Evidence Synthesis Glossary

A method used in clinical trials to prevent selection bias by concealing the upcoming treatment assignment from those involved in enrolling participants. Adequate allocation concealment ensures that the assignment sequence cannot be predicted before or during enrolment.

Arzneimittelmarktneuordnungsgesetz: German drug market reform act requiring early benefit assessment for new pharmaceuticals. Systematic reviews are mandatory for dossier submission to G-BA.

Advanced Therapy Medicinal Products: a category of complex biologics including gene therapies, somatic cell therapies and tissue-engineered products regulated under EU Regulation 1394/2007. ATMPs entered mandatory EU JCA scope from January 2026 due to their high clinical uncertainty, high cost and significant unmet need, all contexts where rigorous, comprehensive systematic evidence synthesis is essential.

A systematic error in the design, conduct, analysis or reporting of a study that leads to results that differ from the truth. Common types include selection bias, performance bias, detection bias, attrition bias and reporting bias. Risk of bias assessment is a core component of systematic reviews.

A procedure in clinical trials where one or more parties (participants, investigators, outcome assessors) are kept unaware of treatment assignments. Single-blind means participants are unaware, double-blind means both participants and investigators are unaware. Blinding helps reduce performance and detection bias.

Biocidal Products Regulation: Regulation (EU) No 528/2012 governing the approval and use of biocidal products in the EU. Requires systematic toxicological and ecotoxicological evidence for active substance approval.

Canadian Agency for Drugs and Technologies in Health: Canada's national HTA body. Note: as of 2023, CADTH was announced to merge under the new Canadian Drug Agency (CDA). Verify current name at time of reading.

Clinical Evaluation Assessment Report: the notified body's report assessing the manufacturer's CER. Template defined in MDCG 2020-13. The CEAR documents the notified body's review of clinical evidence, methodology and conclusions.

Clinical Evaluation Report: mandatory document under EU MDR Article 61 and Annex XIV for medical devices demonstrating safety and performance through systematic literature review and clinical data analysis. Assessed by notified bodies using the CEAR template (MDCG 2020-13).

Classification, Labelling and Packaging: Regulation (EC) No 1272/2008 implementing the UN Globally Harmonized System (GHS) in the EU for classification of chemical hazards.

Cochrane Handbook for Systematic Reviews of Interventions: the gold-standard methodological reference for conducting systematic reviews. Current version: 6.4 (August 2023). Defines best practice for every stage of the systematic review process from protocol to reporting.

A systematic review produced by Cochrane, an international organisation dedicated to producing high-quality, relevant, accessible systematic reviews. Cochrane Reviews follow a standardised methodology and are published in the Cochrane Database of Systematic Reviews.

A range of values within which the true effect is expected to lie with a specified probability (typically 95%). A narrow confidence interval indicates high precision, while a wide interval suggests uncertainty. Confidence intervals that cross the line of no effect indicate a statistically non-significant result.

A statistical framework that produces prediction sets with guaranteed marginal coverage. Unlike standard confidence scores, conformal prediction intervals are provably calibrated; at α = 0.10, the true label falls within the interval ≥90% of the time.

Cosmetic Product Safety Report: mandatory under the EU Cosmetics Regulation (EC) No 1223/2009. Part A compiles safety information including systematic literature review across all relevant toxicological endpoints. Part B contains the safety assessment conclusion by a qualified safety assessor.

Contract Research Organisation: a company that provides outsourced research services to pharmaceutical, biotech and medical device companies. In evidence synthesis, CROs conduct systematic reviews on behalf of sponsors.

Chemical Safety Report: a document required under REACH for substances manufactured or imported at ≥10 tonnes per year. Contains hazard assessment, exposure assessment and risk characterisation. Systematic literature review is a core method for populating the CSR.

Defined Approaches for Skin Sensitisation: combined testing strategies using multiple non-animal methods (e.g. DPRA, KeratinoSens, h-CLAT) to classify skin sensitisation potential. Recognised in SCCS Notes of Guidance 12th Revision and OECD Guideline No. 497.

Direct Peptide Reactivity Assay: in chemico test method for skin sensitisation assessment (OECD TG 442C). A non-animal alternative measuring cysteine and lysine peptide reactivity.

European Chemicals Agency: EU agency managing REACH, CLP, BPR and other chemicals legislation. Based in Helsinki. ECHA maintains the substance registration database and coordinates substance evaluation.

A quantitative measure of the magnitude of a phenomenon or the strength of the relationship between variables. Common effect size measures include risk ratio (RR), odds ratio (OR), mean difference (MD) and standardised mean difference (SMD).

European Food Safety Authority: EU agency providing independent scientific advice on food-related risks. Based in Parma, Italy. Evaluates novel food applications, health claims, food additives and pesticide safety.

EFSA Panel on Nutrition, Novel Foods and Food Allergens: evaluates novel food applications, health claims and allergenicity assessments. Formerly Panel on Dietetic Products, Nutrition and Allergies.

EFSA Panel on Plant Protection Products and their Residues: provides scientific opinions on pesticide active substances, MRLs and dietary exposure assessments.

European Medicines Agency: the EU's centralised medicines regulatory authority, based in Amsterdam. Responsible for the scientific evaluation, supervision and safety monitoring of medicines in the EU.

Joint Clinical Assessment: the EU-level harmonised clinical evaluation of new medicines under Regulation (EU) 2021/2282. Mandatory for oncology medicines from January 2025, ATMPs from January 2026 and all centrally authorised medicines by January 2030. The JCA requires a comprehensive systematic literature review structured around multiple PICO schemes and produces a single EU-wide clinical evidence report used by all Member States as the foundation for national pricing and reimbursement decisions.

European Network for Health Technology Assessment: the predecessor EU collaboration of HTA bodies (2010-2023) that conducted voluntary joint clinical assessments and developed the methodology and templates now formalised under the EU HTA Regulation. EUnetHTA 21 produced 20 guidance documents and 13 templates that directly underpin today's JCA methodology.

A classification performance metric defined as the harmonic mean of Precision and Sensitivity (Recall): F1 = 2 × (Precision × Recall) / (Precision + Recall). In evidence synthesis AI, Sensitivity (not missing relevant studies) is typically prioritised over Precision, making F-beta variants with β > 1 more appropriate than standard F1. Per RAISE 2026 guidance, F1 alone is insufficient for evaluating AI screening tools.

U.S. Food and Drug Administration: the US federal agency responsible for regulating drugs, biologics, medical devices and food safety.

A graphical display used in meta-analyses to show the effect estimates and confidence intervals of individual studies alongside the pooled (combined) effect estimate. Each study is represented by a horizontal line (confidence interval) and a square (point estimate), with the diamond at the bottom representing the overall pooled result.

A list of prescription drugs covered by a health insurance plan, often tiered by cost-sharing level. Formulary decisions are informed by systematic evidence appraisal of clinical effectiveness and cost-effectiveness.

A scatter plot used to detect publication bias in meta-analyses. It plots study effect sizes against a measure of study precision (typically standard error). In the absence of publication bias, the plot should resemble a symmetrical inverted funnel. Asymmetry may indicate the presence of publication bias or other systematic differences.

Gemeinsamer Bundesausschuss (Federal Joint Committee): Germany's highest decision-making body for healthcare. Commissions IQWiG to evaluate added benefit of new drugs and issues binding reimbursement decisions based on AMNOG dossiers.

Grading of Recommendations, Assessment, Development and Evaluations: a framework for rating the certainty of evidence and strength of recommendations in systematic reviews and clinical guidelines.

Human Cell Line Activation Test: in vitro test for skin sensitisation assessment based on dendritic cell activation markers CD86 and CD54 (OECD TG 442E).

In generative AI, hallucination refers to the generation of factually incorrect, fabricated, or unsupported content presented with apparent confidence. In evidence synthesis contexts, hallucination is critically dangerous: fabricated citations, incorrect effect sizes, or non-existent studies cannot be tolerated in regulatory submissions. Puraite's RAG architecture and citation-level provenance system are specifically designed to prevent hallucination.

Haute Autorité de Santé: France's national HTA body. Evaluates drugs for the Transparency Committee (Commission de la Transparence) and issues SMR (medical service rendered) and ASMR (improvement of medical service rendered) ratings.

Regulation (EC) No 1924/2006: governs nutrition and health claims on food in the EU. Requires scientific substantiation through systematic review of the totality of available evidence. EFSA's NDA Panel evaluates claim applications.

Health Economics and Outcomes Research: research that assesses the value of medical treatments, informing coverage, pricing and reimbursement decisions by payers and HTA bodies.

Variability in study results beyond what would be expected by chance alone. Clinical heterogeneity refers to differences in participants, interventions or outcomes. Statistical heterogeneity is assessed using measures such as the I-squared statistic, where values above 50% suggest substantial heterogeneity.

Health Technology Assessment: the systematic evaluation of properties and effects of health technologies used to support coverage and reimbursement decisions. Requires comprehensive systematic evidence review.

The predefined characteristics that studies must have to be included in a systematic review. Typically defined using the PICO framework (Population, Intervention, Comparator, Outcome). Clear inclusion criteria are essential for ensuring the review is focused, reproducible and free from selection bias.

An analysis strategy in clinical trials where all participants are analysed according to their originally assigned group, regardless of whether they completed the intervention or crossed over to another group. ITT analysis preserves the benefits of randomisation and provides a more realistic estimate of treatment effects.

Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (Institute for Quality and Efficiency in Health Care): evaluates drugs, non-drug interventions and diagnostics for the G-BA. Publishes the Allgemeine Methoden (General Methods, current version 7.0, 2023) defining its systematic review methodology.

International standard for application of risk management to medical devices. Current version: ISO 14971:2019 (3rd edition). Systematic literature review supports hazard identification by identifying known complications, adverse events and failure modes associated with the device type.

Indirect Treatment Comparison: statistical method for comparing treatments that have never been directly compared in head-to-head trials, using a shared common comparator. Techniques include Bucher indirect comparison, network meta-analysis (NMA), MAIC and simulated treatment comparison (STC). Required in most EU JCA submissions where direct trial evidence against all required comparators is unavailable.

In Vitro Diagnostics Regulation (EU 2017/746): EU regulation governing in vitro diagnostic medical devices, requiring performance evaluation reports including systematic literature review of diagnostic accuracy.

In vitro skin sensitisation test based on keratinocyte activation (OECD TG 442D). Measures ARE-Nrf2 luciferase activity as an indicator of the keratinisation pathway in skin sensitisation.

A systematic review that is continuously updated as new evidence becomes available, rather than being a static one-time publication. Living reviews are particularly valuable in fast-moving therapeutic areas. Puraite's architecture is designed to support living evidence synthesis through automated literature monitoring and incremental screening.

Matching-Adjusted Indirect Comparison: a population adjustment method used in health technology assessment when individual patient data (IPD) is available for one treatment but only aggregate data for comparators. MAIC reweights the IPD to match the target population's baseline characteristics, enabling valid indirect comparisons. Widely used in oncology HTA and EU JCA submissions.

Medical Device Coordination Group: the EU body that issues guidance documents for implementation of the MDR and IVDR. Key guidance includes MDCG 2020-5 (equivalence), MDCG 2020-6 (legacy devices) and MDCG 2020-13 (CEAR template). MDCG documents are the operative references under MDR, replacing the older MEDDEV guidance.

Medical Device Regulation (EU 2017/745): EU regulation requiring medical devices to undergo clinical evaluation including systematic literature review per Article 61 and Annex XIV. Fully applicable since 26 May 2021. MDCG guidance documents (2020-5, 2020-13) are the operative references, replacing the older MEDDEV guidance.

Statistical synthesis of quantitative results from multiple independent studies. Requires systematic identification and selection of eligible studies, the foundation of which is a rigorous systematic review.

Maximum Residue Level: the maximum concentration of a pesticide residue legally permitted in food. Set under Regulation (EC) No 396/2005. Derivation requires systematic review of toxicological data and dietary exposure assessment.

New Approach Methodologies: umbrella term for non-animal testing methods including in vitro, in chemico, in silico and defined approaches. Increasingly required under EU cosmetics regulation due to the animal testing ban. Replacing traditional animal methods such as LLNA for sensitisation assessment.

A qualitative summary of evidence on a topic that does not follow the strict, predefined methodology of a systematic review. While useful for providing broad overviews, narrative reviews are more susceptible to selection bias and are generally considered lower in the evidence hierarchy.

A quality assessment tool for non-randomised studies (cohort and case-control) in systematic reviews. Evaluates three domains: selection, comparability, and outcome/exposure. Widely used in CER literature reviews and HTA dossiers involving observational evidence.

National Institute for Health and Care Excellence: UK HTA body issuing technology appraisals (TAs) and highly specialised technology evaluations (HSTEs) for NHS England. Its Methods and Process Guide (updated 2022) defines the methodological standards for drug appraisals. Note: the NICE Evidence Standards Framework (ESF) is specifically for digital health technologies, not pharma.

Network Meta-Analysis: statistical technique allowing simultaneous comparison of multiple treatments even when direct head-to-head trials don't exist, using a network of indirect comparisons.

An organisation designated by an EU member state to assess whether medical devices meet regulatory requirements before market placement. Examples include BSI, TÜV SÜD and SGS. Under MDR, notified bodies perform more rigorous clinical evaluation scrutiny than under the former MDD.

Regulation (EU) 2015/2283: requires authorisation for foods not consumed to a significant degree in the EU before 15 May 1997. Applicants must compile systematic safety, allergenicity and nutritional evidence for EFSA assessment.

The average number of patients who need to be treated with the intervention for one additional patient to benefit compared with the control. An NNT of 1 means every patient benefits. Lower NNTs indicate more effective treatments.

A measure of association between an exposure and an outcome. It represents the odds that an outcome will occur given a particular exposure, compared to the odds of the outcome occurring without that exposure. Commonly used in case-control studies and meta-analyses of binary outcomes.

Occupational Exposure Limit: the maximum concentration of a chemical agent in workplace air that workers may be exposed to. IOELVs (indicative) are health-based; BOELVs (binding) also consider technical feasibility. Derived using systematic review of epidemiological and toxicological evidence.

Pharmaceutical Benefits Advisory Committee: Australia's statutory body recommending drugs for subsidised listing on the Pharmaceutical Benefits Scheme (PBS).

Persistent, Bioaccumulative, Toxic / very Persistent, very Bioaccumulative: classifications for substances that persist in the environment and accumulate in living organisms. Assessed under REACH Annex XIII criteria.

Population, Intervention, Comparator, Outcome: the standard framework for defining systematic review eligibility criteria. Puraite uses PICO to structure AI screening criteria and evidence extraction.

A structured approach to formulating a clinical research question. P stands for Population (who), I for Intervention (what treatment), C for Comparator (what alternative) and O for Outcome (what result). The PICO framework is the foundation for developing search strategies and eligibility criteria in systematic reviews.

Extension of PICO adding Timing and Setting, used in comparative effectiveness research and HTA submissions to more precisely define study eligibility for systematic reviews.

Post-Market Clinical Follow-Up: required under EU MDR, involving systematic collection and evaluation of clinical data after CE marking. PMCF plans must be grounded in evidence gap analysis from systematic review.

A requirement that a healthcare provider obtains approval from a payer before a specific treatment, medication or procedure will be covered. Evidence-based prior authorisation criteria should be grounded in systematic review of clinical guidelines.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses: the reporting standard for systematic reviews. Includes the PRISMA flow diagram showing study screening and selection at each stage.

International Prospective Register of Systematic Reviews: the pre-registration database for systematic review protocols maintained by the University of York. Pre-registering a protocol on PROSPERO before conducting a review is a best-practice requirement for HTA submissions and publication in peer-reviewed journals. Registration prevents outcome reporting bias and duplicate reviews.

The tendency for studies with positive or statistically significant results to be published more frequently than those with negative or null results. Publication bias can lead to an overestimation of treatment effects in systematic reviews and meta-analyses.

Quality Assessment of Diagnostic Accuracy Studies: tool for assessing risk of bias and applicability concerns in diagnostic accuracy studies. Evaluates patient selection, index test, reference standard and flow/timing domains. Essential for IVDR performance evaluation literature reviews.

Risk Assessment Committee of ECHA: since 2019, responsible for scientific opinions on occupational exposure limits (previously SCOEL), SVHC identification, restriction proposals and harmonised classification under CLP.

Retrieval-Augmented Generation: AI architecture that combines document retrieval with generative models. Puraite uses RAG to ground AI outputs in actual literature, enabling citation-level provenance.

Responsible use of AI in Evidence SynthEsis: the leading international guidance framework (Version 3, March 2026; submitted to Research Synthesis Methods) for AI tools in systematic evidence synthesis. RAISE covers three core principles: Transparency/Traceability (AI decisions must be auditable), Comprehensiveness (recall must not be sacrificed) and Accountability (continuous post-deployment auditing). Puraite is designed and evaluated in accordance with RAISE 2026. Source: Thomas J, Hair K, Noel-Storr A et al., OSF DOI 10.17605/OSF.IO/FWAUD.

A study design where participants are randomly allocated to receive either the intervention being tested or a control (placebo or standard care). RCTs are considered the gold standard for evaluating the efficacy of interventions because randomisation minimises the risk of confounding and selection bias.

A streamlined form of evidence synthesis that uses simplified systematic review methods to produce evidence summaries in a shorter timeframe. Trade-offs in methodological rigour are made to meet urgent decision-making needs, such as during health crises or policy deadlines.

Randomised Controlled Trial: the gold standard study design for evaluating treatment efficacy. Systematic reviews of RCTs form the highest level of the evidence hierarchy for HTA submissions.

Relative Effectiveness Assessment: the scientific analysis at the core of each EU JCA, comparing the clinical efficacy, effectiveness and safety of a new medicine against comparators in routine clinical practice. REA does not include economic evaluation; that remains national. It is the EU successor to EUnetHTA joint assessments and the foundation for all JCA dossier submissions.

Registration, Evaluation, Authorisation and Restriction of Chemicals: Regulation (EC) No 1907/2006. The EU's main chemicals regulation requiring manufacturers and importers to register substances with ECHA and demonstrate safe use through Chemical Safety Reports.

Systematic assessment of methodological flaws in individual studies that could distort their results. In SLRs, Risk of Bias (RoB) assessment is mandatory for every included study using validated tools such as Cochrane RoB 2 (for RCTs) or ROBINS-I (for observational studies). RoB findings directly inform GRADE certainty-of-evidence ratings.

Risk of Bias 2 tool: Cochrane's revised tool for assessing risk of bias in randomised controlled trials. Evaluates 5 domains: randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Current standard for RCT quality assessment in systematic reviews.

Scientific Committee on Consumer Safety: EU scientific committee providing safety opinions on cosmetic ingredients at the European Commission's request. The SCCS Notes of Guidance (12th Revision, 2023) define data requirements and quality criteria for safety dossiers.

Full title: SCCS Notes of Guidance for the Testing of Cosmetic Ingredients and their Safety Evaluation. Current: 12th Revision, adopted 15 May 2023 (SCCS/1647/22). Defines the expected data structure, toxicological endpoints and study quality criteria for cosmetic ingredient safety assessment.

Scientific Committee on Occupational Exposure Limits: EU committee that derived OEL recommendations from 1995 to 2018. Mandate ended in 2018; since 2019, ECHA's RAC (Risk Assessment Committee) has taken over this function.

A form of evidence synthesis used to map the breadth and nature of evidence on a topic, identify research gaps and clarify concepts, without the full critical appraisal required in a systematic review. Scoping reviews follow PRISMA-ScR reporting guidance and are frequently used as a precursor to full SLRs in regulatory and HTA contexts.

A repeat of the primary analysis using different assumptions, methods or subsets of data to test the robustness of the original findings. In systematic reviews, sensitivity analyses may exclude studies at high risk of bias, use different statistical models or vary the inclusion criteria.

Systematic Literature Review: a structured, reproducible process for identifying, appraising and synthesising all available evidence on a defined research question. Required for HTA, MDR CER and regulatory submissions.

A payer requirement that patients try lower-cost medications before being approved for more expensive alternatives. Step-therapy protocols should be supported by systematic evidence of comparative effectiveness.

Systematic reviews analyse 'studies,' but a single study can be reported across multiple publications, conference abstracts or registry entries. Correctly identifying which papers report on the same underlying study is critical. Misidentifying linked reports can lead to double-counting outcomes or missing key data, distorting the final synthesis results.

An analysis that examines whether the treatment effect varies across different subsets of participants defined by characteristics such as age, sex, disease severity or study quality. Subgroup analyses should be pre-specified in the review protocol to avoid data dredging.

Substance of Very High Concern: substances identified under REACH Article 57 for properties such as CMR (carcinogenic, mutagenic, toxic to reproduction), PBT/vPvB or equivalent concern. Listed on the ECHA Candidate List for eventual authorisation requirement.

A structured, reproducible research methodology that identifies, appraises and synthesises all relevant evidence on a specific research question following a predefined protocol. Considered the highest level of evidence in the evidence hierarchy. Typically follows PRISMA reporting guidelines.

A focused evidence synthesis that addresses a specific, narrow clinical or regulatory question. Often used in health technology assessment (HTA) dossiers or regulatory submissions where a full systematic review is not required but structured evidence gathering is necessary.

A review of existing systematic reviews on a given topic. Also called a review of reviews. Umbrella reviews provide a high-level summary of evidence across multiple systematic reviews, helping to identify consistencies and contradictions across the literature.

Explainable Artificial Intelligence: AI systems designed to produce outputs that can be understood and audited by humans. In regulated contexts, XAI is not optional; every AI decision must be traceable and defensible.