct_id, i.intervention_type, i.intervention_name;

	-- name: current_trials_by_phase_and_status
	-- level: basic
	-- Current trials by phase and overall status. A trial with several phases (for example
	-- PHASE2 and PHASE3) counts once per phase.
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	)
	SELECT p.phase, c.overall_status, count(*) AS trials
	FROM current AS cur
	INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
	INNER JOIN phases AS p ON p.nct_id = c.nct_id AND p.nct_version = c.nct_version
	GROUP BY p.phase, c.overall_status
	ORDER BY p.phase, trials DESC;

	-- name: trial_profile
	-- level: advanced
	-- One row per trial combining six tables at the current version. Each child table is
	-- aggregated to one row per trial first, so the joins cannot multiply rows.
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	),
	phase_agg AS (
	SELECT nct_id, nct_version, string_agg(phase, ', ' ORDER BY phase) AS phases
	FROM phases GROUP BY nct_id, nct_version
	),
	condition_agg AS (
	SELECT nct_id, nct_version, count(*) AS conditions,
	string_agg(condition, '; ' ORDER BY condition) AS condition_list
	FROM conditions GROUP BY nct_id, nct_version
	),
	intervention_agg AS (
	SELECT nct_id, nct_version, count(*) AS interventions,
	string_agg(DISTINCT intervention_type, ', ') AS intervention_types
	FROM interventions GROUP BY nct_id, nct_version
	),
	location_agg AS (
	SELECT nct_id, nct_version, count(*) AS sites, count(DISTINCT country) AS countries
	FROM locations GROUP BY nct_id, nct_version
	),
	collaborator_agg AS (
	SELECT nct_id, nct_version, count(*) AS collaborators
	FROM collaborators GROUP BY nct_id, nct_version
	)
	SELECT c.nct_id, c.overall_status, c.lead_sponsor_name, ph.phases, c.enrollment_count,
	cd.conditions, i.interventions, i.intervention_types,
	coalesce(l.sites, 0) AS sites, coalesce(l.countries, 0) AS countries,
	coalesce(co.collaborators, 0) AS collaborators
	FROM current AS cur
	INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
	LEFT JOIN phase_agg AS ph ON ph.nct_id = c.nct_id AND ph.nct_version = c.nct_version
	LEFT JOIN condition_agg AS cd ON cd.nct_id = c.nct_id AND cd.nct_version = c.nct_version
	LEFT JOIN intervention_agg AS i ON i.nct_id = c.nct_id AND i.nct_version = c.nct_version
	LEFT JOIN location_agg AS l ON l.nct_id = c.nct_id AND l.nct_version = c.nct_version
	LEFT JOIN collaborator_agg AS co ON co.nct_id = c.nct_id AND co.nct_version = c.nct_version
	ORDER BY c.nct_id;

	-- name: registry_as_of_date
	-- level: advanced
	-- What the registry said about each trial on a past date: the latest version posted on or
	-- before that date, with its conditions, compared with today's status. Change as_of below.
	WITH params AS (
	SELECT DATE '2019-01-01' AS as_of
	),
	as_of_version AS (
	SELECT c.nct_id, c.nct_version
	FROM core AS c
	CROSS JOIN params AS p
	WHERE c.last_update_post_date <= p.as_of
	QUALIFY row_number() OVER (PARTITION BY c.nct_id ORDER BY c.nct_version DESC) = 1
	),
	current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	),
	condition_agg AS (
	SELECT nct_id, nct_version, string_agg(condition, '; ' ORDER BY condition) AS conditions
	FROM conditions GROUP BY nct_id, nct_version
	)
	SELECT then_.nct_id, then_.nct_version AS version_then, then_.overall_status AS status_then,
	then_.primary_completion_date AS primary_completion_then, cd.conditions AS conditions_then,
	now_.nct_version AS version_now, now_.overall_status AS status_now,
	now_.primary_completion_date AS primary_completion_now
	FROM as_of_version AS a
	INNER JOIN core AS then_ ON then_.nct_id = a.nct_id AND then_.nct_version = a.nct_version
	LEFT JOIN condition_agg AS cd ON cd.nct_id = then_.nct_id AND cd.nct_version = then_.nct_version
	INNER JOIN current AS cur ON cur.nct_id = then_.nct_id
	INNER JOIN core AS now_ ON now_.nct_id = cur.nct_id AND now_.nct_version = cur.nct_version
	ORDER BY then_.nct_id;

	-- name: primary_completion_date_slips_by_phase
	-- level: advanced
	-- How often sponsors push out a trial's primary completion date, by phase. Each version is
	-- compared with the one before it (core), dated from version_history (ct_gov_version equals
	-- nct_version) and attributed to the phases of the version that made the change.
	-- A trial with several phases counts once per phase. Days are gross: a date moved earlier
	-- and later again counts each push.
	WITH dated AS (
	SELECT c.nct_id, c.nct_version, c.primary_completion_date,
	lag(c.primary_completion_date) OVER (PARTITION BY c.nct_id ORDER BY c.nct_version)
	AS previous_date,
	h.change_date
	FROM core AS c
	INNER JOIN version_history AS h
	ON h.nct_id = c.nct_id AND h.ct_gov_version = c.nct_version
	),
	slips AS (
	SELECT *, date_diff('day', previous_date, primary_completion_date) AS slip_days
	FROM dated
	WHERE primary_completion_date > previous_date
	)
	SELECT p.phase,
	count(DISTINCT s.nct_id) AS trials_pushed,
	count(*) AS pushes,
	round(avg(s.slip_days)) AS avg_days_per_push,
	max(s.slip_days) AS longest_push_days,
	min(s.change_date) AS first_push_posted,
	max(s.change_date) AS last_push_posted
	FROM slips AS s
	INNER JOIN phases AS p ON p.nct_id = s.nct_id AND p.nct_version = s.nct_version
	GROUP BY p.phase
	ORDER BY pushes DESC;

	-- name: primary_outcome_results
	-- level: advanced
	-- Reported results for primary outcomes at the current version: outcome, group, category
	-- and measured value. Joins five tables through the outcome / class / category ordinals.
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	)
	SELECT om.nct_id, om.title AS outcome, om.unit_of_measure,
	g.title AS group_title, cat.title AS category, ms.value, ms.spread
	FROM current AS cur
	INNER JOIN results_outcome_measures AS om
	ON om.nct_id = cur.nct_id AND om.nct_version = cur.nct_version AND om.type = 'PRIMARY'
	INNER JOIN results_outcome_classes AS cl
	ON cl.nct_id = om.nct_id AND cl.nct_version = om.nct_version AND cl.outcome_ord = om.outcome_ord
	INNER JOIN results_outcome_categories AS cat
	ON cat.nct_id = cl.nct_id AND cat.nct_version = cl.nct_version
	AND cat.outcome_ord = cl.outcome_ord AND cat.class_ord = cl.class_ord
	INNER JOIN results_outcome_measurements AS ms
	ON ms.nct_id = cat.nct_id AND ms.nct_version = cat.nct_version
	AND ms.outcome_ord = cat.outcome_ord AND ms.class_ord = cat.class_ord
	AND ms.category_ord = cat.category_ord
	INNER JOIN results_outcome_groups AS g
	ON g.nct_id = ms.nct_id AND g.nct_version = ms.nct_version
	AND g.outcome_ord = ms.outcome_ord AND g.group_id = ms.group_id
	ORDER BY om.nct_id, om.outcome_ord, cl.class_ord, cat.category_ord, g.group_id;

	-- name: serious_adverse_event_rates_by_group
	-- level: advanced
	-- Serious adverse events at the current version: for each trial, arm group and event term,
	-- how many participants were affected out of those at risk (events, event stats and groups).
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	)
	SELECT e.nct_id, g.title AS group_title, e.organ_system, e.term,
	s.num_affected, s.num_at_risk,
	round(100.0 * s.num_affected / nullif(s.num_at_risk, 0), 2) AS pct_affected
	FROM current AS cur
	INNER JOIN results_adverse_serious_events AS e
	ON e.nct_id = cur.nct_id AND e.nct_version = cur.nct_version
	INNER JOIN results_adverse_serious_event_stats AS s
	ON s.nct_id = e.nct_id AND s.nct_version = e.nct_version AND s.event_ord = e.event_ord
	INNER JOIN results_adverse_event_groups AS g
	ON g.nct_id = s.nct_id AND g.nct_version = s.nct_version AND g.group_id = s.group_id
	WHERE s.num_affected > 0
	ORDER BY pct_affected DESC, e.nct_id, e.term
	LIMIT 50;

	-- name: arms_and_their_interventions
	-- level: advanced
	-- Which intervention each study arm receives, at the current version (arm_groups joined to
	-- the arm/intervention link table and to interventions for the description).
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	)
	SELECT a.nct_id, a.arm_label, a.arm_type, i.intervention_type, i.intervention_name
	FROM current AS cur
	INNER JOIN arm_groups AS a ON a.nct_id = cur.nct_id AND a.nct_version = cur.nct_version
	INNER JOIN intervention_arm_links AS l
	ON l.nct_id = a.nct_id AND l.nct_version = a.nct_version AND l.arm_label = a.arm_label
	INNER JOIN interventions AS i
	ON i.nct_id = l.nct_id AND i.nct_version = l.nct_version
	AND i.intervention_name = l.intervention_name AND i.intervention_type = l.intervention_type
	ORDER BY a.nct_id, a.arm_label, i.intervention_name;

	-- name: sponsor_site_footprint
	-- level: advanced
	-- Each lead sponsor's current trials and where they run: trials, sites and countries
	-- (core joined to locations at the current version).
	WITH current AS (
	SELECT nct_id, max(nct_version) AS nct_version FROM core GROUP BY nct_id
	)
	SELECT c.lead_sponsor_name,
	count(DISTINCT c.nct_id) AS trials,
	count(*) AS sites,
	count(DISTINCT l.country) AS countries
	FROM current AS cur
	INNER JOIN core AS c ON c.nct_id = cur.nct_id AND c.nct_version = cur.nct_version
	INNER JOIN locations AS l ON l.nct_id = c.nct_id AND l.nct_version = c.nct_version
	GROUP BY c.lead_sponsor_name
	ORDER BY sites DESC;