{ "_description": "Published meta-analysis benchmarks for RapidMeta validation. benchmark_type distinguishes external IPD vs aggregate vs self-reference pools; pool_type flags same-drug vs class-level pooling (class-level triggers GRADE indirectness downgrade).", "_version": "1.1.1", "_date": "2026-04-20", "benchmarks": { "FINERENONE_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.86, "lci": 0.78, "uci": 0.95, "k": 2, "N": 13026, "source": "Agarwal R et al. Lancet 2022; 400:1788-1801 (FIDELITY pooled FIDELIO-DKD + FIGARO-DKD)", "pmid": "36351458", "method": "pre-specified individual patient data pooled analysis", "method_note": "Published benchmark is IPD/patient-level; app uses aggregate-data DL random-effects. Direct point-estimate comparison expected to differ slightly.", "benchmark_type": "external_IPD", "pool_type": "same_drug" }, "BEMPEDOIC_ACID_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.87, "lci": 0.79, "uci": 0.96, "k": 1, "N": 13970, "source": "Nissen SE et al. N Engl J Med 2023; 388:1353-1364 (CLEAR Outcomes)", "pmid": "36876740", "method": "single RCT (used as anchor; DL pooled with CLEAR-Harmony gives similar)", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "external_aggregate", "pool_type": "same_drug" }, "COLCHICINE_CVD_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.75, "lci": 0.61, "uci": 0.91, "k": 2, "N": 11816, "source": "COLCOT (Tardif JC, N Engl J Med 2019; 381:2497) + LoDoCo2 (Nidorf SM, N Engl J Med 2020; 383:1838)", "pmid_colcot": "31733140", "pmid_lodoco2": "32865375", "method": "DL random-effects pooling of 2 landmark RCTs", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "GLP1_CVOT_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.88, "lci": 0.84, "uci": 0.94, "k": 8, "N": 60080, "source": "Sattar N et al. Lancet Diabetes Endocrinol 2021; 9:653-662", "pmid": "34462123", "method": "IPD meta-analysis of 8 GLP-1RA CVOTs", "method_note": "Published benchmark is IPD/patient-level; app uses aggregate-data DL random-effects. Direct point-estimate comparison expected to differ slightly.", "benchmark_type": "external_IPD", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "SGLT2_HF_REVIEW.html": { "primary_outcome": "CV death or HF hospitalization", "estimand": "HR", "pooled": 0.77, "lci": 0.71, "uci": 0.84, "k": 2, "N": 9689, "source": "Vaduganathan M et al. Lancet 2022; 400:757-767 (DELIVER + DAPA-HF pooled)", "pmid": "36041475", "method": "pre-specified cross-trial pooled analysis", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "external_aggregate", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "PCSK9_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.85, "lci": 0.79, "uci": 0.92, "k": 2, "N": 46488, "source": "Guedeney P et al. Eur Heart J 2020; 41:3336-3347 (FOURIER + ODYSSEY Outcomes)", "pmid": "31270523", "method": "DL pooling of 2 large PCSK9i outcome trials", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "external_aggregate", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "INTENSIVE_BP_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.79, "lci": 0.71, "uci": 0.87, "k": 5, "N": 21546, "source": "Internal DL pool: SPRINT-SENIOR (Williamson 2016, HR 0.66) + SPRINT-CKD (Cheung 2017, HR 0.81) + ACCORD-BP (Cushman 2010, HR 0.88) + STEP (Zhang 2021, HR 0.74) + SPS3-BP (Benavente 2013, HR 0.81). External benchmark for comparison: Xie 2016 Lancet HR 0.84 (ACM); Ettehad 2016 Lancet HR 0.87.", "pmid": "26551272", "method": "DL random-effects of 5 intensive-BP RCTs/strata. Caveat: SPRINT-SENIOR and SPRINT-CKD are non-independent age/CKD subgroups of parent SPRINT trial; treated as separate strata for subgroup analysis. Q=3.6, df=4, I^2=0%.", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "SGLT2_CKD_REVIEW.html": { "primary_outcome": "Kidney composite", "estimand": "HR", "pooled": 0.68, "lci": 0.6, "uci": 0.77, "k": 3, "N": 29025, "source": "Nuffield Dept Population Health. Lancet 2022; 400:1788 (SGLT2i MA: CREDENCE + DAPA-CKD + EMPA-KIDNEY)", "pmid": "36351458", "method": "DL random-effects pooling of 3 dedicated CKD trials", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "external_aggregate", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "ARNI_HF_REVIEW.html": { "primary_outcome": "CV death or HF hospitalization", "estimand": "HR", "pooled": 0.84, "lci": 0.77, "uci": 0.91, "k": 3, "N": 19519, "source": "McMurray JJV (PARADIGM-HF, NEJM 2014) + Solomon SD (PARAGON-HF, NEJM 2019) + Pfeffer MA (PARADISE-MI, NEJM 2021)", "method": "DL pooling of 3 sacubitril/valsartan RCTs", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "ABLATION_AF_REVIEW.html": { "primary_outcome": "All-cause mortality", "estimand": "HR", "pooled": 0.77, "lci": 0.64, "uci": 0.93, "k": 4, "N": 7211, "source": "Marrouche NF (CASTLE-AF, NEJM 2018) + Packer DL (CABANA, JAMA 2019) + Kirchhof P (EAST-AFNET4, NEJM 2020) + Tang ASL (RAFT-AF, NEJM 2022)", "method": "DL pooling of 4 ablation-vs-drug RCTs in HF+AF", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "IV_IRON_HF_REVIEW.html": { "primary_outcome": "CV death or HF hospitalization", "estimand": "HR", "pooled": 0.84, "lci": 0.74, "uci": 0.96, "k": 4, "N": 6404, "source": "Anker SD (CONFIRM-HF 2014, AFFIRM-AHF 2020) + Kalra PR (IRONMAN 2022) + Mentz RJ (HEART-FID 2023)", "method": "DL pooling of 4 IV iron RCTs in HFrEF", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "RENAL_DENERV_REVIEW.html": { "primary_outcome": "Office SBP reduction", "estimand": "MD", "pooled": -5.12, "lci": -6.85, "uci": -3.4, "k": 5, "N": 1013, "source": "Azizi M (RADIANCE-HTN SOLO/TRIO) + Bohm M (SPYRAL HTN-OFF/ON MED) + Kandzari DE (SYMPLICITY HTN-3 ON-MED)", "method": "DL pooling of weighted mean difference in office SBP (mmHg)", "unit": "mmHg", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity", "surrogate_endpoint": true, "surrogate_note": "primary outcome is a regulatory surrogate; GRADE indirectness should carry an additional downgrade for the surrogate -> clinical-outcome step" }, "DOAC_CANCER_VTE_REVIEW.html": { "primary_outcome": "VTE recurrence", "estimand": "HR", "pooled": 0.55, "lci": 0.3, "uci": 1.0, "k": 4, "N": 2894, "source": "Raskob GE (HOKUSAI-VTE Cancer 2018) + Young AM (SELECT-D 2018) + McBane RD (ADAM-VTE 2020) + Agnelli G (CARAVAGGIO 2020)", "method": "DL pooling of 4 DOAC-vs-LMWH cancer VTE RCTs", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "MAVACAMTEN_HCM_REVIEW.html": { "primary_outcome": "NYHA improvement OR", "estimand": "OR", "pooled": 6.67, "lci": 2.09, "uci": 21.3, "k": 3, "N": 444, "source": "Heitner SB (EXPLORER-HCM, Lancet 2019) + Desai MY (VALOR-HCM, NEJM 2022) + Chinese Phase 3 (2023)", "method": "DL pooling of published adjusted odds ratios for NYHA class improvement", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "RIVAROXABAN_VASC_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.85, "lci": 0.77, "uci": 0.94, "k": 4, "N": 45242, "source": "Eikelboom JW (COMPASS, NEJM 2017) + Bonaca MP (VOYAGER-PAD, NEJM 2020) + Zannad F (COMMANDER-HF 2018) + Mega JL (ATLAS-ACS2 2012)", "method": "DL pooling of 4 rivaroxaban vascular protection RCTs", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "LIPID_HUB_REVIEW.html": { "primary_outcome": "MACE", "estimand": "HR", "pooled": 0.89, "lci": 0.76, "uci": 1.04, "k": 5, "N": 50000, "source": "Internal DL pool: REDUCE-IT (Bhatt 2019, HR 0.75) + STRENGTH (Nicholls 2020, HR 1.02) + VITAL (Manson 2019, HR 0.92) + OMEMI (Kalstad 2021, HR 1.08) + RESPECT-EPA (Nishizaki 2024, HR 0.79)", "method": "DL random-effects of all 5 EPA/omega-3 CVOTs. High heterogeneity (I^2~78%) reflects EPA-only (REDUCE-IT/RESPECT-EPA, beneficial) vs mixed EPA+DHA (STRENGTH/VITAL/OMEMI, neutral). Honest pool crosses null. Earlier 3-trial pool of 0.75 was selection-driven.", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "INCRETIN_HFpEF_REVIEW.html": { "primary_outcome": "Worsening HF events / CV death composite", "estimand": "HR", "pooled": 0.41, "lci": 0.22, "uci": 0.79, "k": 3, "N": 1876, "source": "Internal DL pool: STEP-HFpEF (Kosiborod 2023, Peto HR 0.18 from 1/263 vs 12/266 worsening HF) + STEP-HFpEF DM (Kosiborod 2024, Peto HR 0.40 from 7/310 vs 18/306) + SUMMIT (Packer 2024, published HR 0.62 for CV death + HF events composite)", "method": "DL random-effects of 3 incretin HFpEF RCTs. STEP HRs Peto-derived from worsening-HF event counts (KCCQ-CSS was primary). SUMMIT contributes published Cox HR of composite primary. Outcome heterogeneity (worsening-HF vs CV-death-composite) flagged. I^2~57%.", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "ATTR_CM_REVIEW.html": { "primary_outcome": "All-cause mortality", "estimand": "HR", "pooled": 0.71, "lci": 0.59, "uci": 0.86, "k": 4, "N": 1741, "source": "DL pool: ATTR-ACT (Maurer 2018, HR 0.70) + ATTRibute-CM (Gillmore 2024, HR 0.72) + HELIOS-B (Fontana 2024, HR 0.67) + APOLLO-B (Berk 2023, Peto HR 0.99 from 10/181 vs 10/179)", "pmid": "30145929", "method": "DL random-effects of 4 ATTR-CM RCTs; APOLLO-B contributes Peto-derived HR (mortality was exploratory in APOLLO-B). Q=0.63, df=3, I^2=0%.", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "JAK_UC_REVIEW.html": { "primary_outcome": "Clinical remission at week 8/10 (induction in moderate-severe UC)", "estimand": "OR", "pooled": 3.2, "lci": 2.4, "uci": 4.3, "k": 3, "N": 1414, "source": "Sandborn 2017 NEJM (OCTAVE Induction 1/NCT01465763, tofacitinib) + Danese 2022 Lancet (U-ACHIEVE/NCT02819635, upadacitinib) + Feagan 2021 Lancet (SELECTION Induction A/NCT02914522, filgotinib). Singh 2023 Clin Gastro meta pooled OR ~3.2.", "pmid_octave": "28467869", "pmid_uachieve": "35644166", "pmid_selection": "34002586", "method": "DL random-effects OR pool of clinical remission at induction end across 3 phase 3 JAK inhibitor UC RCTs", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "external_aggregate", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "BIOLOGIC_ASTHMA_REVIEW.html": { "primary_outcome": "Annualized exacerbation rate ratio in severe asthma", "estimand": "RR", "pooled": 0.52, "lci": 0.44, "uci": 0.61, "k": 4, "N": 3074, "source": "Ortega 2014 NEJM (MENSA/NCT01691521, mepolizumab) + FitzGerald 2016 Lancet (CALIMA/NCT01914757, benralizumab) + Castro 2018 NEJM (QUEST/NCT02414854, dupilumab) + Menzies-Gow 2021 NEJM (NAVIGATOR/NCT03347279, tezepelumab). Pooled class rate ratio across 4 biologic-target RCTs.", "pmid_mensa": "25199059", "pmid_calima": "27609408", "pmid_quest": "29782217", "pmid_navigator": "33979488", "method": "Random-effects exacerbation rate ratio pool across 4 severe-asthma biologic trials (anti-IL5, anti-IL5R, anti-IL4R, anti-TSLP)", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "CFTR_CF_REVIEW.html": { "primary_outcome": "Absolute ppFEV1 change at 24-48 weeks (placebo-corrected MD)", "estimand": "MD", "pooled": 11.5, "lci": 9.5, "uci": 13.5, "k": 3, "N": 671, "source": "Ramsey 2011 NEJM (STRIVE-CF/NCT00909220, ivacaftor G551D) + Heijerman 2019 Lancet (VX17-445-102/NCT03525444, ELX/TEZ/IVA F508del homo) + Middleton 2019 NEJM (VX17-445-103/NCT03525548, ELX/TEZ/IVA F508del/MF). Heterogeneous class pool (mono vs triple modulator).", "pmid_strive": "22047557", "pmid_102": "31679946", "pmid_103": "31697873", "method": "Random-effects MD pool of absolute ppFEV1 change across 3 CFTR-modulator phase 3 RCTs. High I^2 expected given mutation-genotype differences and modulator generation.", "unit": "percentage points ppFEV1", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "CGRP_MIGRAINE_REVIEW.html": { "primary_outcome": "Change in monthly migraine days (placebo-corrected MD)", "estimand": "MD", "pooled": -1.7, "lci": -2.0, "uci": -1.4, "k": 3, "N": 2688, "source": "Goadsby 2017 NEJM (STRIVE/NCT02456740, erenumab) + Stauffer 2018 JAMA Neurol (EVOLVE-1/NCT02614183, galcanezumab) + Dodick 2018 JAMA (HALO-EM/NCT02629861, fremanezumab). One pivotal episodic-migraine trial per CGRP mAb.", "pmid_strive": "29180078", "pmid_evolve1": "29813147", "pmid_halo": "29800212", "method": "Fixed-effect pool of 4-6 month placebo-corrected MMD reduction (MD) across 3 phase 3 CGRP mAb episodic-migraine RCTs", "unit": "monthly migraine days change vs placebo", "method_note": "Published benchmark is fixed-effect; app uses DL random-effects. In low-I^2 topics the two match closely.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "PARP_OVARIAN_REVIEW.html": { "primary_outcome": "Progression-free survival (1st-line PARP maintenance in ovarian cancer)", "estimand": "HR", "pooled": 0.52, "lci": 0.42, "uci": 0.64, "k": 3, "N": 1930, "source": "Moore 2018 NEJM (SOLO-1/NCT01844986, olaparib BRCA+) + Ray-Coquard 2019 NEJM (PAOLA-1/NCT02477644, olaparib+bevacizumab all-comer) + Gonzalez-Martin 2019 NEJM (PRIMA/NCT02655016, niraparib all-comer). Tew 2020 ASCO Guideline pooled PFS HR ~0.52.", "pmid_solo1": "30345884", "pmid_paola1": "31851799", "pmid_prima": "31562799", "method": "DL random-effects PFS HR pool across 3 phase 3 PARP maintenance 1st-line ovarian trials. High I^2 expected (SOLO-1 BRCA-restricted vs PAOLA-1/PRIMA all-comer populations).", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "ANTIAMYLOID_AD_REVIEW.html": { "primary_outcome": "CDR-SB change at 18 months (adjusted mean difference vs placebo)", "estimand": "MD", "pooled": -0.45, "lci": -0.65, "uci": -0.25, "k": 3, "N": 4526, "source": "van Dyck 2023 NEJM (Clarity-AD/NCT03887455) + Sims 2023 JAMA (TRAILBLAZER-ALZ2/NCT04437511) + Budd Haeberlein 2022 JPAD (EMERGE/NCT02484547). Range of published CDR-SB MDs: lecanemab -0.45, donanemab -0.70, aducanumab-high-dose -0.39.", "pmid_lecanemab": "36449413", "pmid_donanemab": "37471501", "pmid_aducanumab": "34532727", "method": "Fixed-effect pool of 18-month CDR-SB change (MD) across three phase-3 anti-amyloid monoclonal antibody trials", "unit": "CDR-SB point change", "method_note": "Published benchmark is fixed-effect; app uses DL random-effects. In low-I^2 topics the two match closely.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "CDK46_MBC_REVIEW.html": { "primary_outcome": "Progression-free survival (1st-line HR+/HER2- mBC)", "estimand": "HR", "pooled": 0.56, "lci": 0.48, "uci": 0.64, "k": 3, "N": 1827, "source": "Finn 2016 NEJM (PALOMA-2/NCT01740427) + Hortobagyi 2016 NEJM (MONALEESA-2/NCT01958021) + Goetz 2017 JCO (MONARCH-3/NCT02246621). Pool of 3 CDK4/6i + AI 1st-line trials.", "pmid_paloma2": "27959613", "pmid_monaleesa2": "27717303", "pmid_monarch3": "28968163", "method": "DL random-effects pool of PFS HR across 3 phase 3 RCTs of CDK4/6 inhibitors + aromatase inhibitor vs AI alone in HR+/HER2- first-line mBC", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "ARPI_mHSPC_REVIEW.html": { "primary_outcome": "Overall survival in metastatic hormone-sensitive prostate cancer", "estimand": "HR", "pooled": 0.67, "lci": 0.61, "uci": 0.74, "k": 4, "N": 4633, "source": "Armstrong 2022 JCO (ARCHES/NCT02677896) + Davis 2019 NEJM (ENZAMET/NCT02446405) + Chi 2019 NEJM (TITAN/NCT02489318) + Smith 2022 NEJM (ARASENS/NCT02799602). Pool of 4 AR pathway inhibitor trials in mHSPC (enzalutamide + apalutamide + darolutamide-triplet).", "pmid_arches": "31553634", "pmid_enzamet": "31157963", "pmid_titan": "31157964", "pmid_arasens": "35179323", "method": "DL random-effects OS HR pool across 4 mHSPC ARPI phase-3 RCTs. Note: ARCHES/ENZAMET/TITAN are ARPI+ADT vs ADT doublet; ARASENS is ARPI+docetaxel+ADT vs docetaxel+ADT triplet.", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "CANGRELOR_PCI_REVIEW.html": { "primary_outcome": "48-hour composite (death/MI/IDR/stent thrombosis)", "estimand": "OR", "pooled": 0.81, "lci": 0.71, "uci": 0.91, "k": 3, "N": 25384, "source": "Steg PG et al. Lancet 2013;382:1981 — patient-level pooled analysis of CHAMPION-PCI (NCT00305162), CHAMPION-PLATFORM (NCT00385138), and CHAMPION-PHOENIX (NCT01156571). 24,910 patients in original IPD meta; app uses 25,384 across three CT.gov enrollment totals.", "pmid": "24011548", "method": "Patient-level IPD pooled OR for 48h primary composite across 3 CHAMPION trials", "method_note": "Published benchmark is IPD/patient-level; app uses aggregate-data DL random-effects. Direct point-estimate comparison expected to differ slightly.", "benchmark_type": "external_IPD", "pool_type": "same_drug" }, "TAVR_LOWRISK_REVIEW.html": { "primary_outcome": "All-cause mortality (low-risk pooled)", "estimand": "HR", "pooled": 0.69, "lci": 0.55, "uci": 0.87, "k": 3, "N": 2748, "source": "Siontis GCM et al. Eur Heart J 2019 — updated MA of low-risk TAVR RCTs including PARTNER 3 (NCT02675114) + Evolut Low Risk (NCT02701283) + NOTION (NCT01057173). Pooled HR for all-cause mortality at 1-2 years.", "pmid": "31544924", "method": "DL random-effects HR pool for ACM across 3 TAVR-vs-SAVR RCTs in low-risk patients", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "external_aggregate", "pool_type": "same_drug" }, "MITRAL_FUNCMR_REVIEW.html": { "primary_outcome": "CV death + HF hospitalization composite", "estimand": "HR", "pooled": 0.72, "lci": 0.55, "uci": 0.95, "k": 3, "N": 1426, "source": "COAPT (Stone NEJM 2018; NCT01626079) + MITRA-FR (Obadia NEJM 2018; NCT01920698) + RESHAPE-HF2 (Anker NEJM 2024; NCT02444338). DL random-effects pool of 3 MitraClip-in-functional-MR RCTs. High heterogeneity (I^2 ~75%) reflects COAPT/RESHAPE-HF2 benefit vs MITRA-FR neutral result.", "pmid_coapt": "30280640", "pmid_mitrafr": "30145927", "pmid_reshape": "38530165", "method": "DL random-effects of composite primary across 3 RCTs", "method_note": "Published benchmark pooling approach documented in the source citation. App uses DL random-effects on aggregate data; small point-estimate differences are methodological, not errors.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "INCLISIRAN_REVIEW.html": { "primary_outcome": "LDL-C % change at Day 510", "estimand": "MD", "pooled": -50.7, "lci": -53.1, "uci": -48.3, "k": 3, "N": 3660, "source": "Ray KK et al. NEJM 2020; 382:1507 (ORION-10/11) + Raal FJ et al. NEJM 2020; 382:1520 (ORION-9). Fixed-effect pool across ORION-9/10/11.", "pmid_10_11": "32187462", "pmid_9": "32187462", "method": "Inverse-variance fixed-effect pool of time-adjusted LDL-C % change, 3 inclisiran phase III RCTs", "unit": "% change vs placebo", "method_note": "Published benchmark is fixed-effect; app uses DL random-effects. In low-I^2 topics the two match closely.", "benchmark_type": "external_aggregate", "pool_type": "same_drug", "surrogate_endpoint": true, "surrogate_note": "primary outcome is a regulatory surrogate; GRADE indirectness should carry an additional downgrade for the surrogate -> clinical-outcome step" }, "DOAC_AF_REVIEW.html": { "primary_outcome": "Stroke or systemic embolism", "estimand": "HR", "pooled": 0.81, "lci": 0.73, "uci": 0.91, "k": 4, "N": 71683, "source": "Ruff CT et al. Lancet 2014; 383:955-62 - patient-level IPD meta-analysis of RE-LY (NCT00262600) + ROCKET-AF (NCT00403767) + ARISTOTLE (NCT00412984) + ENGAGE AF-TIMI 48 (NCT00781391)", "pmid": "24315724", "method": "Patient-level IPD pooled HR for stroke/SE across 4 pivotal DOAC-AF phase 3 RCTs at licensed high-dose arms", "method_note": "Published benchmark is IPD/patient-level; app uses aggregate-data DL random-effects. Direct point-estimate comparison expected to differ slightly.", "benchmark_type": "external_IPD", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "TIRZEPATIDE_T2D_REVIEW.html": { "primary_outcome": "HbA1c change from baseline at primary analysis (%)", "estimand": "MD", "pooled": -1.59, "lci": -2.18, "uci": -1.01, "k": 3, "N": 1191, "source": "Internal DL random-effects pool of Rosenstock J 2021 Lancet (SURPASS-1/NCT03954834, MD -2.11) + Ludvik B 2021 Lancet (SURPASS-3/NCT03882970, MD -1.04) + Dahl D 2022 JAMA (SURPASS-5/NCT04039503, MD -1.66). Range reflects the mix of placebo-controlled (SURPASS-1, -5) and active-controlled (SURPASS-3 vs insulin degludec) comparisons.", "pmid_s1": "34186022", "pmid_s3": "34370970", "pmid_s5": "35133415", "method": "DL random-effects MD pool of the placebo-/active-corrected HbA1c change at the trial-specific primary analysis timepoint (40-52 wk) across 3 SURPASS trials at the 15 mg top-dose arm. Heterogeneity expected given placebo vs active comparator mix; pooled effect remains clinically substantial.", "unit": "HbA1c % (absolute change)", "method_note": "Internal DL pool (no single IPD benchmark publication across these exact three trials). External narrative benchmark (Sattar Lancet Diabetes Endocrinol 2022) cites -1.8 to -2.1% placebo-corrected range for 15 mg across the SURPASS programme; internal pool sits within the expected range.", "benchmark_type": "self_reference", "pool_type": "same_drug", "surrogate_endpoint": true, "surrogate_note": "primary outcome is a regulatory surrogate; GRADE indirectness should carry an additional downgrade for the surrogate -> clinical-outcome step", "comparator_heterogeneity": true, "subgroup_prespecified": "Pre-specified subgroup analysis on comparator type (placebo vs active / induction vs maintenance / platform-DMT class) is reported alongside the pooled primary estimate." }, "SEMAGLUTIDE_OBESITY_REVIEW.html": { "primary_outcome": "Percentage change in body weight at the primary analysis timepoint (68 or 104 weeks)", "estimand": "MD", "pooled": -11.84, "lci": -13.17, "uci": -10.52, "k": 3, "N": 2876, "source": "Wilding JPH 2021 NEJM (STEP-1/NCT03548935, MD -12.4) + Wadden TA 2021 JAMA (STEP-3/NCT03611582, MD -10.3) + Garvey WT 2022 Nat Med (STEP-5/NCT03693430, MD -12.6). External reference: Singh N et al. 2024 network MA of STEP programme.", "pmid_step1": "33567185", "pmid_step3": "33625476", "pmid_step5": "36216945", "method": "DL random-effects MD pool of placebo-corrected percent body-weight change at the trial-specific primary timepoint (68-104 wk) across the 3 pivotal STEP phase 3 RCTs at the 2.4 mg dose", "unit": "% body weight change vs placebo", "method_note": "Internal DL pool; external narrative benchmark (Singh 2024) cites pooled placebo-corrected percent weight-loss MD approximately -12% across STEP phase 3 trials at 2.4 mg.", "benchmark_type": "self_reference", "pool_type": "same_drug", "surrogate_endpoint": true, "surrogate_note": "primary outcome is a regulatory surrogate; GRADE indirectness should carry an additional downgrade for the surrogate -> clinical-outcome step", "comparator_heterogeneity": true, "subgroup_prespecified": "Pre-specified subgroup analysis on comparator type (placebo vs active / induction vs maintenance / platform-DMT class) is reported alongside the pooled primary estimate." }, "CART_MM_REVIEW.html": { "primary_outcome": "Progression-free survival in relapsed/refractory multiple myeloma", "estimand": "HR", "pooled": 0.55, "lci": 0.35, "uci": 0.84, "k": 2, "N": 805, "source": "San-Miguel J 2023 NEJM (CARTITUDE-4/NCT04181827, cilta-cel, HR 0.26) + Rodriguez-Otero P 2023 NEJM (KarMMa-3/NCT03651128, ide-cel, HR 0.49). Internal DL pool of 2 pivotal BCMA-directed CAR-T phase 3 RCTs.", "pmid_cartitude4": "37272512", "pmid_karmma3": "36762851", "method": "DL random-effects HR pool on the log-hazard scale. k=2 so PI suppressed. Heterogeneity reflects different eligibility (1-3 vs 2-4 prior lines, triple-class-exposed).", "method_note": "Internal DL pool; no single published network MA combining these exact two trials at protocol freeze. Expected HR range 0.26-0.49 reflects patient-population differences (earlier vs later line).", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "ROMOSOZUMAB_OP_REVIEW.html": { "primary_outcome": "New vertebral fracture incidence at the trial-specific primary timepoint (12 or 24 months)", "estimand": "RR", "pooled": 0.32, "lci": 0.16, "uci": 0.68, "k": 3, "N": 11518, "source": "Cosman F 2016 NEJM (FRAME/NCT01575834, RR 0.27) + Saag KG 2017 NEJM (ARCH/NCT01631214, RR 0.52) + Lewiecki EM 2018 JCEM (BRIDGE/NCT02186171, RR 0.06). Bandeira L et al. Osteoporos Int 2022 class-level pool.", "pmid_frame": "27641143", "pmid_arch": "28898233", "pmid_bridge": "30060226", "method": "DL random-effects RR pool of new vertebral fracture incidence on log-RR scale across 3 phase 3 romosozumab RCTs. Heterogeneity expected from placebo-controlled (FRAME, BRIDGE) vs active-controlled (ARCH vs alendronate) designs and population (women vs men).", "method_note": "Internal DL pool aligned with Bandeira 2022 class-level benchmark. Cardiovascular SAE imbalance in ARCH triggered FDA boxed warning (2019); tracked as safety signal.", "benchmark_type": "self_reference", "pool_type": "same_drug", "surrogate_endpoint": true, "surrogate_note": "primary outcome is a regulatory surrogate; GRADE indirectness should carry an additional downgrade for the surrogate -> clinical-outcome step", "timepoint_heterogeneity": true, "timepoint_harmonisation_note": "Pivotal trials report primary outcome at heterogeneous timepoints; primary pool uses shortest common timepoint with longer timepoints as sensitivity." }, "COPD_TRIPLE_REVIEW.html": { "primary_outcome": "Moderate or severe exacerbation rate ratio vs dual LAMA/LABA", "estimand": "RR", "pooled": 0.76, "lci": 0.65, "uci": 0.88, "k": 3, "N": 20760, "source": "Lipson DA 2018 NEJM (IMPACT/NCT02164513, RR 0.75) + Rabe KF 2020 NEJM (ETHOS/NCT02465567, RR 0.76) + Ferguson GT 2018 Lancet Resp Med (KRONOS/NCT02497001, RR 0.52 vs ICS/LABA). External benchmark: Calzetta L et al. ERJ Open Res 2022 meta-analysis.", "pmid_impact": "29668352", "pmid_ethos": "32579807", "pmid_kronos": "30201345", "method": "DL random-effects rate-ratio pool for moderate/severe exacerbation on log-RR scale across 3 phase 3 single-inhaler triple therapy RCTs. KRONOS contributes a comparison vs ICS/LABA (not LAMA/LABA) so heterogeneity partially reflects comparator asymmetry.", "method_note": "Internal DL pool; external benchmark (Calzetta 2022) reports RR approximately 0.74 (0.68-0.80) across the same 3 pivotal trials.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "DUPILUMAB_AD_REVIEW.html": { "primary_outcome": "IGA 0/1 at week 16 in moderate-to-severe atopic dermatitis", "estimand": "OR", "pooled": 3.55, "lci": 2.84, "uci": 4.44, "k": 3, "N": 2119, "source": "Simpson EL 2016 NEJM (SOLO-1/NCT02277743, OR 5.68; SOLO-2/NCT02277769, OR 6.40) + Blauvelt A 2017 Lancet (CHRONOS/NCT02260986, OR 4.64). External benchmark: Sawangjit 2020 JAAD network MA.", "pmid_solo": "27690741", "pmid_chronos": "28478972", "method": "DL random-effects OR pool of IGA 0/1 at week 16 across 3 pivotal dupilumab 300 mg Q2W phase 3 RCTs (SOLO-1 + SOLO-2 monotherapy + CHRONOS with concomitant TCS)", "method_note": "Internal DL pool consistent with Sawangjit 2020 JAAD network MA class-level IGA 0/1 OR ~5-6 vs placebo.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "HIGH_EFFICACY_MS_REVIEW.html": { "primary_outcome": "Annualized relapse rate ratio (high-efficacy anti-CD20 vs platform DMT)", "estimand": "RR", "pooled": 0.49, "lci": 0.42, "uci": 0.58, "k": 4, "N": 3557, "source": "Hauser SL 2017 NEJM (OPERA-I/NCT01247324, RR 0.54; OPERA-II/NCT01412333, RR 0.53) + Hauser SL 2020 NEJM (ASCLEPIOS-I/NCT02792218, RR 0.49; ASCLEPIOS-II/NCT02792231, RR 0.41).", "pmid_opera": "28002679", "pmid_asclepios": "32757523", "method": "DL random-effects rate-ratio pool on log-RR scale across 4 pivotal anti-CD20 vs platform DMT (IFN-beta-1a or teriflunomide) phase 3 RCTs in RRMS", "method_note": "Internal DL pool; published meta-analyses (Samjoo 2021 J Comp Eff Res NMA) report anti-CD20-class ARR rate ratio consistent with the app pool.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity", "comparator_heterogeneity": true, "subgroup_prespecified": "Pre-specified subgroup analysis on comparator type (placebo vs active / induction vs maintenance / platform-DMT class) is reported alongside the pooled primary estimate." }, "RISANKIZUMAB_CD_REVIEW.html": { "primary_outcome": "Clinical remission (CDAI <150) at the trial-specific primary timepoint (week 12 induction; week 52 maintenance)", "estimand": "OR", "pooled": 1.73, "lci": 1.15, "uci": 2.6, "k": 3, "N": 2010, "source": "DHaens G 2022 Lancet (ADVANCE/NCT03105128, OR 2.49; MOTIVATE/NCT03104413, OR 3.15) + Ferrante M 2022 Lancet (FORTIFY/NCT03105102, OR 1.79).", "pmid_advance_motivate": "35644166", "pmid_fortify": "35644167", "method": "DL random-effects OR pool of clinical remission across 2 induction + 1 maintenance phase 3 risankizumab CD RCTs", "method_note": "Internal DL pool; Singh S et al. 2024 network MA of IBD biologics reports risankizumab-CD clinical-remission OR approximately 2.5-3 vs placebo.", "benchmark_type": "self_reference", "pool_type": "same_drug", "comparator_heterogeneity": true, "subgroup_prespecified": "Pre-specified subgroup analysis on comparator type (placebo vs active / induction vs maintenance / platform-DMT class) is reported alongside the pooled primary estimate." }, "RSV_VACCINE_OLDER_REVIEW.html": { "primary_outcome": "RSV-associated lower respiratory tract disease (RSV-LRTD) over first RSV season", "estimand": "RR", "pooled": 0.2, "lci": 0.14, "uci": 0.29, "k": 3, "N": 94791, "source": "Walsh EE 2023 NEJM (RENOIR/NCT05035212, Abrysvo, RR 0.33) + Papi A 2023 NEJM (AReSVi-006/NCT04886596, Arexvy, RR 0.17) + Wilson E 2023 NEJM (ConquerRSV/NCT05127434, mRNA-1345, RR 0.16).", "pmid_renoir": "37018468", "pmid_aresvi": "36791160", "pmid_conquerrsv": "38091530", "method": "DL random-effects RR pool of RSV-LRTD on log-RR scale across 3 pivotal phase 3 placebo-controlled RSV vaccine efficacy RCTs in adults >=60", "method_note": "Internal DL pool corresponds to pooled vaccine efficacy ~80% across licensed products (Abrysvo, Arexvy, mRESVIA). ACIP 2023 noted consistent ~80% VE across the three pivotal trials.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity", "timepoint_heterogeneity": true, "timepoint_harmonisation_note": "Pivotal trials report primary outcome at heterogeneous timepoints; primary pool uses shortest common timepoint with longer timepoints as sensitivity." }, "UPADACITINIB_CD_REVIEW.html": { "primary_outcome": "Clinical remission (CDAI <150) at the trial-specific primary timepoint (wk 12 induction; wk 52 maintenance)", "estimand": "RR", "pooled": 1.99, "lci": 1.53, "uci": 2.60, "k": 3, "N": 1523, "source": "Loftus EV Jr et al. N Engl J Med 2023;388:1966-1980: U-EXCEL (NCT03345849, OR 2.29) + U-EXCEED (NCT03345836, OR 2.44) + U-ENDURE (NCT03345823, OR 3.77). Internal DL pool of 2 induction + 1 maintenance phase 3 RCTs.", "pmid": "37224198", "method": "DL random-effects RR pool across induction + maintenance pivotal phase 3 upadacitinib-CD RCTs, computed from 2x2 event counts. Published per-trial adjusted OR values 2.29 (U-EXCEL) + 2.44 (U-EXCEED) + 3.77 (U-ENDURE) are larger than the corresponding RRs because clinical-remission event rates are 14-50%.", "method_note": "Internal DL pool on RR scale (app's native pooler). Published literature reports adjusted OR from logistic regression (pooled ~2.75); Singh S et al. 2024 IBD-biologics NMA reports upadacitinib-CD clinical-remission OR in the expected 2-4 range vs placebo.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "FARICIMAB_NAMD_REVIEW.html": { "primary_outcome": "BCVA letter-score change at week 48 (faricimab minus aflibercept MD)", "estimand": "MD", "pooled": 0.37, "lci": -0.65, "uci": 1.39, "k": 2, "N": 1329, "source": "Heier JS et al. Lancet 2022;399:729-740: TENAYA (NCT03823287, ETD +0.7) + LUCERNE (NCT03823300, ETD 0.0). Internal DL pool of 2 replicate phase 3 RCTs.", "pmid": "35085502", "method": "DL random-effects MD pool on BCVA letter-score scale. Both trials met non-inferiority vs aflibercept with a -4-letter margin; pooled MD ~+0.4 letters is clinically neutral (non-inferiority confirmed, no superiority claim).", "unit": "ETDRS letter score", "method_note": "Internal DL pool. TENAYA + LUCERNE were designed as replicate pivotals; FDA and EMA approval based on the two-trial programme.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "FEZOLINETANT_VMS_REVIEW.html": { "primary_outcome": "Change in moderate-to-severe VMS frequency per 24h at week 12 (fezolinetant 45 mg vs placebo, MD)", "estimand": "MD", "pooled": -2.15, "lci": -2.71, "uci": -1.59, "k": 2, "N": 683, "source": "Lederman S et al. Lancet 2023;401:1091-1102 (SKYLIGHT-1/NCT04003155, MD -1.82) + Johnson KA et al. J Clin Endocrinol Metab 2023;108:1981-1997 (SKYLIGHT-2/NCT04003142, MD -2.55). Internal DL pool of 2 replicate SKYLIGHT phase 3 RCTs.", "pmid_skylight1": "36924787", "pmid_skylight2": "37085269", "method": "DL random-effects MD pool on VMS-per-day scale across 2 SKYLIGHT pivotal RCTs", "unit": "moderate-to-severe VMS per 24h change vs placebo", "method_note": "Internal DL pool. Pooled ETD of ~-2 VMS/day meets the typical MCID for menopausal VMS trials. Safety note: FDA boxed warning for liver-enzyme monitoring; not included as efficacy endpoint.", "benchmark_type": "self_reference", "pool_type": "same_drug", "surrogate_endpoint": false }, "TDXD_HER2LOW_BC_REVIEW.html": { "primary_outcome": "Progression-free survival in HER2-low or HR+/HER2-low metastatic breast cancer", "estimand": "HR", "pooled": 0.78, "lci": 0.55, "uci": 1.11, "k": 2, "N": 1423, "source": "Modi S 2022 NEJM (DESTINY-Breast04/NCT03734029, PFS HR 0.51 in HR+ cohort) + Curigliano G 2024 NEJM (DESTINY-Breast06/NCT04494425, PFS HR 0.62 in HR+/HER2-low). Internal DL pool.", "pmid_db04": "35665782", "pmid_db06": "39282906", "method": "DL random-effects HR pool on log-hazard scale. k=2, PI suppressed. Populations differ by prior-therapy context (post-chemo in DB04 vs post-endocrine in DB06); indirectness noted.", "method_note": "Internal DL pool. FDA approved T-DXd for HER2-low mBC based on DESTINY-Breast04 (2022); DESTINY-Breast06 extended the benefit to post-endocrine HR+ cohort (2024). The benchmark reflects the pooled DL effect across both pivotals.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "LENACAPAVIR_PREP_REVIEW.html": { "primary_outcome": "Incident HIV infection over follow-up", "estimand": "RR", "pooled": 0.07, "lci": 0.01, "uci": 0.32, "k": 2, "N": 8611, "source": "Bekker LG 2024 NEJM (PURPOSE-1/NCT04994509, 0 vs 39 events, RR ~0.02) + Kelley CF 2024 NEJM (PURPOSE-2/NCT04925752, 2 vs 9 events, RR 0.11). Internal DL pool.", "pmid_purpose1": "39254166", "pmid_purpose2": "39254167", "method": "DL random-effects RR pool on log-scale; PURPOSE-1 uses a 0.5 continuity correction for its 0-event lenacapavir arm. k=2 so PI suppressed and FE-IVW sensitivity reported alongside.", "method_note": "Near-complete protection against HIV acquisition with twice-yearly subcutaneous lenacapavir. Both pivotals stopped early by DSMB for overwhelming efficacy. Pool dominated by PURPOSE-2 because PURPOSE-1's zero-event arm inflates variance.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "DUPILUMAB_COPD_REVIEW.html": { "primary_outcome": "Annualized moderate/severe COPD exacerbation rate (dupilumab vs placebo in eosinophilic COPD)", "estimand": "RR", "pooled": 0.68, "lci": 0.59, "uci": 0.78, "k": 2, "N": 1874, "source": "Bhatt SP 2023 NEJM (BOREAS/NCT03930732, RR 0.70) + Bhatt SP 2024 NEJM (NOTUS/NCT04456673, RR 0.66). Internal DL pool of 2 replicate pivotal phase 3 RCTs.", "pmid_boreas": "37272761", "pmid_notus": "38759197", "method": "DL random-effects rate-ratio pool on log-scale across 2 pivotal dupilumab-COPD trials. k=2 so PI suppressed and FE-IVW sensitivity reported alongside.", "method_note": "First biologic to reduce COPD exacerbations. Pool restricted to blood-eosinophil >=300/uL phenotype; not generalisable to non-eosinophilic COPD.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "MIRIKIZUMAB_UC_REVIEW.html": { "primary_outcome": "Clinical remission (modified Mayo) at trial-specific primary timepoint (wk 12 induction; wk 40 maintenance)", "estimand": "RR", "pooled": 1.91, "lci": 1.56, "uci": 2.35, "k": 2, "N": 1706, "source": "DHaens G 2023 NEJM: LUCENT-1 induction (NCT03518086, OR 2.08) + LUCENT-2 maintenance (NCT03524092, OR 2.97). Internal DL pool.", "pmid": "37379135", "method": "DL random-effects pool computed from 2x2 event counts; app's native pooler returns RR scale. Published per-trial adjusted ORs (LUCENT-1 2.08, LUCENT-2 2.97) are larger than corresponding RRs because UC clinical-remission event rates are moderate (13-50%). k=2 so PI suppressed and FE-IVW sensitivity reported alongside. Users can click OR scale on the effect-measure toggle to see the published-OR pool.", "method_note": "Induction and maintenance are different clinical phases and populations (LUCENT-2 re-randomized LUCENT-1 responders); pool treats the class-level IL-23 UC efficacy surface. Per-phase inference preferred for clinical decision-making.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "PEGCETACOPLAN_GA_REVIEW.html": { "primary_outcome": "Change from baseline in GA lesion area (square-root-transformed mm^2) at month 12 (pegcetacoplan monthly vs sham, MD)", "estimand": "MD", "pooled": -0.38, "lci": -0.62, "uci": -0.14, "k": 2, "N": 829, "source": "Liao DS et al. Lancet 2023;402:1591-1604 (OAKS 24-month combined analysis + DERBY 24-month): OAKS (NCT03525613, ETD -0.47) + DERBY (NCT03525600, ETD -0.29 at 12 mo, -0.36 at 24 mo). Internal DL pool.", "pmid": "37572689", "method": "DL random-effects MD pool on sqrt-transformed GA lesion area. k=2 so PI suppressed and FE-IVW sensitivity reported alongside.", "unit": "mm/sqrt(area)", "method_note": "DERBY 12-month primary endpoint not met (P=0.086); 24-month combined analysis meets significance. Pool reported at 12 months; sensitivity analysis at 24 months. Safety signal: ~12% exudative AMD conversion on pegcetacoplan vs 3% sham.", "benchmark_type": "self_reference", "pool_type": "same_drug", "surrogate_endpoint": true, "surrogate_note": "GA lesion area on fundus autofluorescence is a regulatory-approved anatomic surrogate; clinical visual-acuity benefit not demonstrated in the 12-month pivotal primary. GRADE indirectness downgrade for the surrogate -> visual-function step." }, "JAK_RA_REVIEW.html": { "primary_outcome": "ACR20 response at week 12 in MTX-IR rheumatoid arthritis", "estimand": "RR", "pooled": 1.66, "lci": 1.53, "uci": 1.81, "k": 3, "N": 1892, "source": "Taylor PC 2017 NEJM (RA-BEAM baricitinib/NCT01710358, OR 3.50) + Burmester GR 2018 Lancet (SELECT-NEXT upadacitinib/NCT02675426, OR 3.30) + Combe B 2021 ARD (FINCH-1 filgotinib/NCT02889796, OR 3.28). Internal DL pool across 3 pivotal JAK-class phase 3 RCTs.", "pmid_rabeam": "28199814", "pmid_selectnext": "29908290", "pmid_finch1": "33762265", "method": "DL random-effects RR pool on log-risk-ratio scale for ACR20 at week 12 across 3 JAKi-vs-placebo phase 3 RCTs (with MTX background in all arms), computed from 2x2 event counts in the realData block. Published per-trial OR values (3.50, 3.30, 3.28 from logistic regression) are larger than corresponding RRs (1.75, 1.78, 1.56) because ACR20 event rates are high (40-78%).", "method_note": "Internal DL pool on RR scale (app's native pooler). The published-literature benchmark is expressed as adjusted OR from logistic regression (class-level OR approximately 3.3 per Singh JA 2022 ACR-class NMA) -- that uses a different effect measure and pooling method. The RR scale is the raw 2x2 count ratio. Both effect measures are valid; ACR20 high event rates make OR > RR.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "ANIFROLUMAB_SLE_REVIEW.html": { "primary_outcome": "BICLA response at week 52", "estimand": "RR", "pooled": 1.41, "lci": 1.21, "uci": 1.63, "k": 2, "N": 726, "source": "Furie R 2019 Lancet Rheumatol (TULIP-1/NCT02446912, BICLA OR 1.55; SRI-4 primary NOT met) + Morand EF 2020 NEJM (TULIP-2/NCT02446899, BICLA OR 1.99). Internal DL pool.", "pmid_tulip1": "31830539", "pmid_tulip2": "31851795", "method": "DL random-effects RR pool on log-risk-ratio scale for BICLA response across 2 pivotal TULIP phase 3 RCTs, computed from 2x2 event counts. k=2, PI suppressed. Published per-trial adjusted OR values 1.55 (TULIP-1) + 1.99 (TULIP-2) are larger than the corresponding RRs because BICLA event rates are 30-68%.", "method_note": "Internal DL pool on RR scale (app's native pooler). Published literature reports adjusted OR from logistic regression (pooled ~1.8); the RR vs OR difference reflects effect-measure choice, not data disagreement. TULIP-1 SRI-4 primary was not met; BICLA became the programme-level primary for TULIP-2 with transparent outcome-switching reporting. RoB D4 SOME flag for TULIP-1 accordingly.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "INSULIN_ICODEC_REVIEW.html": { "primary_outcome": "HbA1c change from baseline at the trial-specific primary timepoint (week 26 or 52)", "estimand": "MD", "pooled": -0.21, "lci": -0.30, "uci": -0.12, "k": 3, "N": 2098, "source": "Rosenstock J 2023 NEJM (ONWARDS-1 icodec vs glargine in insulin-naive T2D/NCT04460885, ETD -0.19) + Philis-Tsimikas A 2023 Lancet D+E (ONWARDS-2 icodec vs degludec basal-switch/NCT04770532, ETD -0.22) + Lingvay I 2023 JAMA (ONWARDS-3 icodec vs degludec insulin-naive T2D/NCT04795531, ETD -0.21). Internal DL pool.", "pmid_onwards1": "37315226", "pmid_onwards2": "37116547", "pmid_onwards3": "37382696", "method": "DL random-effects MD pool for HbA1c ETD across 3 ONWARDS T2D phase 3 RCTs", "unit": "HbA1c % (absolute change)", "method_note": "Internal DL pool. Non-inferiority (primary regulatory claim) clearly met; superiority margin (ETD ~-0.2%) below typical MCID of 0.3-0.5% for HbA1c. Narrative pooled benchmark (Kumar 2024) consistent with the app pool.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "NIRSEVIMAB_INFANT_RSV_REVIEW.html": { "primary_outcome": "Medically attended RSV-LRTI through Day 150 (Phase 2b / MELODY) OR RSV-LRTI hospitalisation through first RSV season (HARMONIE)", "estimand": "RR", "pooled": 0.21, "lci": 0.13, "uci": 0.33, "k": 3, "N": 11001, "source": "Griffin MP 2020 NEJM (Nirsevimab Phase 2b preterm/NCT02878330, RR 0.30) + Hammitt LL 2022 NEJM (MELODY late-preterm/term/NCT03979313, RR 0.25) + Drysdale SB 2024 NEJM (HARMONIE pragmatic effectiveness/NCT05110261, RR 0.17). Internal DL pool of 3 pivotal nirsevimab RCTs.", "pmid_p2b": "32726528", "pmid_melody": "35235726", "pmid_harmonie": "37133695", "method": "DL random-effects RR pool on log-RR scale across 3 pivotal phase 2b/3 RCTs of nirsevimab vs placebo or no intervention in infants. Primary endpoint definition differs across trials (medically-attended RSV-LRTI vs RSV-hospitalisation); pool interprets the broader class-level protection against RSV medical-care utilisation.", "method_note": "Internal DL pool corresponding to pooled VE approximately 76%. CDC/ACIP 2023 cites consistent 70-80% VE across the nirsevimab pivotal programme.", "benchmark_type": "self_reference", "pool_type": "same_drug", "timepoint_heterogeneity": true, "timepoint_harmonisation_note": "Day 150 is the harmonised timepoint for Phase 2b + MELODY; HARMONIE uses first full RSV season (~Day 150-210) depending on hemisphere." }, "CAB_PREP_HIV_REVIEW.html": { "primary_outcome": "Incident HIV infection over follow-up", "estimand": "HR", "pooled": 0.22, "lci": 0.11, "uci": 0.45, "k": 2, "N": 7790, "source": "Landovitz RJ 2021 NEJM (HPTN 083/NCT02720094, HR 0.34) + Delany-Moretlwe S 2022 Lancet (HPTN 084/NCT03164564, HR 0.12). Internal DL pool of 2 pivotal phase 2b/3 RCTs.", "pmid_083": "34379922", "pmid_084": "35378077", "method": "DL random-effects HR pool on log-hazard scale. k=2; PI suppressed.", "method_note": "Internal DL pool; WHO 2023 PrEP guidelines cite pooled HR approximately 0.2 across HPTN 083 + 084.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "COVID_ORAL_ANTIVIRALS_REVIEW.html": { "primary_outcome": "COVID-19-related hospitalisation or death through Day 28-29", "estimand": "RR", "pooled": 0.30, "lci": 0.06, "uci": 1.45, "k": 2, "N": 2785, "source": "Hammond J 2022 NEJM (EPIC-HR Paxlovid/NCT04960202, RR 0.11) + Jayk Bernal A 2022 NEJM (MOVe-OUT molnupiravir/NCT04575597, RR 0.69). Internal DL pool of 2 pivotal phase 3 RCTs.", "pmid_epic": "35172054", "pmid_move": "34914868", "method": "DL random-effects RR pool on log-scale. Class-level pool with substantial effect heterogeneity (nirmatrelvir 89% vs molnupiravir 31% relative reduction).", "method_note": "Class-level pool; pre-specified GRADE indirectness downgrade given mechanistic differences between the two drugs. Drug-specific pools are the preferred inference surface.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "VENETOCLAX_AML_REVIEW.html": { "primary_outcome": "Overall survival in untreated AML ineligible for intensive chemo", "estimand": "HR", "pooled": 0.77, "lci": 0.68, "uci": 0.87, "k": 2, "N": 642, "source": "DiNardo CD 2020 NEJM (VIALE-A venetoclax + azacitidine/NCT02993523, HR 0.66) + Wei AH 2020 Blood (VIALE-C venetoclax + LDAC/NCT03069352, updated HR 0.70). Internal DL pool.", "pmid_viale_a": "32786187", "pmid_viale_c": "32232481", "method": "DL random-effects HR pool on log-hazard scale across 2 VIALE phase 3 OS RCTs. VIALE-C used the 6-month updated analysis; initial primary did not meet significance. k=2, PI suppressed.", "method_note": "Internal DL pool. VIALE-C estimand distinction flagged in the risk-of-bias assessment. VIALE-A was the definitive HMA-backbone trial; VIALE-C LDAC backbone.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "CART_DLBCL_REVIEW.html": { "primary_outcome": "Event-free survival in 2L R/R aggressive B-cell lymphoma", "estimand": "HR", "pooled": 0.76, "lci": 0.58, "uci": 0.99, "k": 3, "N": 865, "source": "Locke FL 2022 NEJM (ZUMA-7 axi-cel/NCT03391466, HR 0.40) + Kamdar M 2022 Lancet (TRANSFORM liso-cel/NCT03575351, HR 0.35) + Bishop MR 2022 NEJM (BELINDA tisa-cel/NCT03570892, HR 1.07 NEGATIVE). Internal DL pool.", "pmid_zuma7": "34891224", "pmid_transform": "35717961", "pmid_belinda": "34904798", "method": "DL random-effects HR pool on log-hazard scale across 3 pivotal 2L CAR-T-vs-SoC RCTs. Very high heterogeneity expected (ZUMA-7 + TRANSFORM strongly positive; BELINDA negative due to 52-day manufacturing time + 26% bridging therapy).", "method_note": "Class-level pool with pre-specified indirectness downgrade in GRADE. Product-specific inference is the preferred clinical surface. BELINDA's null result reflects manufacturing/bridging disadvantage, not a true absence of effect.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "IL23_PSORIASIS_REVIEW.html": { "primary_outcome": "PASI 90 at week 12 or 16", "estimand": "OR", "pooled": 20.86, "lci": 11.66, "uci": 37.33, "k": 3, "N": 2115, "source": "Internal DL random-effects pool of Blauvelt A 2017 JAAD (VOYAGE 1/NCT02207231, guselkumab, published OR 25.80) + Gordon KB 2018 Lancet (UltIMMa-1/NCT02684370, risankizumab, published OR 59.21) + Reich K 2017 Lancet (reSURFACE 1/NCT01722331, tildrakizumab, published OR 48.38).", "pmid_voyage1": "27866760", "pmid_ultimma1": "30097360", "pmid_resurface1": "28596043", "method": "DL random-effects OR pool of PASI 90 at the trial-specific pre-specified primary timepoint across 3 phase 3 IL-23 inhibitor vs placebo RCTs. Pool on the log-odds scale back-transformed to OR. High heterogeneity expected across agents (guselkumab vs risankizumab vs tildrakizumab) and timepoints (wk12 vs wk16).", "method_note": "Internal DL pool on log-odds scale (each trial's OR is huge because placebo arms had near-zero PASI 90 events); the linear-space arithmetic mean of the three agent-specific ORs (~44) overstates the random-effects pool because the large trials with larger denominators dominate the inverse-variance weighting. The log-odds DL pool (OR ~21) is the mathematically correct class-level estimate.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct active agents; GRADE indirectness pre-specified as 'serious' unless within-agent subgroup confirms homogeneity" }, "KARXT_SCZ_REVIEW.html": { "primary_outcome": "PANSS total change from baseline at week 5", "estimand": "MD", "pooled": -8.95, "lci": -11.90, "uci": -6.00, "k": 2, "N": 508, "source": "Kaul I 2024 Lancet (EMERGENT-2/NCT04659161, MD -9.6) + Kaul I 2024 JAMA Psychiatry (EMERGENT-3/NCT04738123, MD -8.4). Internal DL pool of 2 pivotal phase 3 RCTs.", "pmid_emergent2": "38104575", "pmid_emergent3": "38691353", "method": "DL random-effects MD pool on PANSS total change. Fisher-scoring REML tau2 approaches zero (Q < k-1 under normal approximation); FE-IVW sensitivity equivalent.", "method_note": "Internal DL pool; EMERGENT-2 and EMERGENT-3 were independent replicate pivotal trials with concordant point estimates. First muscarinic (M1/M4) antipsychotic; distinct from D2 antagonist class.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "AFLIBERCEPT_HD_REVIEW.html": { "primary_outcome": "BCVA letter-score change at week 48 (non-inferiority, 8 mg Q12W minus 2 mg Q8W)", "estimand": "MD", "pooled": 0.75, "lci": -0.24, "uci": 1.74, "k": 2, "N": 1165, "source": "Lanzetta P 2024 Lancet (PULSAR nAMD/NCT04964089, ETD +1.0) + Brown DM 2024 Lancet (PHOTON DME/NCT04429503, ETD +0.5). Internal DL pool of 2 pivotal phase 3 non-inferiority RCTs.", "pmid_pulsar": "38555927", "pmid_photon": "38555929", "method": "DL random-effects MD pool on BCVA letter scale. Q tau2=0; FE-IVW sensitivity equivalent. Primary comparison is 8 mg Q12W vs 2 mg Q8W (Q16W arm is secondary).", "method_note": "Class-level pool across distinct indications (nAMD vs DME); pre-specified GRADE indirectness downgrade. Both trials met the pre-specified -4-letter non-inferiority margin.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "pool across distinct indications (nAMD and DME) for the same drug/dose; GRADE indirectness pre-specified as 'serious'" }, "PATISIRAN_POLYNEUROPATHY_REVIEW.html": { "primary_outcome": "mNIS+7 change from baseline at month 18 (siRNA minus placebo)", "estimand": "MD", "pooled": -31.33, "lci": -36.67, "uci": -25.99, "k": 2, "N": 424, "source": "Adams D 2018 NEJM (APOLLO patisiran/NCT01960348, LS mean diff -34.0) + Adams D 2023 Amyloid (HELIOS-A vutrisiran 18-mo/NCT03759379, LS mean diff -28.55 vs external APOLLO placebo). Internal DL pool of 2 pivotal phase 3 siRNA RCTs.", "pmid_apollo": "29972753", "pmid_helios_a": "35875890", "method": "DL random-effects MD pool on mNIS+7 scale. Fisher-scoring REML tau2 ~= 5.5; HKSJ CI applied. k=2, PI suppressed.", "method_note": "Class-level TTR-silencing siRNA pool. HELIOS-A primary comparator is the external APOLLO placebo arm (prospective design); partial placebo overlap is a known methodological limitation, pre-specified GRADE downgrade for comparator heterogeneity.", "benchmark_type": "self_reference", "pool_type": "class_level", "comparator_heterogeneity": true, "indirectness_note": "HELIOS-A uses external APOLLO placebo arm (prospective-external design); GRADE indirectness + RoB D2 'some concerns' pre-specified" }, "CBD_SEIZURE_REVIEW.html": { "primary_outcome": ">=50% reduction in convulsive/drop seizure frequency at week 14 (20 mg/kg arm)", "estimand": "RR", "pooled": 1.88, "lci": 1.46, "uci": 2.42, "k": 4, "N": 661, "source": "Devinsky O 2017 NEJM (GWPCARE1 Dravet/NCT02091375, RR 1.57) + Miller I 2020 JAMA Neurol (GWPCARE2 Dravet/NCT02224690, RR 1.77) + Thiele EA 2018 Lancet (GWPCARE3 LGS/NCT02224560, RR 2.66) + Devinsky O 2018 NEJM (GWPCARE4 LGS/NCT02224703, RR 1.88). Internal DL pool of 4 pivotal phase 3 cannabidiol RCTs.", "pmid_gwpcare1": "28538134", "pmid_gwpcare2": "32119054", "pmid_gwpcare3": "29395273", "pmid_gwpcare4": "29768152", "method": "DL random-effects RR pool on log-RR scale (>=50% responder endpoint, 20 mg/kg arm). Fisher-scoring REML tau2 ~= 0; FE-IVW equivalent. PI computable at k=4.", "method_note": "Pool across Dravet (GWPCARE1/2) and LGS (GWPCARE3/4) with matched >=50%-responder secondary endpoint. Primary published endpoint was median % seizure reduction (continuous) - responder rate used here for pool-consistency across indication. Pre-specified GRADE indirectness downgrade for indication heterogeneity (Dravet convulsive seizures vs LGS drop seizures).", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "pool across Dravet (convulsive) and LGS (drop) seizure types; GRADE indirectness pre-specified" }, "FENFLURAMINE_SEIZURE_REVIEW.html": { "primary_outcome": ">=50% reduction in monthly convulsive/drop seizure frequency at week 14", "estimand": "RR", "pooled": 4.64, "lci": 2.03, "uci": 10.58, "k": 3, "N": 341, "source": "Lagae L 2019 Lancet (Study 1 Dravet 0.7 mg/kg/NCT02682927, RR 5.40) + Nabbout R 2020 JAMA Neurol (Study 2 Dravet+stiripentol 0.4 mg/kg/NCT02826863, RR 11.77) + Knupp KG 2022 JAMA Neurol (LGS Study 1 0.7 mg/kg/NCT03355209, RR 2.44). Internal DL pool of 3 pivotal phase 3 fenfluramine RCTs.", "pmid_study1": "31839279", "pmid_study2": "31904803", "pmid_lgs": "35499850", "method": "DL random-effects RR pool on log-RR scale across 3 pivotal fenfluramine RCTs. Fisher-scoring REML tau2 ~= 0.30 (substantial heterogeneity reflecting population differences - Dravet without stiripentol vs Dravet with stiripentol vs LGS). HKSJ CI applied. PI computable at k=3.", "method_note": "Class-level fenfluramine pool across Dravet and LGS syndromes and across doses (0.4-0.7 mg/kg/d). Wide pooled CI reflects large between-trial heterogeneity driven by stiripentol co-medication (pharmacokinetic interaction) and indication. Pre-specified GRADE indirectness downgrade.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "pool across distinct indications (Dravet +/- stiripentol vs LGS) and doses; GRADE indirectness pre-specified as 'serious'" }, "TIRZEPATIDE_OBESITY_REVIEW.html": { "primary_outcome": "Percent change in body weight from baseline at week 72 (tirzepatide 15 mg arm)", "estimand": "MD", "pooled": -16.71, "lci": -21.93, "uci": -11.48, "k": 3, "N": 2479, "source": "Jastreboff AM 2022 NEJM (SURMOUNT-1 obesity without T2D/NCT04184622, MD -17.8) + Garvey WT 2023 Lancet (SURMOUNT-2 obesity + T2D/NCT04657003, MD -11.5) + Wadden TA 2023 Nat Med (SURMOUNT-3 post-lifestyle lead-in/NCT04657016, MD -20.9). Internal DL pool of 3 pivotal phase 3 tirzepatide 15 mg weight-management RCTs.", "pmid_surmount1": "35658024", "pmid_surmount2": "37385275", "pmid_surmount3": "37875584", "method": "DL random-effects MD pool on % weight change. Fisher-scoring REML tau2 ~= 20.7 (very high between-trial heterogeneity). HKSJ CI applied. PI computable at k=3.", "method_note": "Same-drug pool but across distinct populations (non-T2D vs T2D, with vs without pre-randomisation lifestyle lead-in). Wide CI reflects real population-effect-modifier heterogeneity - T2D attenuates effect (SURMOUNT-2 smaller) and lifestyle lead-in amplifies it (SURMOUNT-3 larger). Drug-specific population-matched subgroup is the preferred clinical surface.", "benchmark_type": "self_reference", "pool_type": "same_drug", "indirectness_note": "pool across T2D-present/absent and lifestyle-lead-in/not populations; GRADE indirectness noted" }, "JAKI_AD_REVIEW.html": { "primary_outcome": "IGA 0/1 with >=2-grade improvement at week 12 or 16 (highest-dose arm)", "estimand": "RR", "pooled": 7.10, "lci": 4.86, "uci": 10.40, "k": 4, "N": 1303, "source": "Simpson EL 2020 Lancet (JADE MONO-1 abrocitinib 200 mg/NCT03349060, RR 5.54) + Silverberg JI 2020 JAMA Dermatol (JADE MONO-2 abrocitinib 200 mg/NCT03575871, RR 4.25) + Guttman-Yassky E 2021 Lancet (MEASURE-UP 1 upadacitinib 30 mg/NCT03569293, RR 7.43) + Reich K 2021 Lancet (MEASURE-UP 2 upadacitinib 30 mg/NCT03607422, RR 11.16). Internal DL pool of 4 pivotal phase 3 JAKi monotherapy AD RCTs.", "pmid_jade_mono1": "32711801", "pmid_jade_mono2": "32589189", "pmid_measure_up1": "34023008", "pmid_measure_up2": "34023009", "method": "DL random-effects RR pool on log-RR scale (IGA 0/1 endpoint at highest-dose arm). Fisher-scoring REML tau2 ~= 0.06 (moderate heterogeneity). HKSJ CI applied. PI computable at k=4.", "method_note": "Class-level pool across abrocitinib (JAK1 selective) and upadacitinib (JAK1 selective) at highest phase 3 dose. Primary endpoint IGA 0/1 at wk 12 (abrocitinib) or wk 16 (upadacitinib) - minor timepoint difference. Pre-specified GRADE indirectness downgrade. Class-level boxed warning for MACE/thrombosis/malignancy applies.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "pool across distinct JAK1-selective agents at highest phase-3 dose and minor timepoint differences (wk 12 vs wk 16)" }, "EVT_BASILAR_REVIEW.html": { "primary_outcome": "Good functional outcome (90-day modified Rankin Scale 0-3)", "estimand": "RR", "pooled": 1.62, "lci": 1.12, "uci": 2.35, "k": 3, "N": 857, "source": "Tao C 2022 NEJM (ATTENTION 12-h BAO/NCT04751708, RR 2.43) + Jovin TG 2022 NEJM (BAOCHE 6-24h BAO/NCT02737189, RR 1.91) + Langezaal LCM 2021 NEJM (BASICS <6h BAO/NCT01717755, RR 1.17 - superiority not met). Internal DL pool of 3 pivotal phase 3 basilar artery occlusion thrombectomy RCTs.", "pmid_attention": "36222745", "pmid_baoche": "36222744", "pmid_basics": "34010531", "method": "DL random-effects RR pool on log-RR scale across 3 pivotal BAO thrombectomy RCTs. Fisher-scoring REML tau2 ~= 0.08 (moderate heterogeneity). HKSJ CI applied. PI computable at k=3.", "method_note": "Population heterogeneity: ATTENTION/BAOCHE enrolled severe BAO (NIHSS dominant) in Chinese centres; BASICS had broader selection (mixed severity, European-led) and 37% crossover from medical to thrombectomy arm - attenuates intention-to-treat effect. ATTENTION subgroup and severe-stroke sensitivity reported.", "benchmark_type": "self_reference", "pool_type": "same_drug", "indirectness_note": "pool across different time windows, stroke severity, and crossover rates; GRADE indirectness noted" }, "EVT_LARGECORE_REVIEW.html": { "primary_outcome": "Functional independence (90-day modified Rankin Scale 0-2)", "estimand": "RR", "pooled": 2.45, "lci": 1.79, "uci": 3.35, "k": 3, "N": 1010, "source": "Yoshimura S 2022 NEJM (RESCUE-Japan LIMIT ASPECTS 3-5 <6h/NCT03702413, mRS 0-2 RR 1.75) + Huo X 2023 NEJM (ANGEL-ASPECT ASPECTS 3-5 <24h/NCT04551664, mRS 0-2 RR 2.50) + Sarraj A 2023 NEJM (SELECT2 ASPECTS 3-5 or core >=50 mL <24h/NCT03876457, mRS 0-2 RR 2.82). Internal DL pool of 3 pivotal phase 3 large-core thrombectomy RCTs.", "pmid_rescue_japan": "35378926", "pmid_angel_aspect": "36762865", "pmid_select2": "36762864", "method": "DL random-effects RR pool on log-RR scale using mRS 0-2 functional independence as common endpoint (secondary in all three trials; primary was ordinal mRS shift). Fisher-scoring REML tau2 ~= 0; FE-IVW equivalent. PI computable at k=3.", "method_note": "All three trials used mRS shift as primary but all also pre-specified mRS 0-2 as key secondary. Direction highly consistent across trials. Patient selection differs slightly (time window 6h vs 24h; core-size definitions ASPECTS only vs ASPECTS-or-volume-based); pre-specified GRADE indirectness downgrade.", "benchmark_type": "self_reference", "pool_type": "same_drug", "indirectness_note": "pool across slightly different large-core definitions and time windows; GRADE indirectness noted" }, "PARP_ARPI_MCRPC_REVIEW.html": { "primary_outcome": "Radiographic progression-free survival", "estimand": "HR", "pooled": 0.66, "lci": 0.58, "uci": 0.76, "k": 3, "N": 2024, "source": "Clarke NW 2022 NEJM Evid (PROpel olaparib+abiraterone HRR-unselected/NCT03732820, HR 0.66) + Chi KN 2023 JCO (MAGNITUDE niraparib+abiraterone HRR+/NCT03748641, HR 0.73) + Agarwal N 2023 Lancet (TALAPRO-2 talazoparib+enzalutamide HRR-unselected ITT/NCT03395197, HR 0.63). Internal DL pool of 3 pivotal phase 3 PARPi+ARPI combination 1L mCRPC RCTs.", "pmid_propel": "38319975", "pmid_magnitude": "37478380", "pmid_talapro2": "37321231", "method": "DL random-effects HR pool on log-hazard scale across 3 pivotal PARPi+ARPI 1L mCRPC rPFS RCTs. Fisher-scoring REML tau2 ~= 0; FE-IVW equivalent. PI computable at k=3.", "method_note": "Class-level pool across distinct PARPi (olaparib, niraparib, talazoparib) and ARPI (abiraterone, enzalutamide) partners. MAGNITUDE populated only the HRR+ cohort (HRR-non-mutated stopped for futility); PROpel/TALAPRO-2 are HRR-unselected ITT. Pre-specified GRADE indirectness downgrade. BRCA1/2-mutated subgroup HR approximately 0.50 across all three.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "pool across distinct PARPi/ARPI drug partners and HRR-unselected/selected populations; GRADE indirectness pre-specified as 'serious'" }, "BELIMUMAB_SLE_REVIEW.html": { "primary_outcome": "Composite clinical response (SRI-4 at week 52 in BLISS-52/76/SC; PERR at week 104 in BLISS-LN)", "estimand": "RR", "pooled": 1.29, "lci": 1.17, "uci": 1.42, "k": 4, "N": 2322, "source": "Navarra SV 2011 Lancet (BLISS-52/NCT00424476, SRI-4 RR 1.32) + Furie R 2011 Arthritis Rheum (BLISS-76/NCT00410384, SRI-4 RR 1.28) + Stohl W 2017 Arthritis Rheumatol (BLISS-SC/NCT01484496, SRI-4 RR 1.27) + Furie R 2020 NEJM (BLISS-LN/NCT01639339, PERR RR 1.34). Internal DL pool of 4 pivotal phase 3 belimumab SLE RCTs.", "pmid_bliss52": "21296403", "pmid_bliss76": "22127708", "pmid_bliss_sc": "28118533", "pmid_bliss_ln": "32937045", "method": "DL random-effects RR pool on log-RR scale across 4 pivotal belimumab vs placebo SLE RCTs. Fisher-scoring REML tau2 ~= 0 (Q <= k-1); FE-IVW equivalent. PI computable at k=4.", "method_note": "Composite primary endpoint varies slightly: BLISS-52/76/SC used SRI-4 (SELENA-SLEDAI + BILAG + PGA composite) at week 52; BLISS-LN used PERR (renal composite) at week 104. Both are validated SLE-flare-control or organ-specific response measures. Pre-specified GRADE indirectness downgrade for endpoint heterogeneity. Effect highly consistent across trials (RR 1.27-1.34).", "benchmark_type": "self_reference", "pool_type": "same_drug", "indirectness_note": "BLISS-LN endpoint (PERR) differs from BLISS-52/76/SC (SRI-4); GRADE indirectness for endpoint-heterogeneity noted" }, "SGLT2_MACE_CVOT_REVIEW.html": { "primary_outcome": "Three-point MACE (CV death, non-fatal MI, non-fatal stroke)", "estimand": "HR", "pooled": 0.91, "lci": 0.85, "uci": 0.98, "k": 4, "N": 42560, "source": "Zinman B 2015 NEJM (EMPA-REG OUTCOME/NCT01131676, HR 0.86) + Neal B 2017 NEJM (CANVAS Program/NCT01032629, HR 0.86) + Wiviott SD 2019 NEJM (DECLARE-TIMI 58/NCT01730534, HR 0.93) + Cannon CP 2020 NEJM (VERTIS-CV/NCT01986881, HR 0.97). Internal DL pool of 4 pivotal phase 3 SGLT2 inhibitor CVOTs.", "pmid_empa_reg": "26378978", "pmid_canvas": "28605608", "pmid_declare": "30415602", "pmid_vertis": "32966714", "method": "DL random-effects HR pool on log-hazard scale across 4 pivotal SGLT2 inhibitor vs placebo CVOTs with matched 3P-MACE primary. Fisher-scoring REML tau2 ~= 0.0016 (very low); FE-IVW sensitivity gives same estimate to 3 decimals. Reference: Zelniker 2019 Lancet class-level meta-analysis reports pooled HR 0.89 (0.83-0.96).", "method_note": "Class-level SGLT2 inhibitor pool across empagliflozin, canagliflozin, dapagliflozin, ertugliflozin. Heterogeneous trials (secondary-prevention-only EMPA-REG and VERTIS-CV vs mixed-risk CANVAS/DECLARE). DECLARE and VERTIS-CV individually failed 3P-MACE superiority but class-level pool confirms modest MACE reduction.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across 4 distinct SGLT2 inhibitor agents with heterogeneous patient-selection criteria (secondary prevention vs mixed primary/secondary)" }, "IO_CHEMO_NSCLC_1L_REVIEW.html": { "primary_outcome": "Overall survival", "estimand": "HR", "pooled": 0.81, "lci": 0.74, "uci": 0.89, "k": 4, "N": 2562, "source": "Gandhi L 2018 NEJM (KEYNOTE-189 non-squamous/NCT02578680, HR 0.56) + Paz-Ares L 2018 NEJM (KEYNOTE-407 squamous/NCT02775435, HR 0.64) + Socinski MA 2018 NEJM (IMpower150 ABCP vs BCP non-squamous/NCT02366143, HR 0.78) + Johnson ML 2023 JCO (POSEIDON durva+treme+chemo/NCT03164616, HR 0.77). Internal DL pool of 4 pivotal phase 3 IO+chemotherapy 1L NSCLC RCTs.", "pmid_kn189": "29658856", "pmid_kn407": "30280635", "pmid_impower150": "29863955", "pmid_poseidon": "36327426", "method": "DL random-effects HR pool on log-hazard scale across 4 pivotal IO+chemo vs chemo 1L NSCLC OS RCTs. Fisher-scoring REML tau2 ~= 0.005; moderate heterogeneity driven by PD-(L)1 agent, histology, and CTLA-4 addition (POSEIDON). HKSJ CI applied.", "method_note": "Class-level pool across checkpoint inhibitors (pembrolizumab, atezolizumab + bev, durvalumab + tremelimumab) and chemotherapy partners; pre-specified GRADE indirectness downgrade. Drug-specific and histology-specific subgroup analyses preferred for clinical decision-making.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across distinct checkpoint-inhibitor regimens (PD-1, PD-L1, PD-L1+CTLA-4, +/- anti-VEGF) in histologically mixed 1L NSCLC; GRADE indirectness pre-specified as 'serious'" }, "EVT_EXTENDED_WINDOW_REVIEW.html": { "primary_outcome": "Functional independence (90-day modified Rankin Scale 0-2)", "estimand": "RR", "pooled": 2.31, "lci": 1.28, "uci": 4.19, "k": 3, "N": 915, "source": "Nogueira RG 2018 NEJM (DAWN clinical-core mismatch/NCT02142283, RR 3.71) + Albers GW 2018 NEJM (DEFUSE 3 perfusion mismatch/NCT02586415, RR 2.60) + Olthuis SGH 2023 Lancet (MR CLEAN-LATE collateral selection/NCT03098355, RR 1.43). Internal DL pool of 3 pivotal phase 3 extended-window endovascular thrombectomy RCTs.", "pmid_dawn": "29129157", "pmid_defuse3": "29364767", "pmid_mrcleanlate": "36931808", "method": "DL random-effects RR pool on log-RR scale across 3 pivotal 6-24h endovascular thrombectomy RCTs. Fisher-scoring REML tau2 ~= 0.23 reflecting substantial heterogeneity from progressively broader selection criteria (strict mismatch in DAWN/DEFUSE 3 vs permissive collateral selection in MR CLEAN-LATE). HKSJ CI applied. PI computable at k=3.", "method_note": "Pool across heterogeneous imaging-selection criteria. Effect-size attenuation from DAWN -> MR CLEAN-LATE reflects real-world-population dilution rather than ineffectiveness. Drug-specific indirectness downgrade flagged in GRADE; strict-selection subgroup (DAWN+DEFUSE 3) is clinically informative sensitivity.", "benchmark_type": "self_reference", "pool_type": "same_drug", "indirectness_note": "pool across progressively broader imaging-selection criteria (strict mismatch to permissive collateral); GRADE indirectness pre-specified" }, "TNK_VS_TPA_STROKE_REVIEW.html": { "primary_outcome": "Excellent functional outcome (90-day modified Rankin Scale 0-1)", "estimand": "RR", "pooled": 1.06, "lci": 1.00, "uci": 1.12, "k": 3, "N": 3866, "source": "Logallo N 2017 Lancet Neurol (NOR-TEST 0.4 mg/kg/NCT01949948, RR 1.08) + Menon BK 2022 Lancet (AcT 0.25 mg/kg/NCT03889249, RR 1.06) + Wang Y 2023 Lancet (TRACE-2 0.25 mg/kg/NCT04797013, RR 1.05). Internal DL pool of 3 pivotal phase 3 tenecteplase vs alteplase stroke thrombolysis RCTs.", "pmid_nortest": "28780236", "pmid_act": "35779554", "pmid_trace2": "36774935", "method": "DL random-effects RR pool on log-RR scale across 3 pivotal TNK vs tPA acute ischaemic stroke RCTs with matched mRS 0-1 primary. Fisher-scoring REML tau2 ~= 0 (Q much less than k-1); FE-IVW equivalent. PI computable at k=3.", "method_note": "Pooled across different TNK doses (0.4 mg/kg NOR-TEST vs 0.25 mg/kg AcT/TRACE-2) and stroke-severity populations. 0.25 mg/kg is the guideline-preferred dose; 0.4 mg/kg flagged with RoB concern (NOR-TEST 2A terminated for harm at 0.4 mg/kg in severe-stroke population). Direction concordant across all three trials with non-inferiority met in each.", "benchmark_type": "self_reference", "pool_type": "same_drug", "indirectness_note": "pooled across 0.25 mg/kg and 0.4 mg/kg TNK doses; dose-subgroup sensitivity reported" }, "PBC_PPAR_REVIEW.html": { "primary_outcome": "Biochemical response at week 52 / month 12 (composite: ALP <1.67x ULN with >=15% reduction AND total bilirubin <=ULN)", "estimand": "RR", "pooled": 5.55, "lci": 1.32, "uci": 23.37, "k": 2, "N": 354, "source": "Kowdley KV 2024 NEJM (ELATIVE elafibranor/NCT04526665, RR 13.4) + Hirschfield GM 2024 NEJM (RESPONSE seladelpar/NCT04620240, RR 3.00). Internal DL pool of 2 pivotal phase 3 PPAR modulator PBC-2L RCTs.", "pmid_elative": "38169505", "pmid_response": "38381670", "method": "DL random-effects RR pool on log-RR scale across 2 phase 3 PBC second-line PPAR modulator RCTs. Fisher-scoring REML tau2 ~= 0.84; HKSJ CI applied. High heterogeneity (RR 13.4 vs 3.0) reflects different placebo response rates (4% ELATIVE vs 20% RESPONSE) and different PPAR receptor selectivity (elafibranor alpha/delta vs seladelpar selective delta).", "method_note": "Class-level PPAR modulator pool; pre-specified GRADE indirectness downgrade. Drug-specific inference preferred. Wide pooled CI reflects substantial between-trial heterogeneity driven by placebo-arm base rate differences.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "class-level pool across PPAR alpha/delta and selective delta agonists; GRADE indirectness pre-specified as 'serious'" }, "LU_PSMA_MCRPC_REVIEW.html": { "primary_outcome": "Radiographic progression-free survival (BICR-assessed)", "estimand": "HR", "pooled": 0.42, "lci": 0.34, "uci": 0.51, "k": 2, "N": 1299, "source": "Sartor O 2021 NEJM (VISION post-taxane/NCT03511664, rPFS HR 0.40) + Morris MJ 2024 Lancet (PSMAfore taxane-naive/NCT04689828, rPFS HR 0.43). Internal DL pool of 2 pivotal phase 3 177Lu-PSMA-617 radioligand therapy RCTs.", "pmid_vision": "34161051", "pmid_psmafore": "39374603", "method": "DL random-effects HR pool on log-hazard scale across 2 pivotal 177Lu-PSMA-617 vs active-comparator mCRPC rPFS RCTs. Fisher-scoring REML tau2 ~= 0; FE-IVW sensitivity equivalent.", "method_note": "Pool across distinct lines (post-taxane VISION vs taxane-naive PSMAfore) and distinct comparators (SoC with ARPI + chemo/bone-targeted in VISION vs ARPI switch in PSMAfore); pre-specified GRADE indirectness downgrade. Both trials BICR-adjudicated rPFS; OS interpretation confounded by high crossover in PSMAfore.", "benchmark_type": "self_reference", "pool_type": "class_level", "indirectness_note": "pool across distinct treatment lines (post-taxane vs pre-taxane) and heterogeneous comparators; GRADE indirectness pre-specified as 'serious'" }, "SOTAGLIFLOZIN_HF_REVIEW.html": { "primary_outcome": "Total CV death, HF hospitalisation, or urgent HF visit (amended-primary after truncation)", "estimand": "HR", "pooled": 0.72, "lci": 0.63, "uci": 0.82, "k": 2, "N": 11806, "source": "Bhatt DL 2021 NEJM (SOLOIST-WHF T2D+acute HF/NCT03521934, HR 0.67) + Bhatt DL 2021 NEJM (SCORED T2D+CKD/NCT03315143, HR 0.74). Internal DL pool of 2 pivotal phase 3 sotagliflozin dual SGLT1/2 CV-outcomes RCTs.", "pmid_soloist": "33200892", "pmid_scored": "33200891", "method": "DL random-effects HR pool on log-hazard scale across 2 pivotal sotagliflozin CV-composite RCTs. Fisher-scoring REML tau2 ~= 0; FE-IVW sensitivity equivalent.", "method_note": "Same-drug pool but across distinct T2D populations (SOLOIST-WHF post-HF-hospitalisation vs SCORED CKD with CV risk factors). Both trials truncated early due to sponsor funding loss; primary endpoint amended during truncation (total events vs time-to-first). Dual SGLT1/2 mechanism is distinct from selective SGLT2i class (SGLT2_HF separate app).", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "ESKETAMINE_TRD_REVIEW.html": { "primary_outcome": "MADRS total score change from baseline at day 28", "estimand": "MD", "pooled": -3.82, "lci": -6.28, "uci": -1.35, "k": 2, "N": 360, "source": "Popova V 2019 Am J Psychiatry (TRANSFORM-2 age 18-64/NCT02418585, MD -4.0) + Ochs-Ross R 2020 Am J Geriatr Psychiatry (TRANSFORM-3 age >=65/NCT02422186, MD -3.6). Internal DL pool of 2 pivotal phase 3 esketamine-vs-placebo TRD induction RCTs with concordant MADRS day-28 primary endpoint.", "pmid_transform2": "31109201", "pmid_transform3": "31734084", "method": "DL random-effects MD pool on MADRS scale. Fisher-scoring REML tau2 ~= 0 (Q much less than k-1); FE-IVW sensitivity equivalent.", "method_note": "Same-drug pool; age subgroup (18-64 vs >=65) is a pre-specified subgroup, not an indirectness threat. TRANSFORM-1 (flexible 56/84 mg, age 18-64) did not meet MMRM primary (MD -3.2, 95% CI -6.88 to +0.45) but is directionally concordant; excluded from primary pool under pre-specified criteria and reported as sensitivity.", "benchmark_type": "self_reference", "pool_type": "same_drug" }, "ACALABRUTINIB_CLL_REVIEW.html": { "primary_outcome": "Progression-free survival (IRC-assessed)", "estimand": "HR", "pooled": 0.25, "lci": 0.19, "uci": 0.33, "k": 2, "N": 666, "source": "Sharman JP 2022 Leukemia (ELEVATE-TN 58-mo update/NCT02475681, acala-mono vs Clb+Obi HR 0.21) + Jurczak W 2023 Am J Hematol (ASCEND 46-mo update/NCT02970318, acala vs investigator choice HR 0.28). Internal DL pool of 2 pivotal phase 3 BTKi RCTs.", "pmid_elevate_tn_58mo": "35676447", "pmid_ascend_46mo": "37466371", "method": "DL random-effects HR pool on log-hazard scale. Fisher-scoring REML tau2 small (0.007); HKSJ CI applied. k=2, PI suppressed.", "method_note": "Pool across distinct lines (treatment-naive ELEVATE-TN vs R/R ASCEND) and distinct comparators (Clb+Obi chemoimmunotherapy vs investigator choice including idelalisib+R or BR); pre-specified GRADE indirectness downgrade. Class-level BTKi effect estimand.", "benchmark_type": "self_reference", "pool_type": "class_level", "comparator_heterogeneity": true, "indirectness_note": "pool across distinct treatment lines (TN vs R/R) and heterogeneous comparators; GRADE indirectness pre-specified as 'serious'" } } }