Trial biostatistics toolkit

Safety Review Committee packet: Study CP-101, Cohort 4

A worked example on synthetic data. Study CP-101 does not exist, and this is not a study document.

Prepared

25 October 2026

Synthetic data, worked example. Every number below is computed from the TRIALMON synthetic dataset by the same R functions that drive the TRIALMON console. In a real study the tables a committee relies on come from validated programs on ADaM; this packet was not produced by a validated system, and no one has signed it. The packet presents data. It makes no recommendation: the committee adjudicates and decides, and records its decision in the minutes.

Item	Detail
Study	CP-101 (synthetic): first-in-human, multiple ascending dose, open label for illustration
Design	Once-daily oral dosing, Days 1 to 28 (the DLT window); follow-up to Day 43
Data cut	After the last Cohort 4 participant completed the DLT window (Day 29)
Cohorts dosed	C1 10 mg (n=3), C2 30 mg (n=3), C3 60 mg (n=6), C4 120 mg (n=6); 18 participants dosed. Cohort 5 (200 mg) is planned and not dosed.
Settings (SAP parameters)	MedDRA v29.1; NCI CTCAE v6.0 for lab grades (AE grades as recorded, v6.0); potential Hy’s Law window 30 days; ALP < 2x ULN convention off
Status	Draft for committee review. Rendered from synthetic data; not a validated output.

Questions for the committee

Adjudicate a candidate DLT. CP101-018 (Cohort 4, 120 mg) had QTcF 503 ms on Day 22 (baseline 462 ms). Under the synthetic protocol, QTcF above 500 ms is a candidate DLT; the event has not been adjudicated (Sections 5 and 9).
Review the liver findings. CP101-014 (an adjudicated DLT) meets the potential Hy’s Law screen; CP101-004 and CP101-009 had ALT or AST above 3x ULN that needed confirmation (Section 4).
Decide the next step for dose escalation from the options the protocol allows (Section 9). No option is recommended here.

Cohort summary

Cohort	Dose (mg)	Status	Dosed	DLT-evaluable	Stopped treatment	Any TEAE	Grade 3 or higher TEAE	Serious AE	Related or possibly related TEAE
C1	10	Complete	3	3	0	2 (67%)	0 (0%)	0 (0%)	1 (33%)
C2	30	Complete	3	3	0	3 (100%)	0 (0%)	0 (0%)	3 (100%)
C3	60	Complete	6	6	0	6 (100%)	1 (17%)	0 (0%)	4 (67%)
C4	120	DLT window complete; follow-up ongoing	6	6	1	6 (100%)	2 (33%)	1 (17%)	4 (67%)
C5	200	Planned; not dosed	0	0	0	0	0	0	0

Participants counted once per column. TEAE: treatment-emergent adverse event. AE grades as recorded by the investigator (NCI CTCAE v6.0). DLT-evaluable: completed the 28-day DLT window or had a DLT. Synthetic data.

Dose-limiting toxicities

Cohort	Dose (mg)	Dosed	DLT-evaluable	Confirmed DLTs	Pending adjudication
C1	10	3	3	0	0
C2	30	3	3	0	0
C3	60	6	6	1	0
C4	120	6	6	1	1
C5	200	0	0	0	0

DLTs and candidate DLTs
Participant	Cohort	Dose (mg)	Status	Event
CP101-008	C3	60	Adjudicated DLT	Platelet count decreased, grade 3 (Day 15)
CP101-014	C4	120	Adjudicated DLT	Liver injury: potential Hy’s Law (Day 22)
CP101-018	C4	120	Pending adjudication	QTcF 503 ms on Day 22 (candidate DLT; adjudication pending)

Liver tests

Definition used. A participant is a potential Hy’s Law case when ALT or AST is at least 3x ULN and total bilirubin (TBL) is above 2x ULN on or after the day of that aminotransferase elevation, within 30 days of it (the window is an SAP convention). Every draw with ALT or AST at least 3x ULN opens its own window, so a second elevation weeks after the first is screened as well; the tables report the first draw that qualifies. ALP is shown but not used to exclude a case; the SAP option ALP < 2x ULN is off. This follows FDA, Drug-Induced Liver Injury: Premarketing Clinical Evaluation (July 2009), which asks for TBL above 2x ULN “without initial findings of cholestasis (elevated serum ALP)” and for other causes to be ruled out. The screen cannot rule out other causes; the medical monitor does. Separately, any ALT or AST above 3x ULN calls for repeat ALT, AST, ALP and TBL within 48 to 72 hours to confirm it.

eDISH plot: peak ALT or AST against peak total bilirubin for each participant, log scales, with reference lines at 3x and 2x ULN.

eDISH: each point is one participant’s peak post-baseline ALT or AST against peak TBL. The peaks need not come from the same draw, so the upper-right quadrant is a prompt to check timing, not a classification. Lower right: aminotransferase elevation without TBL above 2x ULN. Upper left: TBL above 2x ULN without an aminotransferase elevation (cholestasis, Gilbert syndrome or other causes).

Participants with ALT or AST above 3x ULN, or in the upper-right quadrant
Participant	Cohort	Peak ALT / AST (x ULN)	ALT or AST >= 3x (day)	ALP then (x ULN)	R ratio (pattern)	Peak TBL (x ULN)	TBL > 2x on or after it (day)	Potential Hy’s Law	Quadrant
CP101-014	C4	5.6 / 4.9	15	1.1	4.0 (mixed)	3.1	22	Yes	Upper right
CP101-004	C2	3.3 / 2.4	22	0.8	4.1 (mixed)	2.4		No	Upper right
CP101-009	C3	4.2 / 3.6	15	1.2	3.5 (mixed)	1.6		No	Lower right

Confirmation of aminotransferase elevations above 3x ULN
Participant	ALT or AST > 3x ULN (day)	Value (x ULN)	First full repeat panel (day)	Days later	Within 48 to 72 h	Repeat ALT or AST (x ULN)	FDA (2009) discontinuation consideration
CP101-004	22	3.3	24	2	Yes	2.6
CP101-009	15	4.2	20	5	No	3.4
CP101-014	15	4.4	17	2	Yes	5.1	ALT or AST > 3x ULN with TBL > 2x ULN in the same draw

ALT or AST >= 3x (day): for a potential Hy’s Law case, the draw that opens the qualifying window; otherwise the first such draw. ALP and the R ratio are read from that draw. R ratio = (ALT/ULN)/(ALP/ULN) (at the first draw with ALT at least 3x ULN if only AST reached 3x in that draw): 5 or more hepatocellular, between 2 and 5 mixed, 2 or less cholestatic. “Within 48 to 72 h” is judged on study days (the data carry days, not clock times), as 3 days or fewer. The FDA (2009) discontinuation considerations that these data can show are ALT or AST above 8x ULN, above 5x ULN for more than 2 weeks, and above 3x ULN with TBL above 2x ULN; INR and symptoms are not in these data.

Liver tests by study day for CP101-014, CP101-004 and CP101-009, as multiples of ULN.

QTcF

Participants by maximum post-baseline QTcF category
Category	C2 (N=3)	C3 (N=6)	C4 (N=6)	Total (N=18)
QTcF > 450 ms	1 (33%)	2 (33%)	1 (17%)	4 (22%)
QTcF > 480 ms	1 (33%)	0 (0%)	1 (17%)	2 (11%)
QTcF > 500 ms	0 (0%)	0 (0%)	1 (17%)	1 (6%)
Change from baseline > 30 ms	1 (33%)	0 (0%)	1 (17%)	2 (11%)
Change from baseline > 60 ms	0 (0%)	0 (0%)	0 (0%)	0 (0%)
Baseline QTcF > 450 ms	1 (33%)	0 (0%)	1 (17%)	2 (11%)

Participants in any category
Participant	Cohort	Baseline (ms)	Maximum (ms)	Day	Maximum change (ms)	Screen
CP101-006	C2	453	486	15	33	AMBER
CP101-011	C3	448	474	22	26	watch
CP101-012	C3	446	468	15	22	watch
CP101-018	C4	462	503	22	41	RED

QTcF: Fridericia-corrected QT, triplicate mean. Categories per ICH E14 (May 2005), the E14 Q&A (R3) (December 2015) and the E14/S7B Q&As (February 2022): maximum post-baseline value above 450, 480 and 500 ms, regardless of sex, and maximum increase from baseline above 30 and 60 ms. Categories are cumulative and a value equal to a threshold is not above it. Baseline: Day 1 predose; post-baseline values start at the Day 1 4 h post-dose ECG. Participants already above 450 ms at baseline: 2 of 18. A categorical summary is a screen; concentration-QTc modeling belongs in a prespecified analysis.

Laboratory shifts

Shift from baseline grade to worst post-baseline grade
Test	N	Worst grade 0	1	2	3
Alanine aminotransferase increased	18	11	4	2	1
Aspartate aminotransferase increased	18	13	3	2	0
Blood bilirubin increased	18	13	1	3	1
Alkaline phosphatase increased	18	15	3	0	0
Creatinine increased	18	16	1	1	0
Anemia	18	17	1	0	0
Neutrophil count decreased	18	17	0	1	0
Thrombocytopenia	18	17	0	0	1

Worst post-baseline grade 3 or higher
Participant	Test	Worst grade
CP101-014	Alanine aminotransferase increased	3
CP101-014	Blood bilirubin increased	3
CP101-008	Thrombocytopenia	3

Lab grades: NCI CTCAE v6.0. Grades come from multiples of the ULN or LLN, or from fold over baseline where the term says so, not from absolute values alone. Creatinine increased: under v6.0, fold over baseline applies only when the baseline is below the LLN; under v5.0 the grade is the worse of the x ULN and fold-over-baseline grades. The v6.0 cut points follow admiral 1.5.0 (atoxgr_criteria_ctcv6) and were not re-checked against the NCI v6.0 workbook here; v5.0 follows the NCI v5.0 workbook (2017-11-27). Baseline grades use reference-range criteria only. Hemoglobin grade 4 is a clinical judgment and is not derived from the value.

Adverse events by SOC and preferred term

Treatment-emergent adverse events, participants
System organ class / Preferred term	C1 (N=3)	C2 (N=3)	C3 (N=6)	C4 (N=6)	Total (N=18)
Any TEAE	2 (67%)	3 (100%)	6 (100%)	6 (100%)	17 (94%)
Investigations	0 (0%)	2 (67%)	2 (33%)	3 (50%)	7 (39%)
Alanine aminotransferase increased	0 (0%)	1 (33%)	1 (17%)	1 (17%)	3 (17%)
Electrocardiogram QT prolonged	0 (0%)	1 (33%)	0 (0%)	1 (17%)	2 (11%)
Aspartate aminotransferase increased	0 (0%)	0 (0%)	0 (0%)	1 (17%)	1 (6%)
Blood bilirubin increased	0 (0%)	1 (33%)	0 (0%)	0 (0%)	1 (6%)
Neutrophil count decreased	0 (0%)	0 (0%)	0 (0%)	1 (17%)	1 (6%)
Platelet count decreased	0 (0%)	0 (0%)	1 (17%)	0 (0%)	1 (6%)
Gastrointestinal disorders	1 (33%)	1 (33%)	2 (33%)	2 (33%)	6 (33%)
Nausea	1 (33%)	0 (0%)	2 (33%)	1 (17%)	4 (22%)
Diarrhoea	0 (0%)	1 (33%)	0 (0%)	1 (17%)	2 (11%)
Nervous system disorders	1 (33%)	1 (33%)	1 (17%)	3 (50%)	6 (33%)
Headache	1 (33%)	1 (33%)	1 (17%)	1 (17%)	4 (22%)
Dizziness	0 (0%)	0 (0%)	0 (0%)	2 (33%)	2 (11%)
General disorders and administration site conditions	0 (0%)	0 (0%)	1 (17%)	1 (17%)	2 (11%)
Fatigue	0 (0%)	0 (0%)	1 (17%)	1 (17%)	2 (11%)
Blood and lymphatic system disorders	0 (0%)	0 (0%)	1 (17%)	0 (0%)	1 (6%)
Anaemia	0 (0%)	0 (0%)	1 (17%)	0 (0%)	1 (6%)
Hepatobiliary disorders	0 (0%)	0 (0%)	0 (0%)	1 (17%)	1 (6%)
Hyperbilirubinaemia	0 (0%)	0 (0%)	0 (0%)	1 (17%)	1 (6%)

Serious adverse events
Participant	Preferred term	SOC	Grade	Relationship	Onset (day)	Outcome
CP101-014	Hyperbilirubinaemia	Hepatobiliary disorders	3	Related	22	ongoing

Adverse events coded with MedDRA v29.1 (primary SOC). AE grades as recorded by the investigator (NCI CTCAE v6.0); they are data and are not re-derived when the lab-grading version (v6.0 here) changes. Participants are counted once per row; N is the number dosed in each cohort. SOCs and preferred terms are sorted by the total number of participants. Preferred terms keep MedDRA’s spelling.

Study-level indicators

Indicator	Value	Level	Range	Detail	Source
Participants with a treatment-emergent SAE (trial-level acceptable range)	5.6 % of dosed	GREEN	AMBER > 10, RED > 20	1 of 18 dosed participants	computed from ADAE / ADLB
Scheduled safety-lab visits with no labs	0.7 % of visits	GREEN	AMBER > 2, RED > 5	1 of 141 scheduled visits	computed from ADAE / ADLB
Screen-failure rate	30.8 %	AMBER	AMBER > 30, RED > 40	8 of 26 screened did not meet eligibility (plan: 25%)	operational feed (synthetic)
Enrollment versus plan	-22 % vs plan	AMBER	AMBER < -15, RED < -30	18 dosed against 23 planned by this date	operational feed (synthetic)
Open queries older than 30 days	14 queries	AMBER	AMBER > 10, RED > 25	backlog rising ahead of the C4 data cut	operational feed (synthetic)
Important protocol deviations per dosed participant	0.11 per participant	GREEN	AMBER > 0.2, RED > 0.4	2 important deviations in 18 dosed participants	operational feed (synthetic)
Median days from visit to data entry	2 days	GREEN	AMBER > 5, RED > 10	entry within plan	operational feed (synthetic)

Pre-specified acceptable ranges, as in ICH E6(R3) section 3.10.1.3 (“quality tolerance limits at the trial level”): exceeding one triggers evaluation, not an automatic stop. The operational values are synthetic. Every level in this packet rolls up worst-of: the participant and cohort tier is RED and the study tier is AMBER, so the overall screen is RED.

Dose escalation: data for the decision

Key findings by dose
Dose (mg)	Cohort	DLT-evaluable	DLTs (confirmed + pending)	Potential Hy’s Law	ALT or AST > 3x ULN	QTcF > 500 ms or change > 60 ms	Grade 3 or higher TEAE
10	C1	3	0 + 0	0	0	0	0 (0%)
30	C2	3	0 + 0	0	1	0	0 (0%)
60	C3	6	1 + 0	0	1	0	1 (17%)
120	C4	6	1 + 1	1	1	1	2 (33%)

Participants. Potential Hy’s Law: window 30 days, ALP option off. Grade 3 or higher TEAE counts AE grades as recorded by the investigator (NCI CTCAE v6.0).

What an illustrative 3+3 rule would say (a reference, not a recommendation)
Cohort	Dose (mg)	Illustrative 3+3 rule reference
C1	10	0 of 3: escalation-eligible under the rule
C2	30	0 of 3: escalation-eligible under the rule
C3	60	1 of 6: escalation-eligible under the rule
C4	120	1 of 6 confirmed, 1 pending adjudication. If not a DLT: 1 of 6, escalation-eligible under the rule. If a DLT: 2 of 6, under the rule, the dose exceeds the MTD

Reading the rule reference. Under a 3+3 rule, 0 of 3 or 1 of 6 participants with a DLT is escalation-eligible, 1 of 3 means the cohort expands to 6, and 2 or more mean the dose exceeds the maximum tolerated dose; a dose can be called the MTD only with 6 participants treated at it. For Cohort 4 the reference depends on the committee’s adjudication of CP101-018. 3+3 appears here only as an illustration: model-assisted designs such as BOIN (FDA fit-for-purpose determination, December 2021, for the local BOIN design under a non-informative prior) are the current alternative, and this example implements neither as a decision tool. A rule reference never replaces the committee’s review of the whole safety picture, including the potential Hy’s Law case at 120 mg.

Options the synthetic protocol allows the committee to consider, in protocol order: escalate to 200 mg as planned; escalate to an intermediate dose; enroll further participants at 120 mg; stop escalation and select a lower dose; pause dosing pending further evaluation. They are listed, not ranked.

Committee record (completed by the committee, not by this packet)

DLT adjudication for CP101-018: ______________________________

Decision on dose escalation: ______________________________

Rationale: ______________________________

Chair: ______________________ Date: ______________________

Methods and conventions

Source. The TRIALMON synthetic dataset (adsl, adlb, adeg, adae, cohorts, study_kri), written by R/gen_data.R from one seed. Every table is computed at render time by R/logic.R, the code behind the TRIALMON console and its tests, and the findings the text cites (participants, values, days and counts) are computed from the same objects. The study design and the protocol’s options are fixed text from the synthetic protocol.
Populations. Safety population: all dosed participants. Baseline: the Day 1 predose value. Post-baseline: labs after Day 1; ECGs from the Day 1 4 h post-dose time point on; both include unscheduled draws.
Counting. Participants are counted once per row of a table. Percentages use the dosed participants in the column.
Dictionaries and grading. MedDRA v29.1; NCI CTCAE v6.0 for laboratory grades, derived from the values; AE grades as recorded by the investigator under NCI CTCAE v6.0 (a property of the data: changing the lab-grading version does not re-grade an AE). All three are parameters of this document (meddra_version, ctcae_version, ae_ctcae_version), as are the Hy’s Law window (hys_window_days) and the ALP convention (alp_convention).
Guidance cited. FDA, Drug-Induced Liver Injury: Premarketing Clinical Evaluation (July 2009); ICH E14 (May 2005), E14 Q&A (R3) (10 Dec 2015) and E14/S7B Q&As (Step 4 21 Feb 2022); NCI CTCAE v6.0; ICH E6(R3) (Step 4 6 Jan 2025; EU effective 23 Jul 2025; FDA final guidance Sep 2025); Annex 2 Step 4 3 Jun 2026 (CHMP 25 Jun 2026, EU effective 15 Jan 2027; FDA: draft guidance Dec 2024, not final).
What was executed. Rendered with Quarto and R version 4.6.0 (2026-04-24) on synthetic data; the rules are covered by testthat tests (rule boundaries, the shipped study and the Shiny server). No validated system or independent programmer produced or checked these numbers.
What a real packet adds. Tables from validated programs on ADaM, a named statistician’s sign-off, and, for a DSMB with access to unblinded comparisons, open and closed sections with an independent reporting statistician. This open-label example has no closed section.

--- title: "Safety Review Committee packet: Study CP-101, Cohort 4" subtitle: "A worked example on synthetic data. Study CP-101 does not exist, and this is not a study document." date: "2026-10-25" date-format: "D MMMM YYYY" language: title-block-published: "Prepared" params: meddra_version: "29.1" ctcae_version: "6.0" ae_ctcae_version: "6.0" hys_window_days: 30 alp_convention: false format: html: theme: - darkly - ../../_quarto-mc.scss - _theme-mc.scss include-before-body: ../../_mc-topbar.html toc: true toc-location: left toc-title: On this page embed-resources: true code-tools: true code-fold: true execute: echo: false warning: false message: false knitr: opts_chunk: dev.args: bg: transparent --- ```{r setup} library(ggplot2) source("../trialmon_shiny/R/logic.R") source("_report-helpers.R") tm <- trialmon_load("../trialmon_shiny/data") S <- trialmon_settings(params$ctcae_version, params$hys_window_days, params$alp_convention, params$meddra_version, params$ae_ctcae_version) h <- hys_screen(tm$adlb, S$hys_window, S$alp_convention) q <- qtc_screen(tm$adeg) ds <- dlt_state(tm) cs <- cohort_summary(tm) roll <- trialmon_rollup(tm, S) dosed <- ds[ds$N > 0, ] ctcae_note <- sprintf("Lab grades: NCI CTCAE v%s.", S$ctcae) ae_grade_note <- sprintf("AE grades as recorded by the investigator (NCI CTCAE v%s)", S$ae_ctcae) meddra_note <- sprintf("Adverse events coded with MedDRA v%s (primary SOC). %s; they are data and are not re-derived when the lab-grading version (v%s here) changes.", S$meddra, ae_grade_note, S$ctcae) alp_note <- if (S$alp_convention) "ALP < 2x ULN in the draw with that aminotransferase elevation is also required (an SAP convention)." else "ALP is shown but not used to exclude a case; the SAP option ALP < 2x ULN is off." E6R3 <- paste("ICH E6(R3) (Step 4 6 Jan 2025; EU effective 23 Jul 2025; FDA final guidance Sep 2025);", "Annex 2 Step 4 3 Jun 2026 (CHMP 25 Jun 2026, EU effective 15 Jan 2027; FDA: draft guidance Dec 2024, not final)") # the findings the text cites, computed from the same objects as the tables and_list <- function(x) if (length(x) <= 1) paste(x, collapse = "") else paste(paste(x[-length(x)], collapse = ", "), "and", x[length(x)]) coh_name <- function(cc) sub("^C", "Cohort ", cc) who <- function(id) { s <- tm$adsl[match(id, tm$adsl$USUBJID), ]; sprintf("%s (%s, %g mg)", id, coh_name(s$COHORT), s$DOSEMG) } is_dlt <- function(id) tm$adsl$DLTFL[match(id, tm$adsl$USUBJID)] == "Y" pend <- tm$adsl[tm$adsl$DLTADJ == "Pending", ] hys_ids <- h$USUBJID[h$hys] conf_ids <- h$USUBJID[!h$hys & !is.na(h$at3x_day)] hys_doses <- unique(tm$adsl$DOSEMG[tm$adsl$USUBJID %in% hys_ids]) pend_txt <- vapply(pend$USUBJID, function(id) { x <- q[q$USUBJID == id, ] if (nrow(x) && x$gt500) sprintf("%s had QTcF %d ms on Day %d (baseline %d ms). Under the synthetic protocol, QTcF above 500 ms is a candidate DLT; the event has not been adjudicated", who(id), x$max_qtcf, x$max_day, x$base) else sprintf("%s has a candidate DLT that has not been adjudicated: %s", who(id), pend$DLTDESC[pend$USUBJID == id]) }, "") tag <- function(ids) ifelse(is_dlt(ids), paste(ids, "(an adjudicated DLT)"), ids) liver_txt <- c( if (length(hys_ids)) sprintf("%s %s the potential Hy's Law screen", and_list(tag(hys_ids)), if (length(hys_ids) == 1) "meets" else "meet"), if (length(conf_ids)) sprintf("%s had ALT or AST above 3x ULN that needed confirmation", and_list(tag(conf_ids)))) if (!length(liver_txt)) liver_txt <- "No participant meets the potential Hy's Law screen or had ALT or AST above 3x ULN" next_dose <- tm$cohorts[grepl("^Planned", tm$cohorts$STATUS), ][1, ] top <- dosed[which.max(dosed$DOSEMG), ] pend_coh <- ds[ds$PENDING > 0 & ds$STATE != ds$STATE_IF_DLT, ] ``` ::: {.callout-important appearance="simple"} **Synthetic data, worked example.** Every number below is computed from the TRIALMON synthetic dataset by the same R functions that drive the [TRIALMON console](../index.html). In a real study the tables a committee relies on come from validated programs on ADaM; this packet was not produced by a validated system, and no one has signed it. The packet presents data. It makes no recommendation: the committee adjudicates and decides, and records its decision in the minutes. ::: ```{r front, results='asis'} cat(html_table(data.frame( Item = c("Study", "Design", "Data cut", "Cohorts dosed", "Settings (SAP parameters)", "Status"), Detail = c("CP-101 (synthetic): first-in-human, multiple ascending dose, open label for illustration", "Once-daily oral dosing, Days 1 to 28 (the DLT window); follow-up to Day 43", sprintf("After the last %s participant completed the DLT window (Day %d)", coh_name(top$COHORT), tm$cohorts$DLTWIN[tm$cohorts$COHORT == top$COHORT] + 1), sprintf("%s; %d participants dosed.%s", paste(sprintf("%s %g mg (n=%d)", dosed$COHORT, dosed$DOSEMG, dosed$N), collapse = ", "), sum(dosed$N), if (is.na(next_dose$COHORT)) "" else sprintf(" %s (%g mg) is planned and not dosed.", coh_name(next_dose$COHORT), next_dose$DOSEMG)), sprintf("MedDRA v%s; NCI CTCAE v%s for lab grades (AE grades as recorded, v%s); potential Hy's Law window %g days; ALP < 2x ULN convention %s", S$meddra, S$ctcae, S$ae_ctcae, S$hys_window, if (S$alp_convention) "on" else "off"), "Draft for committee review. Rendered from synthetic data; not a validated output.")))) ``` # Questions for the committee ```{r questions, results='asis'} qs <- c( if (nrow(pend)) sprintf("**Adjudicate a candidate DLT.** %s (Sections [5](#qtcf) and [9](#dose-escalation-data-for-the-decision)).", paste(pend_txt, collapse = ". ")), sprintf("**Review the liver findings.** %s (Section [4](#liver-tests)).", paste(liver_txt, collapse = "; ")), "**Decide the next step for dose escalation** from the options the protocol allows (Section [9](#dose-escalation-data-for-the-decision)). No option is recommended here.") cat(paste0(seq_along(qs), ". ", qs, collapse = "\n"), "\n") ``` # Cohort summary ```{r cohort-summary, results='asis'} d <- cs cat(html_table(data.frame( Cohort = d$COHORT, `Dose (mg)` = d$DOSEMG, Status = d$STATUS, Dosed = d$N, `DLT-evaluable` = d$DLT_EVAL, `Stopped treatment` = d$DISC, `Any TEAE` = npct(d$ANY_TEAE, d$N), `Grade 3 or higher TEAE` = npct(d$G3, d$N), `Serious AE` = npct(d$SAE, d$N), `Related or possibly related TEAE` = npct(d$REL, d$N), check.names = FALSE), num = c("Dose (mg)", "Dosed", "DLT-evaluable", "Stopped treatment", "Any TEAE", "Grade 3 or higher TEAE", "Serious AE", "Related or possibly related TEAE"))) ``` [Participants counted once per column. TEAE: treatment-emergent adverse event. `r ae_grade_note`. DLT-evaluable: completed the 28-day DLT window or had a DLT. Synthetic data.]{.tm-foot} # Dose-limiting toxicities {#dlts} ```{r dlt, results='asis'} cat(html_table(data.frame( Cohort = ds$COHORT, `Dose (mg)` = ds$DOSEMG, Dosed = ds$N, `DLT-evaluable` = ds$EVALUABLE, `Confirmed DLTs` = ds$DLT, `Pending adjudication` = ds$PENDING, check.names = FALSE), num = c("Dose (mg)", "Dosed", "DLT-evaluable", "Confirmed DLTs", "Pending adjudication"))) ev <- tm$adsl[tm$adsl$DLTFL == "Y" | tm$adsl$DLTADJ == "Pending", ] cat(html_table(data.frame( Participant = ev$USUBJID, Cohort = ev$COHORT, `Dose (mg)` = ev$DOSEMG, Status = ifelse(ev$DLTADJ == "Pending", "Pending adjudication", "Adjudicated DLT"), Event = ev$DLTDESC, check.names = FALSE), num = "Dose (mg)", caption = "DLTs and candidate DLTs")) ``` # Liver tests {#liver-tests} **Definition used.** A participant is a *potential Hy's Law* case when ALT or AST is at least 3x ULN and total bilirubin (TBL) is above 2x ULN **on or after** the day of that aminotransferase elevation, within `r S$hys_window` days of it (the window is an SAP convention). Every draw with ALT or AST at least 3x ULN opens its own window, so a second elevation weeks after the first is screened as well; the tables report the first draw that qualifies. `r alp_note` This follows FDA, *Drug-Induced Liver Injury: Premarketing Clinical Evaluation* (July 2009), which asks for TBL above 2x ULN "without initial findings of cholestasis (elevated serum ALP)" and for other causes to be ruled out. The screen cannot rule out other causes; the medical monitor does. Separately, any ALT or AST **above** 3x ULN calls for repeat ALT, AST, ALP and TBL within 48 to 72 hours to confirm it. ```{r edish, fig.width=7.6, fig.height=4.9, fig.alt="eDISH plot: peak ALT or AST against peak total bilirubin for each participant, log scales, with reference lines at 3x and 2x ULN."} both(function(look) edish_plot(h, look)) ``` [eDISH: each point is one participant's peak post-baseline ALT or AST against peak TBL. The peaks need not come from the same draw, so the upper-right quadrant is a prompt to check timing, not a classification. Lower right: aminotransferase elevation without TBL above 2x ULN. Upper left: TBL above 2x ULN without an aminotransferase elevation (cholestasis, Gilbert syndrome or other causes).]{.tm-foot} ```{r liver-tables, results='asis'} x <- h[h$hys | h$hys_quadrant_only | !is.na(h$at3x_day), ] x <- x[order(-x$hys, x$USUBJID), ] coh <- tm$adsl$COHORT[match(x$USUBJID, tm$adsl$USUBJID)] cat(html_table(data.frame( Participant = x$USUBJID, Cohort = coh, `Peak ALT / AST (x ULN)` = sprintf("%.1f / %.1f", x$peak_alt, x$peak_ast), `ALT or AST >= 3x (day)` = x$at3_day, `ALP then (x ULN)` = sprintf("%.1f", x$alp_at3), `R ratio (pattern)` = ifelse(is.na(x$r_at3), "", sprintf("%.1f (%s)", x$r_at3, x$pattern)), `Peak TBL (x ULN)` = sprintf("%.1f", x$peak_tbl), `TBL > 2x on or after it (day)` = x$tbl_day, `Potential Hy's Law` = ifelse(x$hys, "Yes", "No"), Quadrant = sub(":.*", "", x$quadrant), check.names = FALSE), num = c("ALT or AST >= 3x (day)", "ALP then (x ULN)", "Peak TBL (x ULN)", "TBL > 2x on or after it (day)"), caption = "Participants with ALT or AST above 3x ULN, or in the upper-right quadrant")) y <- h[!is.na(h$at3x_day), ] cat(html_table(data.frame( Participant = y$USUBJID, `ALT or AST > 3x ULN (day)` = y$at3x_day, `Value (x ULN)` = sprintf("%.1f", y$at3x_value), `First full repeat panel (day)` = y$repeat_day, `Days later` = y$repeat_gap, `Within 48 to 72 h` = ifelse(is.na(y$repeat_gap), "no repeat", ifelse(y$repeat_gap <= 3, "Yes", "No")), `Repeat ALT or AST (x ULN)` = sprintf("%.1f", y$repeat_at), `FDA (2009) discontinuation consideration` = y$stop_reason, check.names = FALSE), num = c("ALT or AST > 3x ULN (day)", "Value (x ULN)", "First full repeat panel (day)", "Days later", "Repeat ALT or AST (x ULN)"), caption = "Confirmation of aminotransferase elevations above 3x ULN")) ``` [ALT or AST >= 3x (day): for a potential Hy's Law case, the draw that opens the qualifying window; otherwise the first such draw. ALP and the R ratio are read from that draw. R ratio = (ALT/ULN)/(ALP/ULN) (at the first draw with ALT at least 3x ULN if only AST reached 3x in that draw): 5 or more hepatocellular, between 2 and 5 mixed, 2 or less cholestatic. "Within 48 to 72 h" is judged on study days (the data carry days, not clock times), as 3 days or fewer. The FDA (2009) discontinuation considerations that these data can show are ALT or AST above 8x ULN, above 5x ULN for more than 2 weeks, and above 3x ULN with TBL above 2x ULN; INR and symptoms are not in these data.]{.tm-foot} ```{r traj-ids} traj_ids <- head(x$USUBJID[x$hys | !is.na(x$at3x_day)], 4) ``` ```{r traj, eval=length(traj_ids) > 0, fig.width=7.6, fig.height=3.4, fig.alt=paste0("Liver tests by study day for ", and_list(traj_ids), ", as multiples of ULN.")} tr <- do.call(rbind, lapply(traj_ids, function(i) cbind(liver_trajectory(tm, i), USUBJID = i))) tr$USUBJID <- factor(tr$USUBJID, levels = traj_ids) both(function(look) trajectory_plot(tr, look, NULL) + facet_wrap(~USUBJID, nrow = 1)) ``` # QTcF {#qtcf} ```{r qtc, results='asis'} q$COHORT <- tm$adsl$COHORT[match(q$USUBJID, tm$adsl$USUBJID)] cats <- c(`QTcF > 450 ms` = "gt450", `QTcF > 480 ms` = "gt480", `QTcF > 500 ms` = "gt500", `Change from baseline > 30 ms` = "chg30", `Change from baseline > 60 ms` = "chg60", `Baseline QTcF > 450 ms` = "base450") grp <- c(split(q, q$COHORT), list(Total = q)) out <- data.frame(Category = names(cats), check.names = FALSE) for (g in names(grp)) out[[sprintf("%s (N=%d)", g, nrow(grp[[g]]))]] <- vapply(cats, function(cc) npct(sum(grp[[g]][[cc]]), nrow(grp[[g]])), "") cat(html_table(out, num = names(out)[-1], caption = "Participants by maximum post-baseline QTcF category")) z <- q[q$gt450 | q$chg30, ] cat(html_table(data.frame(Participant = z$USUBJID, Cohort = z$COHORT, `Baseline (ms)` = z$base, `Maximum (ms)` = z$max_qtcf, Day = z$max_day, `Maximum change (ms)` = z$max_chg, Screen = sev_html(z$sev), check.names = FALSE), num = c("Baseline (ms)", "Maximum (ms)", "Day", "Maximum change (ms)"), raw = "Screen", caption = "Participants in any category")) ``` [QTcF: Fridericia-corrected QT, triplicate mean. Categories per ICH E14 (May 2005), the E14 Q&A (R3) (December 2015) and the E14/S7B Q&As (February 2022): maximum post-baseline value above 450, 480 and 500 ms, regardless of sex, and maximum increase from baseline above 30 and 60 ms. Categories are cumulative and a value equal to a threshold is not above it. Baseline: Day 1 predose; post-baseline values start at the Day 1 4 h post-dose ECG. Participants already above 450 ms at baseline: `r sum(q$base450)` of `r nrow(q)`. A categorical summary is a screen; concentration-QTc modeling belongs in a prespecified analysis.]{.tm-foot} # Laboratory shifts ```{r shift, results='asis'} s <- lab_shift(tm$adlb, S$ctcae) cat(html_table(data.frame(Test = s$TERM, `Baseline grade` = s$BGRADE, N = s$N, `Worst grade 0` = s$G0, `1` = s$G1, `2` = s$G2, `3` = s$G3, `4` = s$G4, check.names = FALSE), num = c("Baseline grade", "N", "Worst grade 0", "1", "2", "3", "4"), caption = "Shift from baseline grade to worst post-baseline grade")) g <- lab_grades(tm$adlb, S$ctcae); g <- g[!is.na(g$WGRADE) & g$WGRADE >= 3, ] cat(html_table(data.frame(Participant = g$USUBJID, Test = g$TERM, `Baseline grade` = g$BGRADE, `Worst grade` = g$WGRADE, check.names = FALSE), num = c("Baseline grade", "Worst grade"), caption = "Worst post-baseline grade 3 or higher")) ``` [`r ctcae_note` Grades come from multiples of the ULN or LLN, or from fold over baseline where the term says so, not from absolute values alone. Creatinine increased: under v6.0, fold over baseline applies only when the baseline is below the LLN; under v5.0 the grade is the worse of the x ULN and fold-over-baseline grades. The v6.0 cut points follow admiral 1.5.0 (`atoxgr_criteria_ctcv6`) and were not re-checked against the NCI v6.0 workbook here; v5.0 follows the NCI v5.0 workbook (2017-11-27). Baseline grades use reference-range criteria only. Hemoglobin grade 4 is a clinical judgment and is not derived from the value.]{.tm-foot} # Adverse events by SOC and preferred term ```{r ae, results='asis'} a <- ae_soc_pt(tm); N <- attr(a, "N") lab <- ifelse(a$PT == "", a$SOC, paste0("    ", esc(a$PT))) out <- data.frame(`System organ class / Preferred term` = lab, check.names = FALSE) for (gname in names(N)) out[[sprintf("%s (N=%d)", gname, N[[gname]])]] <- npct(a[[gname]], N[[gname]]) cat(html_table(out, num = names(out)[-1], raw = "System organ class / Preferred term", row_class = ifelse(a$PT == "", "soc", ""), caption = "Treatment-emergent adverse events, participants")) sae <- tm$adae[tm$adae$AESER == "Y", ] cat(html_table(data.frame(Participant = sae$USUBJID, `Preferred term` = sae$AEDECOD, SOC = sae$AEBODSYS, Grade = sae$AETOXGR, Relationship = sae$AREL, `Onset (day)` = sae$ASTDY, Outcome = ifelse(is.na(sae$AENDY), "ongoing", paste("resolved Day", sae$AENDY)), check.names = FALSE), num = c("Grade", "Onset (day)"), caption = "Serious adverse events")) ``` [`r meddra_note` Participants are counted once per row; N is the number dosed in each cohort. SOCs and preferred terms are sorted by the total number of participants. Preferred terms keep MedDRA's spelling.]{.tm-foot} # Study-level indicators ```{r study, results='asis'} st <- study_signals(tm) cat(html_table(data.frame(Indicator = st$INDICATOR, Value = paste(st$VALUE, st$UNIT), Level = sev_html(st$LEVEL), Range = st$RULE, Detail = st$DETAIL, Source = st$SOURCE, check.names = FALSE), raw = "Level")) ``` [Pre-specified acceptable ranges, as in ICH E6(R3) section 3.10.1.3 ("quality tolerance limits at the trial level"): exceeding one triggers evaluation, not an automatic stop. The operational values are synthetic. Every level in this packet rolls up worst-of: the participant and cohort tier is `r roll$participant` and the study tier is `r roll$study`, so the overall screen is `r roll$overall`.]{.tm-foot} # Dose escalation: data for the decision ```{r esc, results='asis'} by_coh <- function(col) vapply(dosed$COHORT, function(cc) { ids <- tm$adsl$USUBJID[tm$adsl$COHORT == cc]; col(ids) }, "") cat(html_table(data.frame( `Dose (mg)` = dosed$DOSEMG, Cohort = dosed$COHORT, `DLT-evaluable` = dosed$EVALUABLE, `DLTs (confirmed + pending)` = sprintf("%d + %d", dosed$DLT, dosed$PENDING), `Potential Hy's Law` = by_coh(function(ids) as.character(sum(h$hys[h$USUBJID %in% ids]))), `ALT or AST > 3x ULN` = by_coh(function(ids) as.character(sum(!is.na(h$at3x_day[h$USUBJID %in% ids])))), `QTcF > 500 ms or change > 60 ms` = by_coh(function(ids) as.character(sum(q$gt500[q$USUBJID %in% ids] | q$chg60[q$USUBJID %in% ids]))), `Grade 3 or higher TEAE` = npct(cs$G3[cs$N > 0], cs$N[cs$N > 0]), check.names = FALSE), num = c("Dose (mg)", "DLT-evaluable", "DLTs (confirmed + pending)", "Potential Hy's Law", "ALT or AST > 3x ULN", "QTcF > 500 ms or change > 60 ms", "Grade 3 or higher TEAE"), caption = "Key findings by dose")) cat(sprintf('<p class="tm-foot">Participants. Potential Hy\'s Law: window %g days, ALP option %s. Grade 3 or higher TEAE counts AE grades as recorded by the investigator (NCI CTCAE v%s).</p>\n', S$hys_window, if (S$alp_convention) "on" else "off", S$ae_ctcae)) cat(html_table(data.frame(Cohort = dosed$COHORT, `Dose (mg)` = dosed$DOSEMG, `Illustrative 3+3 rule reference` = dosed$text, check.names = FALSE), num = "Dose (mg)", caption = "What an illustrative 3+3 rule would say (a reference, not a recommendation)")) ``` **Reading the rule reference.** Under a 3+3 rule, 0 of 3 or 1 of 6 participants with a DLT is escalation-eligible, 1 of 3 means the cohort expands to 6, and 2 or more mean the dose exceeds the maximum tolerated dose; a dose can be called the MTD only with 6 participants treated at it. `r if (nrow(pend_coh)) sprintf("For %s the reference depends on the committee's adjudication of %s.", and_list(coh_name(pend_coh$COHORT)), and_list(pend$USUBJID[pend$COHORT %in% pend_coh$COHORT])) else ""` 3+3 appears here only as an illustration: model-assisted designs such as BOIN (FDA fit-for-purpose determination, December 2021, for the local BOIN design under a non-informative prior) are the current alternative, and this example implements neither as a decision tool. A rule reference never replaces the committee's review of the whole safety picture, including `r if (length(hys_ids)) sprintf("the potential Hy's Law %s at %s", if (length(hys_ids) == 1) "case" else "cases", and_list(paste(hys_doses, "mg"))) else "the liver findings in Section 4"`. **Options the synthetic protocol allows the committee to consider,** in protocol order: escalate to `r if (is.na(next_dose$COHORT)) "the next planned dose" else paste(next_dose$DOSEMG, "mg")` as planned; escalate to an intermediate dose; enroll further participants at `r top$DOSEMG` mg; stop escalation and select a lower dose; pause dosing pending further evaluation. They are listed, not ranked. ::: {.tm-decision} **Committee record** (completed by the committee, not by this packet) `r if (nrow(pend)) paste0("DLT adjudication for ", and_list(pend$USUBJID), ":") else "DLT adjudications:"` ______________________________ Decision on dose escalation: ______________________________ Rationale: ______________________________ Chair: ______________________ Date: ______________________ ::: # Methods and conventions - **Source.** The TRIALMON synthetic dataset (`adsl`, `adlb`, `adeg`, `adae`, `cohorts`, `study_kri`), written by `R/gen_data.R` from one seed. Every table is computed at render time by `R/logic.R`, the code behind the TRIALMON console and its tests, and the findings the text cites (participants, values, days and counts) are computed from the same objects. The study design and the protocol's options are fixed text from the synthetic protocol. - **Populations.** Safety population: all dosed participants. Baseline: the Day 1 predose value. Post-baseline: labs after Day 1; ECGs from the Day 1 4 h post-dose time point on; both include unscheduled draws. - **Counting.** Participants are counted once per row of a table. Percentages use the dosed participants in the column. - **Dictionaries and grading.** MedDRA v`r S$meddra`; NCI CTCAE v`r S$ctcae` for laboratory grades, derived from the values; AE grades as recorded by the investigator under NCI CTCAE v`r S$ae_ctcae` (a property of the data: changing the lab-grading version does not re-grade an AE). All three are parameters of this document (`meddra_version`, `ctcae_version`, `ae_ctcae_version`), as are the Hy's Law window (`hys_window_days`) and the ALP convention (`alp_convention`). - **Guidance cited.** FDA, *Drug-Induced Liver Injury: Premarketing Clinical Evaluation* (July 2009); ICH E14 (May 2005), E14 Q&A (R3) (10 Dec 2015) and E14/S7B Q&As (Step 4 21 Feb 2022); NCI CTCAE v`r S$ctcae`; `r E6R3`. - **What was executed.** Rendered with Quarto and `r R.version.string` on synthetic data; the rules are covered by testthat tests (rule boundaries, the shipped study and the Shiny server). No validated system or independent programmer produced or checked these numbers. - **What a real packet adds.** Tables from validated programs on ADaM, a named statistician's sign-off, and, for a DSMB with access to unblinded comparisons, open and closed sections with an independent reporting statistician. This open-label example has no closed section.