I'll pull the relevant safety profile, precedent criteria, and lab thresholds to ground this in precedent.
{
"drug_class": "PI3K inhibitor",
"class_effect_aes": [
"Hyperglycemia (on-target PI3K-alpha effect; can be severe/ketoacidotic)",
"Hepatotoxicity / transaminase elevation",
"Severe diarrhea / colitis (delta-isoform)",
"Rash (including severe cutaneous reactions)",
"Pneumonitis; opportunistic infection (PJP)"
],
"monitoring": [
"Baseline + on-treatment fasting glucose and HbA1c",
"Frequent LFT monitoring (esp. early cycles)",
"Vigilance for diarrhea/colitis; consider PJP prophylaxis",
"Skin assessment for rash"
],
"contraindicated_comorbidities": [
"Poorly controlled diabetes mellitus / baseline HbA1c above threshold",
"Significant baseline hepatic impairment",
"History of severe cutaneous reactions / drug-induced colitis",
"Active pneumonitis"
],
"ddi_flags": [
"Strong CYP3A4 inducers/inhibitors alter exposure",
"Anti-hyperglycemics may be needed concomitantly; coordinate with endocrinology",
"Avoid agents that compound hepatotoxicity"
],
"notes": "Hyperglycemia and hepatotoxicity dominate; baseline glycemic control and LFT thresholds are the key safety-driven inclusion/exclusion levers."
}{
"indication": "Gastric/GEJ",
"precedent_trials": [
"SPOTLIGHT/GLOW (zolbetuximab CLDN18.2)",
"CheckMate-649 (1L IO)"
],
"inclusion": [
"Histologically confirmed gastric or GEJ adenocarcinoma",
"Biomarker-defined where relevant (e.g. CLDN18.2-positive, HER2 status)",
"Measurable or evaluable disease per RECIST v1.1",
"ECOG 0-1",
"Adequate organ function (see lab thresholds)",
"Able to take oral medication / adequate nutrition (if oral agent)"
],
"exclusion": [
"Active GI bleeding or gastric outlet obstruction",
"Untreated CNS metastases",
"Significant cardiovascular disease (QTc, recent MI) if agent warrants",
"Prior lines beyond protocol limit / inadequate washout",
"Uncontrolled intercurrent illness or active infection"
],
"notes": "Gastric trials emphasize nutrition/GI integrity and biomarker selection; CLDN18.2 agents add CLDN18.2-positivity as a key inclusion."
}{
"hematologic": [
{
"parameter": "Absolute neutrophil count (ANC)",
"typical_threshold": ">= 1.5 x 10^9/L",
"rationale": "Adequate marrow reserve before a myelosuppressive agent; lower (>=1.0) sometimes allowed if marrow involvement."
},
{
"parameter": "Platelets",
"typical_threshold": ">= 100 x 10^9/L (>=75 if marrow involvement)",
"rationale": "Bleeding risk and tolerance of thrombocytopenia."
},
{
"parameter": "Hemoglobin",
"typical_threshold": ">= 9.0 g/dL (transfusion-independent)",
"rationale": "Baseline reserve; many protocols require independence from recent transfusion."
}
],
"hepatic": [
{
"parameter": "Total bilirubin",
"typical_threshold": "<= 1.5 x ULN (<=3 x ULN if Gilbert's/liver mets)",
"rationale": "Hepatic clearance reserve; relaxed for documented Gilbert's or hepatic metastases."
},
{
"parameter": "AST / ALT",
"typical_threshold": "<= 3 x ULN (<=5 x ULN if liver metastases)",
"rationale": "Baseline hepatocellular integrity; key for hepatotoxic agents (PI3Ki, JAKi)."
},
{
"parameter": "Albumin",
"typical_threshold": ">= 2.8-3.0 g/dL (some protocols)",
"rationale": "Synthetic liver function / nutritional status, esp. GI cancers."
}
],
"renal": [
{
"parameter": "Creatinine clearance (Cockcroft-Gault)",
"typical_threshold": ">= 50-60 mL/min",
"rationale": "Renal elimination reserve; threshold depends on renal clearance of the drug."
},
{
"parameter": "Serum creatinine",
"typical_threshold": "<= 1.5 x ULN",
"rationale": "Simpler surrogate when CrCl not used."
}
],
"cardiac": [
{
"parameter": "Left-ventricular ejection fraction (LVEF)",
"typical_threshold": ">= 50% (institutional LLN)",
"rationale": "Required for cardiotoxic agents (anti-HER2, anthracycline exposure)."
},
{
"parameter": "QTc (Fridericia)",
"typical_threshold": "<= 470 ms (women) / <= 450 ms (men)",
"rationale": "Exclude baseline prolongation for QT-liable agents; pair with congenital long-QT exclusion."
}
],
"coagulation_metabolic": [
{
"parameter": "INR / aPTT (if on anticoagulation off-study)",
"typical_threshold": "within therapeutic range / <= 1.5 x ULN",
"rationale": "Bleeding risk, esp. for agents with hemorrhage signal (BTKi)."
},
{
"parameter": "Fasting glucose / HbA1c",
"typical_threshold": "HbA1c <= 6.5-8.0% per protocol",
"rationale": "Baseline glycemic control for hyperglycemia-liable agents (PI3K-alpha)."
}
],
"notes": "Teaching-snapshot typical oncology cutoffs. Exact thresholds and relaxations (marrow involvement, liver mets, Gilbert's) are drug- and protocol-specific and must be set against the investigator's brochure and standard of care."
}Design note: The two dominant, on-target/class liabilities for a PI3Kα inhibitor are hyperglycemia (mechanistic, can be ketoacidotic) and hepatotoxicity (transaminitis). These criteria are built to screen out the genuinely high-risk patient while preserving enrollment — i.e., they gate on glycemic control and hepatic integrity but avoid blanket exclusion of liver-metastatic, Gilbert's, or diet-controlled-diabetic patients who are common and treatable in gastric/GEJ.
Disease / population
1. Histologically or cytologically confirmed adenocarcinoma of the stomach or gastroesophageal junction (GEJ), locally advanced unresectable or metastatic.
2. Biomarker-selected: documented activating PIK3CA mutation (and/or PI3K-pathway alteration per protocol) by a validated assay on tumor tissue or ctDNA. HER2 status documented (positive/negative) and prior anti-HER2 therapy permitted per pathway logic.
3. Prior systemic therapy consistent with the protocol line (e.g., progression on/after ≥1 prior line including a platinum/fluoropyrimidine); Phase 1 dose-escalation may allow more heavily pre-treated patients.
4. Measurable or evaluable disease per RECIST v1.1.
Performance / organ function
5. ECOG performance status 0–1.
6. Life expectancy ≥ 12 weeks.
7. Able to swallow and retain oral medication; adequate nutrition (no uncontrolled gastric outlet obstruction). Feeding-tube administration permitted only if compatible with the formulation.
Hematologic (standard, not the limiting axis here — keep accessible)
8. ANC ≥ 1.5 × 10⁹/L.
9. Platelets ≥ 100 × 10⁹/L.
10. Hemoglobin ≥ 9.0 g/dL (transfusion permitted, but independent of transfusion within 7 days of assessment).
Hepatic (tightened — this is a hepatotoxic agent)
11. Total bilirubin ≤ 1.5 × ULN; ≤ 3 × ULN if documented Gilbert's syndrome.
12. AST and ALT ≤ 3 × ULN; ≤ 5 × ULN in the presence of liver metastases (common in gastric/GEJ — relaxation preserves enrollment without ignoring the hepatotoxicity signal).
13. Albumin ≥ 2.8 g/dL (synthetic function + nutritional surrogate; lower end of range chosen deliberately to not over-exclude GI-cancer patients).
Renal
14. Creatinine clearance ≥ 50 mL/min (Cockcroft-Gault), or serum creatinine ≤ 1.5 × ULN.
Glycemic — the defining safety gate
15. Fasting plasma glucose ≤ 140 mg/dL (≤ 7.8 mmol/L) AND HbA1c ≤ 7.0% at screening.
- Rationale: HbA1c ≤ 7.0% (rather than ≤6.5%) admits well-controlled diabetics — a defensible middle ground that doesn't exclude the large diet- or metformin-controlled population, while still requiring genuine baseline control before adding a hyperglycemia-inducing agent.
- Diet- and/or oral-agent-controlled type 2 diabetes is permitted; insulin-dependent diabetes is permitted only if HbA1c ≤ 7.0% and a glycemic-management plan is in place with endocrinology co-management.
Reproductive / general
16. Adequate contraception; negative pregnancy test in persons of childbearing potential.
17. Recovery from prior-therapy toxicity to ≤ Grade 1 (excluding alopecia, stable neuropathy); washout per protocol (typically ≥ 3 weeks or 5 half-lives, whichever shorter).
Glycemic (mechanism-driven)
1. Type 1 diabetes mellitus.
2. Poorly controlled type 2 diabetes (HbA1c > 7.0% or fasting glucose > 140 mg/dL at screening despite optimization).
3. History of diabetic ketoacidosis or hyperosmolar hyperglycemic state.
Hepatic (mechanism-driven)
4. Active or chronic hepatitis B (HBV DNA detectable) or hepatitis C (HCV RNA detectable) without adequate suppression; known cirrhosis Child-Pugh B/C.
5. History of drug-induced liver injury, autoimmune hepatitis, or non-malignant hepatic disease causing transaminitis above the inclusion thresholds.
6. Concurrent use of known hepatotoxic agents that cannot be substituted.
Class-effect (other PI3Ki toxicities)
7. History of severe cutaneous adverse reaction (SJS/TEN/DRESS) or drug-induced colitis.
8. Active (non-infectious) pneumonitis / interstitial lung disease, or history of grade ≥2 pneumonitis.
GI integrity (gastric-specific)
9. Active GI bleeding or uncontrolled gastric outlet obstruction.
10. Conditions impairing oral absorption (uncontrolled malabsorption, prior surgery materially affecting absorption) that would compromise drug exposure.
General oncology
11. Untreated or symptomatic CNS metastases (treated, stable, off escalating steroids ≥ 2 weeks is eligible).
12. Uncontrolled intercurrent illness or active infection requiring systemic therapy.
13. Clinically significant cardiovascular disease (recent MI, unstable arrhythmia); QTc(F) > 470 ms or congenital long-QT if the specific agent carries a QT signal.
Drug interactions
14. Strong CYP3A4 inhibitors or inducers within the protocol washout (alters PI3Kα exposure) where substitution is not feasible.
| Lever | Choice | Why it's defensible / non-over-restrictive |
|---|---|---|
| HbA1c | ≤ 7.0% | Excludes uncontrolled diabetes (real ketoacidosis risk) but admits well-controlled/diet-controlled patients vs. an overly tight ≤6.5%. |
| Diabetes | Controlled T2DM allowed | Outright diabetes exclusion would needlessly shrink the gastric population; gated on control + endocrine co-management instead. |
| AST/ALT & bilirubin | Liver-mets and Gilbert's relaxations | Liver metastases are common in gastric/GEJ; the relaxation (5× ULN / 3× ULN) prevents excluding otherwise-eligible patients while keeping the hepatotoxicity guardrail. |
| Albumin | ≥ 2.8 g/dL | Lower end of the standard range — GI-cancer patients are frequently mildly hypoalbuminemic; a 3.0 cutoff would over-exclude. |
| Hematologic | Standard cutoffs | Not the limiting axis for PI3Kα; kept at conventional values rather than tightening. |
Open items to confirm against the Investigator's Brochure before finalizing: (1) exact PI3K-pathway biomarker definition and assay (tissue vs. ctDNA); (2) the agent's specific QT and CYP3A4 profile (drives criteria #13/#14); (3) whether PJP prophylaxis is mandated (raised in the safety profile); (4) protocol line-of-therapy limits, which differ between the Phase 1 escalation and Phase 2 expansion portions.
Would you like me to draft the on-treatment glucose/LFT monitoring and dose-modification schema that pairs with these baseline gates, or split the criteria into distinct Phase 1 (escalation) vs. Phase 2 (expansion) versions?
Design note: The two dominant, on-target/class liabilities for a PI3Kα inhibitor are hyperglycemia (mechanistic, can be ketoacidotic) and hepatotoxicity (transaminitis). These criteria are built to screen out the genuinely high-risk patient while preserving enrollment — i.e., they gate on glycemic control and hepatic integrity but avoid blanket exclusion of liver-metastatic, Gilbert's, or diet-controlled-diabetic patients who are common and treatable in gastric/GEJ.
Disease / population
1. Histologically or cytologically confirmed adenocarcinoma of the stomach or gastroesophageal junction (GEJ), locally advanced unresectable or metastatic.
2. Biomarker-selected: documented activating PIK3CA mutation (and/or PI3K-pathway alteration per protocol) by a validated assay on tumor tissue or ctDNA. HER2 status documented (positive/negative) and prior anti-HER2 therapy permitted per pathway logic.
3. Prior systemic therapy consistent with the protocol line (e.g., progression on/after ≥1 prior line including a platinum/fluoropyrimidine); Phase 1 dose-escalation may allow more heavily pre-treated patients.
4. Measurable or evaluable disease per RECIST v1.1.
Performance / organ function
5. ECOG performance status 0–1.
6. Life expectancy ≥ 12 weeks.
7. Able to swallow and retain oral medication; adequate nutrition (no uncontrolled gastric outlet obstruction). Feeding-tube administration permitted only if compatible with the formulation.
Hematologic (standard, not the limiting axis here — keep accessible)
8. ANC ≥ 1.5 × 10⁹/L.
9. Platelets ≥ 100 × 10⁹/L.
10. Hemoglobin ≥ 9.0 g/dL (transfusion permitted, but independent of transfusion within 7 days of assessment).
Hepatic (tightened — this is a hepatotoxic agent)
11. Total bilirubin ≤ 1.5 × ULN; ≤ 3 × ULN if documented Gilbert's syndrome.
12. AST and ALT ≤ 3 × ULN; ≤ 5 × ULN in the presence of liver metastases (common in gastric/GEJ — relaxation preserves enrollment without ignoring the hepatotoxicity signal).
13. Albumin ≥ 2.8 g/dL (synthetic function + nutritional surrogate; lower end of range chosen deliberately to not over-exclude GI-cancer patients).
Renal
14. Creatinine clearance ≥ 50 mL/min (Cockcroft-Gault), or serum creatinine ≤ 1.5 × ULN.
Glycemic — the defining safety gate
15. Fasting plasma glucose ≤ 140 mg/dL (≤ 7.8 mmol/L) AND HbA1c ≤ 7.0% at screening.
- Rationale: HbA1c ≤ 7.0% (rather than ≤6.5%) admits well-controlled diabetics — a defensible middle ground that doesn't exclude the large diet- or metformin-controlled population, while still requiring genuine baseline control before adding a hyperglycemia-inducing agent.
- Diet- and/or oral-agent-controlled type 2 diabetes is permitted; insulin-dependent diabetes is permitted only if HbA1c ≤ 7.0% and a glycemic-management plan is in place with endocrinology co-management.
Reproductive / general
16. Adequate contraception; negative pregnancy test in persons of childbearing potential.
17. Recovery from prior-therapy toxicity to ≤ Grade 1 (excluding alopecia, stable neuropathy); washout per protocol (typically ≥ 3 weeks or 5 half-lives, whichever shorter).
Glycemic (mechanism-driven)
1. Type 1 diabetes mellitus.
2. Poorly controlled type 2 diabetes (HbA1c > 7.0% or fasting glucose > 140 mg/dL at screening despite optimization).
3. History of diabetic ketoacidosis or hyperosmolar hyperglycemic state.
Hepatic (mechanism-driven)
4. Active or chronic hepatitis B (HBV DNA detectable) or hepatitis C (HCV RNA detectable) without adequate suppression; known cirrhosis Child-Pugh B/C.
5. History of drug-induced liver injury, autoimmune hepatitis, or non-malignant hepatic disease causing transaminitis above the inclusion thresholds.
6. Concurrent use of known hepatotoxic agents that cannot be substituted.
Class-effect (other PI3Ki toxicities)
7. History of severe cutaneous adverse reaction (SJS/TEN/DRESS) or drug-induced colitis.
8. Active (non-infectious) pneumonitis / interstitial lung disease, or history of grade ≥2 pneumonitis.
GI integrity (gastric-specific)
9. Active GI bleeding or uncontrolled gastric outlet obstruction.
10. Conditions impairing oral absorption (uncontrolled malabsorption, prior surgery materially affecting absorption) that would compromise drug exposure.
General oncology
11. Untreated or symptomatic CNS metastases (treated, stable, off escalating steroids ≥ 2 weeks is eligible).
12. Uncontrolled intercurrent illness or active infection requiring systemic therapy.
13. Clinically significant cardiovascular disease (recent MI, unstable arrhythmia); QTc(F) > 470 ms or congenital long-QT if the specific agent carries a QT signal.
Drug interactions
14. Strong CYP3A4 inhibitors or inducers within the protocol washout (alters PI3Kα exposure) where substitution is not feasible.
| Lever | Choice | Why it's defensible / non-over-restrictive |
|---|---|---|
| HbA1c | ≤ 7.0% | Excludes uncontrolled diabetes (real ketoacidosis risk) but admits well-controlled/diet-controlled patients vs. an overly tight ≤6.5%. |
| Diabetes | Controlled T2DM allowed | Outright diabetes exclusion would needlessly shrink the gastric population; gated on control + endocrine co-management instead. |
| AST/ALT & bilirubin | Liver-mets and Gilbert's relaxations | Liver metastases are common in gastric/GEJ; the relaxation (5× ULN / 3× ULN) prevents excluding otherwise-eligible patients while keeping the hepatotoxicity guardrail. |
| Albumin | ≥ 2.8 g/dL | Lower end of the standard range — GI-cancer patients are frequently mildly hypoalbuminemic; a 3.0 cutoff would over-exclude. |
| Hematologic | Standard cutoffs | Not the limiting axis for PI3Kα; kept at conventional values rather than tightening. |
Open items to confirm against the Investigator's Brochure before finalizing: (1) exact PI3K-pathway biomarker definition and assay (tissue vs. ctDNA); (2) the agent's specific QT and CYP3A4 profile (drives criteria #13/#14); (3) whether PJP prophylaxis is mandated (raised in the safety profile); (4) protocol line-of-therapy limits, which differ between the Phase 1 escalation and Phase 2 expansion portions.
Would you like me to draft the on-treatment glucose/LFT monitoring and dose-modification schema that pairs with these baseline gates, or split the criteria into distinct Phase 1 (escalation) vs. Phase 2 (expansion) versions?
Expected-coverage checklist for drug class 'PI3K inhibitor' — a grading aid, NOT ground truth. A strong I/E list visibly protects against each liability (exclusion / lab cutoff / monitoring) while keeping the target population enrollable.
| Class-effect liability to address | Why it constrains eligibility |
|---|---|
| Hyperglycemia (on-target PI3K-alpha effect; can be severe/ketoacidotic) | Require baseline glycemic control (HbA1c threshold); exclude uncontrolled diabetes. |
| Hepatotoxicity / transaminase elevation | Set LFT/bilirubin thresholds; exclude significant hepatic impairment. |
| Severe diarrhea / colitis (delta-isoform) | Exclude baseline colitis; supportive-care/monitoring plan. |
| Rash (including severe cutaneous reactions) | Should be reflected in an exclusion, lab threshold, or monitoring requirement. |
| Pneumonitis; opportunistic infection (PJP) | Exclude pneumonitis history; pulmonary symptom work-up protocol. |