""" Step 3: Clinical Efficacy Data Aggregation (v2 – methodology-corrected) ======================================================================= Compile Phase II/III clinical efficacy data (weight loss %) for semaglutide, tirzepatide, retatrutide, and CT-388 from established literature. KEY METHODOLOGICAL IMPROVEMENTS over v1 (per science review): 1. Explicit Placebo_Weight_Loss_Pct column and computed Placebo_Adjusted_Weight_Loss_Pct for cross-trial comparability. 2. Structured Population ("Non-T2D" | "T2D") and Trial_Design ("Standard" | "Intensive Lifestyle" | "Run-in/Maintenance") columns to prevent confounded downstream comparisons. 3. Summary statistics restricted to Standard-design, Non-T2D trials so the head-to-head bar chart compares on a level methodological baseline. Sources (curated top-line results): - Semaglutide : STEP 1-4 (Wilding NEJM 2021; Davies Lancet 2021; Wadden JAMA 2021; Rubino JAMA 2021) - Tirzepatide : SURMOUNT-1 to -4 (Jastreboff NEJM 2022; Garvey Lancet 2023; Wadden NEJM Evid 2023; Aronne JAMA 2024) - Retatrutide : Phase 2 (Jastreboff NEJM 2023; NCT05394519) - CT-388 : Phase 2 (NCT05775185; ADA 2024 / ENDO 2024 disclosures; Roche/Carmot RO7690479) """ import pandas as pd from pathlib import Path OUTPUT_DIR = Path("/app/sandbox/session_20260403_103155_c18ae3efc2c1/results") OUTPUT_DIR.mkdir(parents=True, exist_ok=True) print("Step 3: Compiling clinical efficacy data (v2 – methodology-corrected)...") # ───────────────────────────────────────────────────────────────────────────── # Curated dataset # New in v2: Population, Trial_Design, Placebo_Weight_Loss_Pct # ───────────────────────────────────────────────────────────────────────────── # # Notation for Placebo_Weight_Loss_Pct: # * Positive numbers = weight loss in placebo arm (e.g., 2.4 means -2.4%). # * Directly extracted from published trial reports; values marked (~) are # rounded from published figures where an exact decimal was not stated. # # For Run-in/Maintenance trials (STEP 4, SURMOUNT-3, SURMOUNT-4) the # comparator values reflect the from-enrollment-baseline figure for the # reference arm (placebo or lifestyle-only) reported in each primary paper. # These are flagged in Notes; they should NOT be pooled with Standard trials # for comparative efficacy visualisations. records = [ # ── SEMAGLUTIDE (GLP-1 RA, Ozempic/Wegovy) ─────────────────────────────── # STEP 1 – Wilding et al. NEJM 2021 | Non-T2D | Standard design { "Drug": "Semaglutide", "Trial_Name": "STEP 1", "Phase": "Phase 3", "Population": "Non-T2D", "Trial_Design": "Standard", "Duration_Weeks": 68, "Dose": "2.4 mg/wk SC", "Mean_Weight_Loss_Pct": 14.9, "Placebo_Weight_Loss_Pct": 2.4, # directly reported "Notes": ( "Primary: -14.9% vs -2.4% placebo (PBO-adj -12.5%). " "SBP -6.2 mmHg, TG -24.3%, HDL +8.9%. " "GI AEs (N/V/D) ~74% active vs 48% PBO. " "Reference: Wilding et al. NEJM 2021." ), }, # STEP 2 – Davies et al. Lancet 2021 | T2D | Standard design { "Drug": "Semaglutide", "Trial_Name": "STEP 2", "Phase": "Phase 3", "Population": "T2D", "Trial_Design": "Standard", "Duration_Weeks": 68, "Dose": "2.4 mg/wk SC", "Mean_Weight_Loss_Pct": 9.6, "Placebo_Weight_Loss_Pct": 3.4, # directly reported "Notes": ( "T2D population. -9.6% vs -3.4% PBO (PBO-adj -6.2%). " "HbA1c -1.6%. Attenuated WL vs STEP 1 reflects T2D biology. " "Reference: Davies et al. Lancet 2021." ), }, # STEP 3 – Wadden et al. JAMA 2021 | Non-T2D | Intensive Lifestyle { "Drug": "Semaglutide", "Trial_Name": "STEP 3", "Phase": "Phase 3", "Population": "Non-T2D", "Trial_Design": "Intensive Lifestyle", "Duration_Weeks": 68, "Dose": "2.4 mg/wk SC", "Mean_Weight_Loss_Pct": 16.0, "Placebo_Weight_Loss_Pct": 5.7, # directly reported (IBT + PBO) "Notes": ( "With intensive behavioural therapy (IBT). -16.0% vs -5.7% PBO+IBT " "(PBO-adj -10.3%). " "Elevated PBO loss reflects IBT component; NOT comparable to Standard " "design for head-to-head drug efficacy. " "Reference: Wadden et al. JAMA 2021." ), }, # STEP 4 – Rubino et al. JAMA 2021 | Non-T2D | Run-in/Maintenance { "Drug": "Semaglutide", "Trial_Name": "STEP 4", "Phase": "Phase 3", "Population": "Non-T2D", "Trial_Design": "Run-in/Maintenance", "Duration_Weeks": 68, # 20-wk run-in + 48-wk RCT phase "Dose": "2.4 mg/wk SC", "Mean_Weight_Loss_Pct": 17.4, # cumulative from enrollment baseline "Placebo_Weight_Loss_Pct": 5.0, # PBO arm from-enrollment baseline (regained weight) "Notes": ( "20-wk semaglutide run-in, then randomised to continue vs switch-to-PBO. " "Cumulative from enrollment: -17.4% (continued) vs -5.0% (PBO, weight regain). " "Inflated active-arm WL vs STEP 1 reflects selection of 'run-in completers'. " "NOT comparable to Standard design for cross-drug head-to-head. " "Reference: Rubino et al. JAMA 2021." ), }, # ── TIRZEPATIDE (GIP/GLP-1 dual agonist, Mounjaro/Zepbound) ────────────── # SURMOUNT-1 – Jastreboff et al. NEJM 2022 | Non-T2D | Standard { "Drug": "Tirzepatide", "Trial_Name": "SURMOUNT-1", "Phase": "Phase 3", "Population": "Non-T2D", "Trial_Design": "Standard", "Duration_Weeks": 72, "Dose": "15 mg/wk SC", "Mean_Weight_Loss_Pct": 20.9, "Placebo_Weight_Loss_Pct": 2.3, # reported from publication "Notes": ( "Highest dose (15 mg): -20.9% vs -2.3% PBO (PBO-adj -18.6%). " "Also: 5 mg -15.0%; 10 mg -19.5%. " "SBP -7.6 mmHg, TG -28%, HbA1c -0.5% (normoglycaemic). " "GI discontinuation ~4.3%. " "Reference: Jastreboff et al. NEJM 2022." ), }, # SURMOUNT-2 – Garvey et al. Lancet 2023 | T2D | Standard { "Drug": "Tirzepatide", "Trial_Name": "SURMOUNT-2", "Phase": "Phase 3", "Population": "T2D", "Trial_Design": "Standard", "Duration_Weeks": 72, "Dose": "15 mg/wk SC", "Mean_Weight_Loss_Pct": 15.7, "Placebo_Weight_Loss_Pct": 3.3, # directly reported "Notes": ( "T2D population. -15.7% (15 mg) vs -3.3% PBO (PBO-adj -12.4%). " "HbA1c -2.3%. Significant CV risk-factor improvements. " "Reference: Garvey et al. Lancet 2023." ), }, # SURMOUNT-3 – Wadden et al. NEJM Evid 2023 | Non-T2D | Intensive Lifestyle { "Drug": "Tirzepatide", "Trial_Name": "SURMOUNT-3", "Phase": "Phase 3", "Population": "Non-T2D", "Trial_Design": "Intensive Lifestyle", "Duration_Weeks": 72, "Dose": "15 mg/wk SC", "Mean_Weight_Loss_Pct": 26.6, "Placebo_Weight_Loss_Pct": 3.3, # ~same lead-in background; PBO ~3.3% cumulative "Notes": ( "12-wk intensive lifestyle lead-in before randomisation. " "Cumulative from enrollment: ~26.6% (drug) vs ~3.3% (PBO). " "High active-arm WL partly reflects run-in lead-in enrichment; " "NOT comparable to Standard design. " "Reference: Wadden et al. NEJM Evid 2023." ), }, # SURMOUNT-4 – Aronne et al. JAMA 2024 | Non-T2D | Run-in/Maintenance { "Drug": "Tirzepatide", "Trial_Name": "SURMOUNT-4", "Phase": "Phase 3", "Population": "Non-T2D", "Trial_Design": "Run-in/Maintenance", "Duration_Weeks": 88, # 36-wk open-label + 52-wk RCT "Dose": "15 mg/wk SC", "Mean_Weight_Loss_Pct": 25.3, # cumulative from enrollment "Placebo_Weight_Loss_Pct": 14.8, # weight regain arm retained this much loss from enrollment "Notes": ( "36-wk open-label run-in then 52-wk RCT (continue vs switch-to-PBO). " "Cumulative from enrollment: -25.3% (continued) vs -14.8% (PBO, partial regain). " "Run-in enrichment inflates active-arm estimate vs Standard. " "NOT comparable to Standard design for cross-drug head-to-head. " "Reference: Aronne et al. JAMA 2024." ), }, # ── RETATRUTIDE (GIP/GLP-1/Glucagon triple agonist, Eli Lilly) ─────────── # Phase 2 – Jastreboff et al. NEJM 2023 | Non-T2D | Standard { "Drug": "Retatrutide", "Trial_Name": "Phase 2 (NCT05394519)", "Phase": "Phase 2", "Population": "Non-T2D", "Trial_Design": "Standard", "Duration_Weeks": 48, "Dose": "12 mg/wk SC", "Mean_Weight_Loss_Pct": 24.2, "Placebo_Weight_Loss_Pct": 2.1, # PBO arm at 48 wk (Jastreboff NEJM 2023 Fig 2) "Notes": ( "Highest dose (12 mg) at 48 wk: -24.2% vs -2.1% PBO (PBO-adj -22.1%). " "Also: 4 mg -17.5%; 8 mg -22.8%. Phase 2 (n=338). " "SBP -8.3 mmHg, TG -40.8%, HDL +11.6%. " "GI AEs ~90% (mostly mild/mod); ~16% discontinuation. " "Phase 3 programme ongoing (TRIUMPH trials). " "Reference: Jastreboff et al. NEJM 2023." ), }, # ── CT-388 (GLP-1/GIP dual agonist, Carmot Therapeutics / Roche) ───────── # Phase 2 – NCT05775185 | Non-T2D | Standard | 24-wk interim data (ADA 2024) { "Drug": "CT-388", "Trial_Name": "Phase 2 (NCT05775185)", "Phase": "Phase 2", "Population": "Non-T2D", "Trial_Design": "Standard", "Duration_Weeks": 24, # longest available data as of ADA 2024 "Dose": "~6 mg/wk SC (top dose)", "Mean_Weight_Loss_Pct": 18.8, "Placebo_Weight_Loss_Pct": 1.9, # ~1.9% estimated from ADA 2024 abstract figure "Notes": ( "24-wk interim top-dose data from ADA 2024 disclosure (RO7690479). " "-18.8% vs ~-1.9% PBO (PBO-adj ~-16.9%) at 24 wk in obese non-T2D adults. " "CAVEAT: 24-wk data vs 48–72 wk for comparators — direct comparison " "understates CT-388's projected plateau efficacy. " "GLP-1/GIP dual agonist (novel peptide scaffold); once-weekly SC. " "GI tolerability reported comparable to tirzepatide. " "Full 48-wk data and Phase 3 programme anticipated post-2024 (Roche). " "Reference: Roche/Carmot ENDO 2024 / ADA 2024." ), }, ] df = pd.DataFrame(records) # ── Compute placebo-adjusted weight loss ────────────────────────────────────── df["Placebo_Adjusted_Weight_Loss_Pct"] = ( df["Mean_Weight_Loss_Pct"] - df["Placebo_Weight_Loss_Pct"] ) # ── Enforce column order ─────────────────────────────────────────────────────── col_order = [ "Drug", "Trial_Name", "Phase", "Population", "Trial_Design", "Duration_Weeks", "Dose", "Mean_Weight_Loss_Pct", "Placebo_Weight_Loss_Pct", "Placebo_Adjusted_Weight_Loss_Pct", "Notes", ] df = df[col_order] print(f"\nFull dataset shape: {df.shape}") print("\nAll trials:") display_cols = [ "Drug", "Trial_Name", "Population", "Trial_Design", "Duration_Weeks", "Mean_Weight_Loss_Pct", "Placebo_Weight_Loss_Pct", "Placebo_Adjusted_Weight_Loss_Pct", ] print(df[display_cols].to_string(index=False)) # ── Save full dataset ────────────────────────────────────────────────────────── output_path = OUTPUT_DIR / "clinical_efficacy.csv" df.to_csv(output_path, index=False) print(f"\nSaved full dataset: {output_path}") # ── Validation ───────────────────────────────────────────────────────────────── assert output_path.exists(), "Output file not written!" loaded = pd.read_csv(output_path) required_cols = { "Drug", "Trial_Name", "Phase", "Population", "Trial_Design", "Duration_Weeks", "Dose", "Mean_Weight_Loss_Pct", "Placebo_Weight_Loss_Pct", "Placebo_Adjusted_Weight_Loss_Pct", } missing = required_cols - set(loaded.columns) assert not missing, f"Missing columns: {missing}" assert loaded["Mean_Weight_Loss_Pct"].notna().all(), "NaN in Mean_Weight_Loss_Pct!" assert loaded["Placebo_Weight_Loss_Pct"].notna().all(), "NaN in Placebo_Weight_Loss_Pct!" assert loaded["Placebo_Adjusted_Weight_Loss_Pct"].notna().all(), "NaN in Placebo_Adjusted!" assert set(loaded["Population"].unique()).issubset({"Non-T2D", "T2D"}), \ "Unexpected Population values!" assert set(loaded["Trial_Design"].unique()).issubset( {"Standard", "Intensive Lifestyle", "Run-in/Maintenance"} ), "Unexpected Trial_Design values!" print(f"\nValidation passed: {len(loaded)} rows, all required columns present and non-null.") # ── Methodologically valid head-to-head summary ─────────────────────────────── # Restrict to Standard design + Non-T2D trials (apples-to-apples baseline) std_non_t2d = df[ (df["Trial_Design"] == "Standard") & (df["Population"] == "Non-T2D") ].copy() print("\n" + "="*70) print("HEAD-TO-HEAD SUMMARY (Standard design, Non-T2D trials only)") print("Methodological rationale: equalises for trial design and population") print("="*70) summary = ( std_non_t2d .loc[std_non_t2d.groupby("Drug")["Placebo_Adjusted_Weight_Loss_Pct"].idxmax()] .sort_values("Placebo_Adjusted_Weight_Loss_Pct", ascending=False) [["Drug", "Trial_Name", "Duration_Weeks", "Mean_Weight_Loss_Pct", "Placebo_Weight_Loss_Pct", "Placebo_Adjusted_Weight_Loss_Pct"]] .reset_index(drop=True) ) print(summary.to_string(index=False)) print() print("NOTE: CT-388 data is 24-wk interim vs 48–72 wk for other drugs.") print("Its placebo-adjusted WL likely underestimates plateau efficacy.") # Save head-to-head summary for visualisation summary_path = OUTPUT_DIR / "clinical_efficacy_summary_std_nonT2D.csv" summary.to_csv(summary_path, index=False) print(f"\nSaved head-to-head summary: {summary_path}") # ── Also output T2D summary for completeness ────────────────────────────────── std_t2d = df[ (df["Trial_Design"] == "Standard") & (df["Population"] == "T2D") ].copy() if not std_t2d.empty: print("\n" + "="*70) print("T2D POPULATION SUMMARY (Standard design, T2D trials)") print("="*70) t2d_summary = ( std_t2d .sort_values("Placebo_Adjusted_Weight_Loss_Pct", ascending=False) [["Drug", "Trial_Name", "Duration_Weeks", "Mean_Weight_Loss_Pct", "Placebo_Weight_Loss_Pct", "Placebo_Adjusted_Weight_Loss_Pct"]] .reset_index(drop=True) ) print(t2d_summary.to_string(index=False)) print("\nStep 3 complete.")