Review workflow — what to do and when
1
Review new SAEs and deaths (0–24h)
Any new SAE since last review: verify narrative completeness, investigator causality, and regulatory reporting obligation. For every on-study death: construct the full timeline — last dose → symptom onset → hospitalization → death. Attribution must be finalized before closing the narrative. Fatal/life-threatening SUSARs: 7-day reporting clock (ICH E2A).
2
Check critical laboratory alerts
ALT/AST >3× ULN + bilirubin >2× ULN → possible Hy's Law — individual narrative review same day. Troponin elevation in any IO patient → presume immune-mediated myocarditis until ruled out. ANC <500/mm³ + fever ≥38.3°C → febrile neutropenia, assess SAE criteria. QTc >500ms or delta >60ms from baseline → urgent cardiac evaluation. These cannot wait for the next visit.
3
Review patients scheduled for next dose (rolling 7-day window)
For every patient with a dose in the next 7 days: active grade ≥2 AEs? Labs out of range per protocol criteria? Cardiac or tumor assessments pending? New concomitant with a clinically relevant interaction? This review must be complete before the patient arrives at the site — not after dosing.
4
Review new tumor assessments
Verify assigned response is consistent with target lesion measurements. In IO: identify any iUPD and confirm a confirmatory scan is scheduled within 4–8 weeks. Flag any patient withdrawn for "progression" without iCPD confirmation — protocol deviation and efficacy data loss.
5
Aggregate trend review (weekly / monthly)
Is any AE increasing in incidence over cycles? Does discontinuation rate due to toxicity exceed the protocol-expected rate? Is there an imbalance in SAEs or grade 3–4 events between arms? Does any site show a markedly different AE profile — systematic under-reporting signal?
6
Document and communicate findings
Signals → written communication to sponsor with clinical rationale and timeline. Protocol deviations → site communication with corrective action. Potential SUSARs → regulatory escalation per ICH E2A timelines. All findings: date, data source, clinical reasoning, recommended action.
Core principle: If a finding makes you think "I'll revisit this next week" — that is the signal to act now. In oncology, signals that wait become events.
Multiple attribution — the central challengeCore principle
Oncology patients are never on a single drug. Before attributing any AE to the investigational product, simultaneously assess: What exactly happened? Does any concomitant medication fully or partially explain it? Is the temporal relationship with the study drug biologically plausible? Answering these three questions sequentially rather than simultaneously is the most common data review error in oncology.
Critical drug–concomitant interactionsAlways screen
CYP3A4 inhibitorsAzoles (fluconazole, voriconazole), macrolides, protease inhibitors → elevate TKI plasma levels (imatinib, ibrutinib, palbociclib, osimertinib) → functional overdose. Was an antifungal started 2–4 weeks before the AE?
CYP3A4 inducersPhenytoin, carbamazepine, rifampicin, St. John's Wort → reduce study drug levels → sub-exposure. Apparent efficacy failure may be pharmacokinetic, not pharmacodynamic.
QTc prolongersOndansetron IV, fluoroquinolones, azithromycin, azoles, haloperidol, citalopram → additive/synergistic risk. Review full CM list before attributing QTc prolongation to study drug.
Hepatotoxic agentsAcetaminophen >2g/day, amoxicillin-clavulanate, high-dose statins → simulate or potentiate DILI. Directly impacts Hy's Law Criterion 3 (exclusion of other causes).
Nephrotoxic agentsNSAIDs + cisplatin, or + anti-VEGF, or + checkpoint inhibitors → synergistic acute kidney injury. Ask about OTC NSAID use — rarely reported spontaneously.
Corticosteroids in IOUsed to manage irAEs but immunosuppressive — may reduce checkpoint inhibitor efficacy. Track: dose, duration, indication. Patients on prolonged steroids may show worse PFS — confounding factor for efficacy analysis.
Herbal supplementsSilymarin, kava, Chinese herbal products → hepatotoxic potential, rarely reported without active questioning. Critical for Hy's Law Criterion 3 exclusion.
Concomitant as AE surrogate: Antiemetic without documented nausea. Opioid without documented pain. Antidiarrheal without documented diarrhea. Colony-stimulating factor without documented neutropenia. These are footprints of unreported AE entries.
Under-reporting signals — what should be there but isn'tScreen systematically
- Absence of class-expected AEs (no rash with EGFR inhibitor; no hypertension with anti-VEGF; no diarrhea with IO) → biologically implausible, systematic under-reporting
- AE resolves in <48h when biologically implausible (grade 3 hepatitis resolving in 2 days)
- Multiple AEs with identical start date → retrospective data entry, not prospective capture
- Dose reduction with no grade ≥2 AE documented to justify it
- AE with no resolution date in patients who discontinued — resolution assumed, not verified
- Causality "not related" assigned for AE with positive rechallenge (AE returned on re-exposure)
- MedDRA term inconsistent with clinical description — affects aggregate safety tables
ANC — neutrophils (most critical hematologic parameter)Critical
ANC <500/mm³CTCAE Grade 4. If fever ≥38.3°C → febrile neutropenia = SAE. Risk of fatal sepsis. Verify site acted per protocol: empiric antibiotics initiated, hospitalization criteria applied.
ANC 500–1000/mm³Grade 3. Dose delay usually required per protocol. Verify no concurrent active infection. Was G-CSF administered per protocol or as rescue?
Prolonged neutropenia>21 days without ANC recovery → cumulative bone marrow toxicity signal. Alarm even at grade 3. Requires narrative review and consideration of dose modification.
Expected nadirDays 10–14 post-chemotherapy with spontaneous recovery. If timing or recovery curve deviates from drug-expected profile → investigate. Concurrent infection or bone marrow tumor involvement are confounders.
Key question: Does the nadir depth and recovery curve match what is expected for this specific drug and schedule? Deviation requires biological explanation.
Platelets, hemoglobin and lymphocytesContextual
Platelets <25,000/mm³Grade 4. Was there bleeding? Transfusion required? Count alone without clinical context is insufficient for interpretation.
Hemoglobin — relative dropMeasure drop relative to patient's true baseline. A drop from 14 to 9 g/dL is clinically very different from a stable 9 g/dL in a patient whose baseline was 8 g/dL.
Lymphopenia in IOMay indicate tumor redistribution (response signal) or systemic toxicity. Must correlate with clinical symptoms, imaging, and irAE assessment — cannot be interpreted from the value alone.
Hepatic function — read in three dimensionsHy's Law risk
ALTMost specific marker of hepatocellular injury and DILI. Nearly exclusively hepatic. Track every cycle without exception. Rising trend matters more than any single value.
ASTLess specific — also in cardiac and skeletal muscle. Elevated AST with normal ALT → suspect muscle damage, not liver. AST/ALT >2:1 suggests alcohol; in DILI and immune hepatitis, ALT typically exceeds AST.
Total + direct bilirubinAlways fractionate. Elevated direct → excretory/obstructive problem. Elevated indirect → hemolysis or Gilbert syndrome. Both elevated with transaminases → Hy's Law territory.
ALP + GGTCholestasis markers. R-value = (ALT/ULN) ÷ (ALP/ULN): R>5 hepatocellular (higher DILI risk), R<2 cholestatic, 2–5 mixed. Document R-value in every hepatic AE narrative.
Albumin + PT/INRSynthetic function markers. Falling albumin + rising INR with elevated transaminases = hepatic synthetic failure. This is an emergency, not an observation.
Renal function — catch the trend earlyWatch trends
Creatinine trendRises only after >50% GFR loss — a late marker. A rise of 0.2 mg/dL per cycle is a signal before any CTCAE Grade 2 threshold is met. Do not wait for the threshold.
Proteinuria (anti-VEGF)Dipstick ≥2+ → quantify with spot protein:creatinine ratio or 24h collection. Proteinuria >2g/24h → anti-VEGF suspension in most protocols.
MagnesiumCritical with EGFR inhibitors (cetuximab, panitumumab). Severe hypomagnesemia → arrhythmia risk. Must be checked every cycle — frequently omitted from routine panels.
PhosphorusHypophosphatemia is a class effect of FGFR inhibitors. Severe cases cause muscle weakness and rhabdomyolysis — which then elevates AST and confounds hepatic interpretation.
Endocrine irAEs — insidious and often permanentIO-specific
Endocrine irAEs are frequently underdiagnosed because their symptoms (fatigue, mood change, weight change) are attributed to cancer or treatment burden. Most result in permanent endocrine dysfunction requiring lifelong replacement.
TSH + free T4Hypothyroidism and hyperthyroidism are the most common endocrine irAEs. TSH must be measured every cycle or every 6–8 weeks. TSH >10 mIU/L in an asymptomatic patient still requires action.
Cortisol + ACTHImmune-mediated adrenal insufficiency (hypophysitis): profound fatigue + hypotension + hyponatremia during IO → urgent morning cortisol. Missed adrenal crisis = preventable death.
Glucose + HbA1cImmune-mediated Type 1 diabetes from PD-1/PD-L1 inhibitors. Can present as acute DKA with no prior diabetes history. New hyperglycemia >250 mg/dL in IO patient → urgent evaluation. HbA1c may be normal in acute-onset cases.
Rule: Fatigue in an IO patient is an endocrine irAE until proven otherwise. TSH, cortisol, and glucose should be reflex-checked before attributing fatigue to disease progression.
irAE recognition frameworkIO trials
TimingCan occur weeks to months after last IO dose. A patient 3 months post-IO discontinuation who develops colitis — irAE causality must still be considered.
Multi-organ involvementSimultaneous irAEs across organ systems are common. Colitis + hepatitis + thyroiditis = high-grade irAE syndrome, not three unrelated events.
Response to steroidsImprovement with systemic corticosteroids supports irAE diagnosis. No improvement within 48–72h → reassess, escalate to steroid-sparing agents, consider biopsy.
Pneumonitis signalNew dyspnea + SpO2 <94% or drop >3% from baseline in IO patient → pneumonitis evaluation before CT report. SpO2 at every visit is a critical data point.
Immune-mediated myocarditis — mortality 25–50%: Any troponin elevation above ULN in an IO patient requires same-day cardiac evaluation. New conduction abnormality (any AV block, new bundle branch block) in an IO patient = myocarditis until proven otherwise. This cannot wait for the next scheduled visit.
Hy's Law — FDA gold standard for drug-induced liver injury (DILI)
ALT or AST
Criterion 1
≥ 3× ULN — hepatocellular pattern required (not pure cholestasis)
Total bilirubin
Criterion 2
≥ 2× ULN with direct fraction >35% of total — excludes Gilbert syndrome
Criterion 3
No biliary obstruction + no other cause explaining the combination
Why Hy's Law is a regulatory red flagFDA critical
ALT elevation + impaired hepatic excretion (elevated bilirubin) indicates the liver has lost sufficient functional mass to compensate. Historically ~10% of patients meeting Hy's Law criteria develop fulminant hepatic failure. FDA estimates 1 confirmed case predicts ~10 severe jaundice cases and 1 death in the post-approval population. A single confirmed case can halt a development program.
Temple's Corollary: Many patients with ALT ≥3× ULN without full Hy's Law criteria indicate that DILI potential exists — the denominator is growing. Track the proportion of patients with ALT ≥3× ULN as a leading indicator, even before any Hy's Law case is confirmed.
Criterion 3 — exclusion checklist (mandatory for every case)Must complete
- Viral hepatitis excluded: HAV IgM, HBsAg + HBV DNA, HCV antibody + RNA (reactivation risk with IO and rituximab-containing regimens)
- Biliary obstruction excluded by imaging (ultrasound or CT) — especially with periportal lymphadenopathy or pancreatic head lesions
- Hepatic metastases assessed — bilirubin driven by intrahepatic tumor progression is NOT Hy's Law, it is disease progression. CT correlation mandatory.
- Hepatotoxic concomitants reviewed: acetaminophen dose, amoxicillin-clavulanate, statins, azoles, methotrexate, amiodarone, herbal supplements
- Alcohol use assessed (AST/ALT ratio — in alcoholic hepatitis AST:ALT >2:1)
- Autoimmune hepatitis excluded: ANA, anti-smooth muscle antibody, IgG levels — especially in IO where immune hepatitis mimics Hy's Law
- R-value calculated and documented: (ALT/ULN) ÷ (ALP/ULN). R>5 = hepatocellular. R<2 = cholestatic. Document in case narrative.
- Timeline confirmed: bilirubin elevation follows or is concurrent with ALT elevation — not preceded it (which suggests pre-existing biliary problem)
Oncology-specific considerationsContext
Hepatic metastasesMay elevate baseline bilirubin independent of drug effect. Establish pre-treatment hepatic function and document extent of liver involvement before the first dose. A patient with 60% hepatic replacement by tumor cannot be evaluated by standard Hy's Law criteria without contextual adjustment.
IO combination regimensIO + targeted therapy increases immune-mediated hepatitis risk. Mechanism is T-cell mediated inflammation, not direct hepatocellular toxicity — but can meet Hy's Law criteria. Liver biopsy showing CD8+ T-cell infiltration is diagnostic.
Hy's Law scatter plotAt every data cut: plot peak ALT/ULN (x-axis) vs peak total bilirubin/ULN (y-axis) for all patients. Any point in the upper-right quadrant (ALT ≥3× AND bili ≥2×) requires individual narrative review. This is the standard FDA expects in every clinical study report.
QTc thresholds — absolute values AND deltasMemorize both
<450ms (M) / <470ms (F)Normal. Continue per protocol with standard monitoring frequency.
450–480msGrade 1 CTCAE. Review all QTc-prolonging concomitants. Check electrolytes (K, Mg, Ca). Increase ECG monitoring frequency.
480–500msGrade 2. Temporary suspension per protocol. Cardiology consultation. Correct all reversible factors before re-dosing.
>500msGrade 3 — critical. Immediate suspension. Urgent cardiology evaluation. Risk of Torsades de Pointes → ventricular fibrillation → sudden death.
Delta >60ms from baselineIndependent alarm regardless of absolute value. QTc 410ms → 475ms = 65ms delta = red flag even though 475ms is below the absolute threshold. Always track delta from each patient's own baseline.
Fridericia formula (FDA preferred): QTcF = QT ÷ ∛(RR in seconds). Bazett overestimates QTc at HR >100 bpm — common in oncology patients with fever, anemia, or anxiety — generating false positives. Verify which formula the central ECG lab uses and whether heart rate changes are confounding the measurement.
QTc-prolonging concomitants in oncology patientsScreen every cycle
AntiemeticsOndansetron IV (especially ≥32mg single dose), domperidona, metoclopramide — routine in oncology supportive care
AntibioticsAzithromycin, clarithromycin, fluoroquinolones (ciprofloxacin, levofloxacin) — frequently started during febrile neutropenia episodes
AntifungalsFluconazole, voriconazole, posaconazole — used as prophylaxis during neutropenia. Also CYP3A4 inhibitors: double interaction risk (QTc prolongation + TKI level elevation)
AntipsychoticsHaloperidol (delirium management), quetiapine — common in inpatient oncology settings
Electrolyte disturbancesHypokalemia, hypomagnesemia, hypocalcemia directly prolong QTc independent of any drug. EGFR inhibitors cause hypomagnesemia — compounding QTc risk if study drug also prolongs QT.
LVEF — structured monitoringCardiac function
Clinically significant dropDrop ≥10 percentage points WITH resultant LVEF <50%, OR drop ≥15 percentage points regardless of final value → suspension criteria in most protocols with anthracyclines or anti-HER2 agents.
Measurement variabilityEchocardiography has ±5–8% interobserver variability. Do not over-interpret small changes. Trend across ≥3 measurements is far more reliable than any single data point.
Modality consistencyEcho and MUGA are not interchangeable in the same patient. Verify the same modality is used throughout. A mid-study switch introduces a systematic bias that cannot be corrected retrospectively.
Cumulative anthracycline doseTrack lifetime dose including pre-study treatment. Doxorubicin >400 mg/m², epirubicin >900 mg/m² = high-risk threshold. Damage is largely irreversible — unlike anti-HER2 dysfunction which is typically reversible upon discontinuation.
Baseline is mandatory: Without a pre-treatment LVEF, no post-treatment value can be meaningfully interpreted. Verify it was obtained before the first dose, with the correct modality, and documented in the CRF.
Immune-mediated myocarditis — most lethal irAE25–50% mortality
Detected late because it presents atypically. Investigators frequently attribute symptoms to cancer, anemia, or treatment fatigue. Active surveillance in the data is required — cannot wait for spontaneous reporting.
Troponin (any elevation)Any troponin above ULN in an IO patient = presumed myocarditis until ruled out. Same-day cardiac evaluation — not at the next scheduled visit.
Conduction abnormalitiesNew AV block of any degree, new bundle branch block (especially LBBB) — most frequent ECG finding in IO myocarditis. Can appear before any symptom or troponin elevation.
Diffuse ST-T changesNon-coronary distribution ST-T changes suggest diffuse myocardial inflammation. The ECG narrative comment must distinguish this from regional ischemic pattern.
BNP / NT-proBNPVentricular wall stress marker. Elevated BNP + elevated troponin = dual signal → act immediately. BNP elevation alone in IO patient without cardiac history → check troponin even if not scheduled.
Time is absolute: IO myocarditis can progress to complete heart block, ventricular arrhythmia, and death within hours to days. A troponin elevation reported on Friday afternoon cannot wait until Monday.
Cardiotoxicity mechanism by drug classReference
AnthracyclinesFree radical generation + mitochondrial dysfunction in cardiomyocytes. Cumulative, dose-dependent, largely irreversible. Damage may present months to years after treatment completion.
Anti-HER2 (trastuzumab)HER2 inhibition in cardiomyocytes disrupts a survival pathway. Functional dysfunction without direct cell death — therefore largely reversible upon discontinuation.
TKIs (sunitinib, sorafenib)Off-target kinase inhibition → hypertension (anti-VEGF effect), QTc prolongation (hERG channel), LVEF decrease. Hypertension is a class AE — must be monitored at every visit.
Checkpoint inhibitorsT-cell mediated immune attack on myocardium. Risk highest with combination PD-1 + CTLA-4 blockade. Standard cardiac monitoring is insufficient — must add troponin and ECG surveillance.
Proteasome inhibitorsCarfilzomib (more than bortezomib) associated with cardiac failure and hypertension. Baseline cardiac evaluation required before initiating.
Response assessment — RECIST 1.1 vs iRECIST
RECIST 1.1 response categoriesSolid tumors
| Response | Definition | What to verify in data review |
|---|---|---|
| CR | Complete disappearance of all target and non-target lesions. No new lesions. | All lesions truly zero? Or just below measurability threshold? Non-target lesions fully resolved? |
| PR | ≥30% decrease in sum of diameters from baseline. | Baseline sum used as reference (not nadir). Imaging modality consistent with baseline? |
| SD | Between PR and PD criteria. Neither sufficient shrinkage nor growth. | Minimum duration confirmed (≥6–8 weeks per protocol)? SD at first assessment without confirmed duration does not qualify. |
| PD | ≥20% increase AND ≥5mm absolute increase from nadir, OR unequivocal new lesion. | Is the new lesion truly new? Was the 5mm absolute threshold met? Both criteria required. |
iRECIST — immunotherapy-specific frameworkIO trials
| Term | Meaning | Action required in data review |
|---|---|---|
| iUPD | Unconfirmed progressive disease — first apparent progression per RECIST 1.1 in IO trial | Ensure confirmatory scan is scheduled 4–8 weeks later. Patient must NOT be withdrawn at iUPD. Flag any premature withdrawal — protocol deviation. |
| iCPD | Confirmed progressive disease — progression confirmed at repeat assessment | True endpoint. Verify confirmation assessment was done within the protocol window. |
| Pseudoprogression | Apparent tumor growth (inflammatory infiltrate) before regression begins | Patient with iUPD who subsequently achieves PR or SD = pseudoprogression. Distinguish from hyperprogression. |
| Hyperprogression | Accelerated tumor growth after IO initiation (>2× pre-treatment growth rate) | Compare growth trajectory before and after IO initiation. Red flag — IO may be driving tumor growth. |
Most common efficacy gap in IO trials: Patients withdrawn at iUPD without confirmatory assessment. This removes potentially responding patients from the efficacy analysis and inflates the apparent PD rate. Actively track all iUPD cases and their confirmation status.
Endpoints — what to verify and whenBy endpoint
ORRVerify response confirmation at ≥4 weeks. Consistent imaging modality across assessments. Concordance between investigator and IRC assessments documented.
DoRResponse date = first assessment meeting criteria, not date report was finalized. Censoring rules applied correctly for patients without documented progression.
PFSAssessment windows respected — off-schedule scans affect PFS calculation. Censoring of patients who died without documented progression follows protocol-specified rules.
OSLast contact dates current — patient listed as alive with last contact >6 months ago is a data gap. Post-study crossover treatment documented for all patients — required for OS sensitivity analysis.
Select a CRF domain
Medical history (MH)Baseline context
ECOG PSMost critical field. Must be assessed at baseline AND updated every visit — not carried forward unchanged. ECOG that never changes across 6 active chemotherapy cycles is almost certainly not being reassessed. Baseline ECOG determines the risk profile for interpreting all subsequent AEs.
Comorbidities vs CMComorbidities in MH should have corresponding medications in CM (diabetes → metformin; hypertension → antihypertensive). A comorbidity with no corresponding medication — or a medication with no corresponding condition — is a documentation gap requiring query.
Hepatic / renal baselinePre-existing liver disease (cirrhosis, chronic hepatitis, steatosis) and renal impairment radically change interpretation of lab toxicities. Must be explicitly documented, not inferred from lab values.
Cardiac historyPrior MI, arrhythmia, heart failure, baseline LVEF — directly affects QTc and LVEF monitoring thresholds and interpretation throughout the study.
Prior therapy (TH)Exposure history
Cumulative anthracycline doseSum all prior anthracycline exposure across all prior lines. A patient who received doxorubicin 300 mg/m² in adjuvant breast cancer 5 years ago has a significantly reduced cardiac safety margin entering the current study.
Prior radiation fieldsMediastinal radiation → increased cardiac risk. Pulmonary radiation → reduced reserve, higher pneumonitis risk with IO. Hepatic radiation → compromised function. The radiation field is as important as the dose.
Prior IO exposurePre-existing irAEs from prior checkpoint inhibitors may recur or be potentiated. A patient with prior grade 2 colitis on pembrolizumab entering a nivolumab study is at higher GI irAE risk.
Reason for prior discontinuationIf a previous line was stopped for hepatotoxicity and the study drug is also hepatotoxic — flag for enhanced hepatic monitoring from cycle 1, day 1.
Inclusion / exclusion (IE)Protocol integrity
Screening vs baseline labsScreening labs determine eligibility; baseline labs establish safety monitoring reference. Creatinine 1.0 at screening but 1.8 at baseline → patient deteriorated between visits. Was the investigator aware? Is the patient truly eligible?
Borderline valuesValues at exactly the eligibility cut-off deserve scrutiny. Were prior values available? Is the value confirmed? A single borderline value used to qualify a patient is a fragile data point.
Prohibited concomitantsAny prohibited medication in CM from the start of the study = protocol deviation. Cross-reference the CM domain against the IE exclusion criteria systematically for every patient.
Drug exposure (EX)Dose integrity
Relative dose intensity (RDI)Actual cumulative dose ÷ planned cumulative dose × 100%. RDI <85% indicates significant dose reductions or delays affecting both safety interpretation (AEs at reduced dose ≠ AEs at full dose) and efficacy analysis (sub-therapeutic exposure may explain apparent lack of response).
Modification justificationEvery dose reduction or delay must be traceable to a documented AE or lab finding. A reduction with no supporting AE entry is a data gap. An active grade ≥2 AE with no dose action must also be explained and documented.
Dose during active AETreatment administered on a date when a grade 3–4 AE was active = critical finding. Either AE was not recognized at dosing time, protocol criteria were not applied, or the AE start date is incorrect.
Concomitant medications (CM)AE attribution
Indication as AE surrogateConcomitant drug indication is frequently more informative than the AE domain. Antiemetic without nausea. Opioid without pain. Antidiarrheal without diarrhea. Colony-stimulating factor without documented neutropenia. These are footprints of unreported AE entries.
Start date correlationNew concomitant starting during the study without a documented reason is a data quality signal. Must correlate with either pre-study history (chronic medications) or a specific clinical event.
Corticosteroids in IOFor every corticosteroid course: indication (irAE vs other), starting dose (mg/day prednisone equivalent), taper schedule, and duration. Critical for irAE management adequacy and efficacy confounding analysis.
Adverse events (AE / SAE)Core safety domain
Internal consistencyGrade vs action taken must be consistent. Grade 3 AE with no dose modification, no hospitalization, and no concomitant medication to manage it is implausible — either the grade is overstated or management is undocumented.
Expected class AEsAbsence of class-expected AEs across an entire cohort is a systematic under-reporting signal, not a favorable safety finding. The reviewer who knows the drug mechanism knows what AEs to expect.
Resolution without evidenceAEs marked "resolved" in patients who discontinued without documented follow-up assessment — resolution was assumed, not confirmed. Flag and query.
Causality assignmentsCausality "not related" is not final. Positive dechallenge + positive rechallenge = strong causal evidence regardless of investigator opinion. Document and escalate for sponsor medical review.
Tumor assessment (TU)Efficacy integrity
Assessment window complianceOff-schedule scans affect PFS calculation. Verify assessment dates fall within protocol-specified windows. Systematic late assessments at one site indicate an operational problem.
Non-target and new lesionsMost frequently incomplete fields. Must be captured at every timepoint. Missing non-target data = incomplete response assessment = potentially invalid RECIST assignment.
iUPD tracking in IOEvery iUPD patient must have a confirmatory assessment within the protocol-specified timeframe. Maintain a live list of all iUPD patients and their confirmation status at every data review.
Vital signs (VS)Often overlooked
Blood pressure (anti-VEGF)Hypertension is a class AE with anti-VEGF agents. BP must be trended from baseline. A rise from 120/80 to 155/95 is CTCAE Grade 2 even if the value is clinically normal by population standards — must be documented as AE and managed.
SpO2 in IOOxygen saturation at every visit is an early pneumonitis signal. SpO2 <94% or drop >3% from baseline in IO patient with respiratory symptoms → pneumonitis evaluation before CT report.
WeightRapid gain (>3–5 kg in 2 weeks) → fluid retention — possible cardiac decompensation or steroid-induced edema. Progressive loss → disease progression or treatment intolerance signal.
TemperatureFever ≥38.3°C in patient with ANC <1000/mm³ = febrile neutropenia. Cross-reference fever in VS with ANC values from the same visit — sites frequently document fever but fail to generate the AE entry.
ECG (EG)Cardiac safety
Triplicate complianceMany protocols require triplicate ECGs (3 recordings, 1–2 min apart) with the average QTc reported. Verify the site performed triplicates and not a single ECG — a single value may not represent the true QTc.
Timing relative to dosePre-dose and post-dose ECG timing per protocol (e.g., 2–4h post-infusion for peak drug concentration). An ECG done at the wrong timepoint is not interpretable for the intended safety assessment.
Heart rate at ECGIf HR >100 bpm (tachycardia from fever, anemia) and Bazett formula was used, QTc is artificially elevated. Check HR recorded with the ECG and whether Fridericia was applied.
Conduction findingsBeyond QTc — new PR prolongation, new bundle branch block, new AV block. These are not captured by QTc alone. The cardiologist narrative comment must be reviewed, not just the QTc number. In IO patients, new conduction changes = myocarditis signal.
Biomarkers / PK (MB)Translational signals
Biomarker timingBaseline tumor biopsy must be obtained before first dose. A "baseline" biopsy taken after 2 treatment cycles is not a baseline — introduces a critical confound for biomarker-response correlation.
PK outliersPatients in the top exposure quartile (high AUC) with severe toxicity and patients in the bottom quartile (low AUC) without response are the most informative for understanding the therapeutic window.
PD-L1 / TMB / MSIBiomarker status must be pre-specified and documented before treatment. Missing biomarker data at data cut reduces ability to perform biomarker-efficacy subgroup analyses — a key regulatory expectation for IO trials.
Quality of life / PROPatient perspective
PRO vs AE concordancePatient reports severe pain and fatigue in the PRO but AE domain shows only Grade 1 events → discrepancy suggesting under-reporting. PRO data is a validation tool for the AE domain.
Administration timingPROs must be completed before the patient receives tumor assessment results. Post-result PRO completion is biased — knowledge of response or progression changes responses.
Missing PRO dataHigh rates of missing questionnaires at a specific site may indicate incorrect administration procedures, or that patients are too ill to complete them — itself clinically meaningful and requiring documentation.
Survival / follow-up (DS/SV)Most critical endpoint
Cause of death attributionDeath may be due to disease progression, study drug toxicity, post-study treatment toxicity, or an unrelated cause. Each requires different attribution. Review the full narrative for every on-study death and every death within 30 days of last dose. "Disease progression" as cause of death in a patient with grade 4 hepatotoxicity the prior week requires investigation.
Last contact date currencyPatient listed as alive with last contact >6 months ago → data currency problem. Site must confirm survival status. Stale last-contact dates systematically underestimate OS event rates.
Post-study crossoverControl arm patients receiving the experimental drug after progression confounds OS analysis. Post-study treatment must be documented for all patients — required for OS sensitivity analysis per the statistical analysis plan.
How CRF domains form a clinical narrativeIntegration framework
The data reviewer reads the CRF as a clinician reads a chart — looking for internal coherence and for breaks in the narrative thread. When the thread breaks, that is where the gap lives.
MH + Prior Therapy
→
Defines baseline risk profile. Pre-existing comorbidities and prior drug exposures set the context for interpreting every downstream finding.
IE criteria
→
Confirms the patient was correctly enrolled. Eligibility violations invalidate patient data and may represent an ongoing safety risk.
EX (exposure)
→
Defines what the patient actually received. Without knowing the dose received, AEs cannot be graded in exposure context and efficacy cannot be interpreted.
CM (concomitants)
→
Provides alternative explanations for every AE. Always review CM before finalizing AE causality. Also reveals unreported AEs through indication-AE mismatch.
VS + Labs + ECG
→
Objective measurements that corroborate or contradict the AE narrative. A patient with "mild fatigue" (grade 1) but a 5 g/dL hemoglobin drop has a discrepancy requiring investigation.
AE domain
→
Primary safety data source. Must be internally consistent with EX, CM, VS, labs — and concordant with the PRO patient perspective.
TU (tumor assessment)
→
Efficacy data. In IO, interpreted with iRECIST. Correlate with irAE data — a patient with colitis + tumor shrinkage tells a different story than tumor shrinkage alone.
PRO / QoL
→
Patient-reported validation of the AE domain. Systematic discordance = under-reporting signal at the site level.
DS / SV
→
Final outcome. Integrates all preceding domains. Cause of death must be traceable through the entire prior clinical narrative — it should never be a surprise.
When the thread breaks: Concomitant with no corresponding AE. Dose reduction with no documented reason. Death whose cause does not follow from the clinical narrative. Lab value that contradicts the AE grade. These breaks are where the investigative review work begins.
Clinical case walkthrough — IO trial, solid tumor (NSCLC)Simulated case
58-year-old, stage IV NSCLC, ECOG 1, no prior IO exposure, enrolled in a PD-1 inhibitor trial. Follow the narrative thread the data reviewer must construct.
W2
Week 2 — ALT 2.1× ULN
Below Hy's Law threshold. Note it. Review CM — no new hepatotoxic concomitants. Review MH — no prior liver disease. Baseline ALT was normal. Delta is meaningful. Flag for enhanced monitoring and request next cycle ALT be prioritized in the data cleaning queue.
W6
Week 6 — ALT 4.2×, AST 3.8×, total bilirubin 1.9× ULN
Approaching Hy's Law. Immediately review: direct bilirubin fraction (35% → Criterion 2 borderline). ALP — R-value = 3.1 (mixed pattern). Review CM — amoxicillin-clavulanate started 3 weeks ago for sinusitis. Is this antibiotic DILI or immune-mediated hepatitis from IO? Both are possible. Criterion 3 cannot be excluded yet. Escalate to medical reviewer.
W6+3d
Medical review initiated — request viral hepatitis panel and liver ultrasound
HBsAg, HBV DNA, HCV antibody, HAV IgM — all negative. Ultrasound: no biliary obstruction, liver metastases stable per prior CT. Amoxicillin-clavulanate stopped at W6. Document: Criterion 3 partially addressed, antibiotic as contributing factor not fully excluded. Study drug hold recommended per protocol. Repeat labs in 1 week.
W7
Week 7 — ALT 6.8×, bilirubin 2.4× ULN. Antibiotic stopped 10 days ago.
ALT continued rising after antibiotic discontinuation — critical signal. If purely antibiotic DILI, improvement should have begun. Continued rise despite antibiotic cessation shifts probability to immune-mediated hepatitis from IO. Hy's Law Criterion 1 and 2 both now met. Criterion 3: antibiotic less likely as sole cause. Declare potential Hy's Law case. SAE filed. Sponsor notification. Liver biopsy recommended.
W8
Liver biopsy — CD8+ T-cell infiltration. Diagnosis: immune-mediated hepatitis.
Confirms IO causality. High-dose corticosteroids initiated (prednisone 1–2 mg/kg/day). Update SAE narrative with biopsy findings. Confirm causality as "related" to study drug. Document that antibiotic was a confounding factor now excluded. Monitor ALT trend on steroids — improvement expected within 48–72h. Document Criterion 3 final conclusion.
What this case demonstrates: The narrative required simultaneous tracking of: ALT trend (labs), antibiotic timing (CM), bilirubin fractionation (labs), hepatic imaging (TU domain correlation), viral serology (additional testing), and biopsy (definitive attribution). No single domain told the full story. The reviewer integrated all of them.
Seven guiding questions — sequential data reviewDecision framework
These questions apply to every patient at every review cycle. Each answer either closes the loop or opens an investigation.
1. Is everything that should be there, there?All scheduled assessments completed? Labs, ECGs, tumor scans, PROs — on time and within protocol windows? Missing data is not a neutral event; it creates uncertainty that affects safety and efficacy interpretation.
2. Is the clinical story internally consistent?Do AEs, labs, vital signs, and PROs tell the same story? A patient with "mild" AEs but multiple symptom-management concomitants, progressive weight loss, and declining ECOG is not clinically mild.
3. What else could explain this finding?For every new AE or lab abnormality: disease progression? Concomitant medication? Pre-existing condition? Intercurrent illness? Only after excluding alternatives can the study drug be attributed with confidence.
4. What does the trend say that the point doesn't?A value within normal range that has doubled over 3 cycles is more informative than a mildly abnormal stable value. Always read values as data points on a trend line, not isolated measurements.
5. Is this patient safe to receive the next dose?The operational question that drives every per-patient review. All prior questions converge here. The answer must be a clear yes, no, or conditional yes with specific conditions — before the next dose date.
6. Is this part of a pattern across the cohort?Is this the 4th patient with grade 3 ALT elevation? Is the rate of immune-mediated events approaching the protocol's safety stopping rule threshold? Individual events become signals when they aggregate.
7. What action is required and by when?Every finding that cannot be closed with "monitor per protocol" requires a specific action, a responsible party, and a timeline. "Continue to observe" is not findings management — it is risk deferral.
Regulatory and scientific standards
ICH E2AClinical safety data management: definitions and standards for expedited reporting. Governs SAE and SUSAR timelines: 7 days fatal/life-threatening; 15 days serious unexpected.
ICH E6 (R2/R3)Good Clinical Practice. Defines sponsor oversight responsibilities, source data verification, and risk-based monitoring principles applicable to all clinical roles.
ICH E9 (R1)Statistical principles and estimands framework. Defines how intercurrent events (dose modifications, AE-driven discontinuations) are handled in efficacy analysis. Directly relevant to OS and PFS interpretation.
CTCAE v5.0Common Terminology Criteria for Adverse Events. The grading system for all AEs in oncology trials. Every reviewer must know grade definitions for key toxicities of the study drug class.
RECIST 1.1Response Evaluation Criteria in Solid Tumors. Standard for tumor measurement and response assessment with chemotherapy or targeted therapy.
iRECISTRECIST modification for immunotherapy. Introduces iUPD/iCPD framework to accommodate pseudoprogression. Seymour et al., Lancet Oncology 2017.
FDA DILI GuidanceHy's Law definition, Temple's Corollary, R-value classification, and hepatic safety data expectations in clinical study reports.
ASCO/SITC irAE GuidelinesManagement guidelines for immune-related adverse events from checkpoint inhibitor therapy. Grading, management algorithms, and steroid dosing by organ system and grade.
EMA Oncology GuidelineEMA/CHMP guideline on evaluation of anticancer medicinal products. Covers endpoint selection, biomarker strategy, and benefit-risk assessment framework.
MedDRAMedical Dictionary for Regulatory Activities. AE coding hierarchy: SOC → HLGT → HLT → PT → LLT. Coding accuracy at PT level directly affects aggregate safety tables and signal detection.