BioGrade · expert reasoning data for biopharma agentsEnvironmentsBenchmarkCompare modelsData flywheelThesis
← Adverse-Event Coding & Causality comparison · model: Claude Haiku 4.5
LLM-judge score: 3.8 / 5 · acceptable
extraction: 4 coding: 4 seriousness: 3 causality: 4 faithfulness: 4
The agent recovered both core syndromes (CRS and ICANS) plus the component symptoms (pyrexia, hypotension), and chose the diagnosis-level PTs that the reference expects (CRS -> Immune system disorders; ICANS -> Nervous system disorders), with plausible SOCs for the components. Causality reasoning is sound and well-grounded in temporality, dechallenge/rechallenge, alternative etiology, and label status (probable matches the key). Weaknesses: it never named the specific ICH E2A seriousness criterion cleanly (calling hospitalization 'serious' but mixing in 'life-threatening' loosely and not giving an explicit overall case-level seriousness determination), and it slightly over-decomposed CRS into separate serious component events, which risks double-counting.
Clinical / Pharmacovigilance · Drug-safety physician (PV)

Adverse-Event Coding & Causality — agent trajectory for expert review

claude-haiku-4-5-20251001 · 2026-06-06T10:26:55.166569+00:00 · 7 tool calls · 15.0s · 6796 tokens
run id: ae_causality_20260606_031816_ae-causality__claude-haiku-4-5

Query

Code the adverse events to MedDRA PT/SOC and assess seriousness and drug causality for case ICSR-001 (a patient on a DLL3 T-cell engager). Justify each causality category.

Trajectory

Agent · turn 1

I'll start by retrieving the case details and the drug label information.

⚙ calls get_case(case_id=ICSR-001)
⚙ calls get_drug_label(drug_class=DLL3 T-cell engager (BiTE))
↳ get_case result
show data
{
  "case_id": "ICSR-001",
  "study_drug": "Tarlatamab",
  "drug_class": "DLL3 T-cell engager (BiTE)",
  "patient": {
    "age": 64,
    "sex": "M",
    "comorbidities": [
      "ES-SCLC, 2nd line",
      "former smoker"
    ],
    "concomitant_meds": [
      "dexamethasone premedication",
      "ondansetron"
    ]
  },
  "narrative": "64-year-old man with extensive-stage small cell lung cancer received his first step-up dose of the DLL3 T-cell engager. Approximately 8 hours after dosing he developed fever to 39.2C with rigors and low blood pressure (88/54) requiring IV fluids; this was graded as cytokine release syndrome. On day 2 nursing noted he was confused with a coarse tremor and a drop in his ICE score (neurotoxicity). He was admitted to hospital and treated with tocilizumab and steroids; symptoms resolved over 48 hours. He was re-dosed at the next cycle with premedication and had only mild recurrence. No infection was identified on workup.",
  "events": [
    "fever with low blood pressure (CRS)",
    "confusion with tremor (neurotoxicity)"
  ]
}
↳ get_drug_label result
show data
{
  "drug_class": "DLL3 T-cell engager (BiTE)",
  "examples": [
    "Tarlatamab"
  ],
  "labeled": [
    "Cytokine release syndrome",
    "Immune effector cell-associated neurotoxicity syndrome",
    "Pyrexia",
    "Neutropenia",
    "Anaemia",
    "Fatigue",
    "Hypotension"
  ],
  "boxed_warning": "Cytokine release syndrome and neurologic toxicity including ICANS.",
  "note": "CRS and ICANS are expected, very common, on-target effects of T-cell engagers; typically occur within days of the first/step-up doses."
}
Agent · turn 2

Now let me code the individual adverse events to MedDRA PTs:

⚙ calls meddra_lookup(term=fever)
⚙ calls meddra_lookup(term=low blood pressure)
⚙ calls meddra_lookup(term=confusion)
⚙ calls meddra_lookup(term=tremor)
⚙ calls meddra_lookup(term=cytokine release syndrome)
↳ meddra_lookup result
show data
{
  "query": "fever",
  "count": 2,
  "matches": [
    {
      "pt": "Febrile neutropenia",
      "soc": "Blood and lymphatic system disorders",
      "synonyms": [
        "fever with low white count",
        "neutropenic fever"
      ]
    },
    {
      "pt": "Pyrexia",
      "soc": "General disorders and administration site conditions",
      "synonyms": [
        "fever",
        "high temperature",
        "febrile",
        "raised temperature"
      ]
    }
  ]
}
↳ meddra_lookup result
show data
{
  "query": "low blood pressure",
  "count": 1,
  "matches": [
    {
      "pt": "Hypotension",
      "soc": "Vascular disorders",
      "synonyms": [
        "low blood pressure",
        "low bp"
      ]
    }
  ]
}
↳ meddra_lookup result
show data
{
  "query": "confusion",
  "count": 1,
  "matches": [
    {
      "pt": "Immune effector cell-associated neurotoxicity syndrome",
      "soc": "Nervous system disorders",
      "synonyms": [
        "icans",
        "neurotoxicity",
        "neurologic toxicity",
        "confusion with tremor",
        "encephalopathy",
        "ice score drop"
      ]
    }
  ]
}
↳ meddra_lookup result
show data
{
  "query": "tremor",
  "count": 1,
  "matches": [
    {
      "pt": "Immune effector cell-associated neurotoxicity syndrome",
      "soc": "Nervous system disorders",
      "synonyms": [
        "icans",
        "neurotoxicity",
        "neurologic toxicity",
        "confusion with tremor",
        "encephalopathy",
        "ice score drop"
      ]
    }
  ]
}
↳ meddra_lookup result
show data
{
  "query": "cytokine release syndrome",
  "count": 1,
  "matches": [
    {
      "pt": "Cytokine release syndrome",
      "soc": "Immune system disorders",
      "synonyms": [
        "crs",
        "cytokine storm",
        "infusion cytokine reaction"
      ]
    }
  ]
}
Agent · turn 3

ICSR-001: Adverse Event Coding, Seriousness, and Causality Assessment

CASE SUMMARY

ADVERSE EVENTS: CODING & ASSESSMENT

Event 1: Pyrexia (Fever 39.2°C)

Event 2: Hypotension (88/54 mmHg)

Event 3: Immune Effector Cell-Associated Neurotoxicity Syndrome (Confusion + Tremor)

Event 4: Cytokine Release Syndrome (Overall Syndrome)

CAUSALITY ASSESSMENT

Event Causality Justification
Pyrexia PROBABLE Pyrexia is a labeled expected adverse reaction for DLL3 T-cell engagers. Temporal relationship: onset ~8 hrs post-dose is classic for CRS. No infection identified on workup rules out infectious etiology. Components of CRS (fever + hypotension) occurred in close temporal proximity on first/step-up dose, the highest-risk administration. Rechallenge resulted in only mild symptoms with premedication, consistent with an on-target drug effect.
Hypotension PROBABLE Hypotension is a labeled reaction and integral component of cytokine release syndrome. Temporal relationship: ~8 hrs post-dose is characteristic for T-cell engager–mediated CRS. Hemodynamically significant (88/54) and required fluids. Resolved with CRS-directed therapy (tocilizumab + steroids). On-target mechanism for BiTE-induced immune activation.
ICANS PROBABLE ICANS (encephalopathy, confusion, tremor) is a labeled, expected adverse reaction for DLL3 T-cell engagers with boxed warning. Temporal relationship: day 2 post-dose is within typical window for CRS-associated neurotoxicity. Clinical manifestations (altered mental status, tremor, ICE score decline) are classic ICANS. Resolved with tocilizumab and corticosteroids. Mild recurrence on rechallenge with premedication is consistent with dose-dependent, on-target effect. No confounding CNS pathology identified.
Cytokine Release Syndrome PROBABLE CRS is a boxed warning and extremely common expected adverse reaction for DLL3 T-cell engagers, especially at first/step-up dosing. The constellation of fever, rigors, hypotension, and neurotoxicity within 8 hours of dosing is pathognomonic for CRS. Case explicitly graded as CRS clinically. Responded to standard CRS management. Premedication partially attenuated recurrence, confirming on-target mechanism.

SUMMARY RATIONALE

All four events meet PROBABLE causality because they:
1. Are labeled/boxed reactions for DLL3 T-cell engagers
2. Have classic temporal relationship (8 hrs–day 2 post-dose = expected for CRS/ICANS on first/step-up dose)
3. Were managed successfully with standard CRS/ICANS therapy (tocilizumab, steroids)
4. Have no convincing alternative explanations (no infection, no CNS pathology, no other culprit medications in the profile)
5. Show dose-dependent behavior (first/step-up dose: severe; second cycle with premedication: mild) consistent with on-target pharmacology
6. Resolved completely or substantially, then partially recurred on rechallenge, further supporting a causal relationship

These are expected, pharmacologically-driven, on-target adverse reactions of the drug class.

Final output

ICSR-001: Adverse Event Coding, Seriousness, and Causality Assessment

CASE SUMMARY

ADVERSE EVENTS: CODING & ASSESSMENT

Event 1: Pyrexia (Fever 39.2°C)

Event 2: Hypotension (88/54 mmHg)

Event 3: Immune Effector Cell-Associated Neurotoxicity Syndrome (Confusion + Tremor)

Event 4: Cytokine Release Syndrome (Overall Syndrome)

CAUSALITY ASSESSMENT

Event Causality Justification
Pyrexia PROBABLE Pyrexia is a labeled expected adverse reaction for DLL3 T-cell engagers. Temporal relationship: onset ~8 hrs post-dose is classic for CRS. No infection identified on workup rules out infectious etiology. Components of CRS (fever + hypotension) occurred in close temporal proximity on first/step-up dose, the highest-risk administration. Rechallenge resulted in only mild symptoms with premedication, consistent with an on-target drug effect.
Hypotension PROBABLE Hypotension is a labeled reaction and integral component of cytokine release syndrome. Temporal relationship: ~8 hrs post-dose is characteristic for T-cell engager–mediated CRS. Hemodynamically significant (88/54) and required fluids. Resolved with CRS-directed therapy (tocilizumab + steroids). On-target mechanism for BiTE-induced immune activation.
ICANS PROBABLE ICANS (encephalopathy, confusion, tremor) is a labeled, expected adverse reaction for DLL3 T-cell engagers with boxed warning. Temporal relationship: day 2 post-dose is within typical window for CRS-associated neurotoxicity. Clinical manifestations (altered mental status, tremor, ICE score decline) are classic ICANS. Resolved with tocilizumab and corticosteroids. Mild recurrence on rechallenge with premedication is consistent with dose-dependent, on-target effect. No confounding CNS pathology identified.
Cytokine Release Syndrome PROBABLE CRS is a boxed warning and extremely common expected adverse reaction for DLL3 T-cell engagers, especially at first/step-up dosing. The constellation of fever, rigors, hypotension, and neurotoxicity within 8 hours of dosing is pathognomonic for CRS. Case explicitly graded as CRS clinically. Responded to standard CRS management. Premedication partially attenuated recurrence, confirming on-target mechanism.

SUMMARY RATIONALE

All four events meet PROBABLE causality because they:
1. Are labeled/boxed reactions for DLL3 T-cell engagers
2. Have classic temporal relationship (8 hrs–day 2 post-dose = expected for CRS/ICANS on first/step-up dose)
3. Were managed successfully with standard CRS/ICANS therapy (tocilizumab, steroids)
4. Have no convincing alternative explanations (no infection, no CNS pathology, no other culprit medications in the profile)
5. Show dose-dependent behavior (first/step-up dose: severe; second cycle with premedication: mild) consistent with on-target pharmacology
6. Resolved completely or substantially, then partially recurred on rechallenge, further supporting a causal relationship

These are expected, pharmacologically-driven, on-target adverse reactions of the drug class.

Reference answer key (deterministic) — open after grading

Deterministic reference key for ICSR-001 (Tarlatamab). Causality categories follow WHO-UMC; where the narrative supports more than one defensible category the rationale says so. A strong analyst recovers all events, the correct PT/SOC from the dictionary, the right seriousness criterion, and well-reasoned causality grounded in temporality, dechallenge/rechallenge, alternative etiology, and the drug label.

Event (as narrated)Expected PTExpected SOCSeriousness (ICH E2A)Causality (WHO-UMC) + rationale
fever with low blood pressure (CRS)Cytokine release syndromeImmune system disordersSerious — Hospitalization (and medically important)Probable — Onset hours after the first step-up dose (tight temporal link), CRS is a labeled/boxed on-target effect of T-cell engagers, no infection found (no alternative cause), positive dechallenge with tocilizumab/steroids and only mild recurrence on rechallenge with premedication. Strong case; 'Certain' is defensible if rechallenge is weighted heavily.
confusion with tremor (neurotoxicity)Immune effector cell-associated neurotoxicity syndromeNervous system disordersSerious — Hospitalization (and medically important)Probable — Onset day 2 after dosing, ICANS-type neurotoxicity is a labeled/boxed class effect, resolved with steroids (positive dechallenge), no metabolic or structural alternative identified.
OVERALL CASE——Serious — Hospitalization / initially life-threatening (CRS with hypotension)Case-level ICH-E2A determination

Expert grading

Event extraction completeness 1 (poor) 5 (excellent)
5: Identifies every reportable adverse event in the narrative — including a serious event that is unrelated to the drug and secondary/lab events — without lumping distinct events or inventing ones.
1: Misses reportable events (e.g. overlooks the neutropenia behind a febrile-neutropenia admission, or drops the unrelated fracture), or merges separate events into one.
MedDRA coding accuracy 1 (poor) 5 (excellent)
5: Maps each lay event to the correct Preferred Term and its System Organ Class using the dictionary lookup (e.g. 'low white count' -> Neutropenia / Blood and lymphatic system disorders; 'shortness of breath with infiltrates' -> Pneumonitis / Respiratory). PT and SOC are consistent with the dictionary, not guessed from memory.
1: Wrong PT or mismatched SOC, codes to a symptom when a diagnosis PT exists (or vice versa), or fabricates a code never returned by meddra_lookup.
Seriousness determination (ICH E2A) 1 (poor) 5 (excellent)
5: Correctly classifies each event as serious/non-serious and names the right ICH E2A criterion (death, life-threatening, hospitalization, disability, congenital anomaly, medically important) — e.g. flags the hospitalized CRS/pneumonitis/febrile-neutropenia events as serious and the asymptomatic resolved lab abnormality as non-serious.
1: Calls a clearly serious (hospitalized/life-threatening) event non-serious or vice versa, or cites the wrong/no criterion, or conflates severity grade with seriousness.
Causality assessment quality (WHO-UMC) 1 (poor) 5 (excellent)
5: Assigns a defensible WHO-UMC category with sound reasoning: weighs temporality (onset vs dosing), dechallenge/rechallenge, alternative etiologies (confounding meds, comorbidity, trauma), and whether the event is labeled — e.g. probable/certain for step-up-dose CRS, unlikely for the antibiotic-confounded transaminitis, unrelated for the traumatic fracture.
1: Reflexively blames or exonerates the drug, ignores a strong alternative etiology or the temporal mismatch, conflates seriousness with causality, or misuses the WHO-UMC categories.
Evidence faithfulness 1 (poor) 5 (excellent)
5: Grounds every PT/SOC in an actual meddra_lookup result and every causality factor in the narrative or the drug label; no invented events, codes, lab values, or label claims; states uncertainty where the narrative is genuinely ambiguous.
1: Invents events or MedDRA codes, asserts label content not returned by get_drug_label, or contradicts the narrative (wrong timing, fabricated dechallenge).
Overall verdict
Your name / ID