Evidence, Guidelines, Registries, and Living-Textbook Method

1.1Evidence appraisal and methodological standards

Evidence-based vascular care integrates the best current evidence with patient anatomy, operative risk, life expectancy, symptom burden, care goals, and local procedural performance Sackett 1996.

Randomized trials are evaluated for reporting completeness using CONSORT 2010 and for methodological bias using RoB 2 RoB 2 (2019) · CONSORT 2010. Observational evidence, including cohort, case-control, and registry analyses, is evaluated for reporting transparency using STROBE and for risk of bias using ROBINS-I ROBINS-I 2016 · STROBE 2007. Diagnostic accuracy studies underpinning vascular imaging and physiologic tests are evaluated for reporting completeness using STARD 2015 and for methodological quality and applicability using QUADAS-2 STARD 2015 · QUADAS-2 (2011). Systematic reviews and meta-analyses are evaluated for reporting completeness using PRISMA 2020 and for methodological credibility using AMSTAR 2 PRISMA 2020 · Page 2020 · AMSTAR 2 (2017).

1.2Guideline development and recommendation strength

Guidelines translate evidence into structured practice recommendations. The GRADE methodology provides the standard vocabulary for rating evidence certainty and guidance strength GRADE 2008 · GRADE 2004. GRADE rates certainty of evidence as high, moderate, low, or very low. Randomized-trial evidence starts at high certainty and observational evidence at low certainty; certainty is rated down for risk of bias, inconsistency, indirectness, imprecision, and publication bias, and rated up for a large effect magnitude, a dose-response gradient, or plausible residual confounding that would bias against the observed effect. The Evidence to Decision framework incorporates clinical priority, expected effects, evidence certainty, patient values, resources, cost-effectiveness, equity, acceptability, and feasibility GRADE Evidence to Decision 2016.

A strong recommendation applies to most patients and supports default practice unless contraindicated by patient-specific factors. A conditional recommendation requires explicit individual selection based on benefit-harm trade-offs and patient preferences GRADE Evidence to Decision 2016 · GRADE 2008. ACC/AHA guidelines grade every recommendation by Class of Recommendation and Level of Evidence ACC/AHA 2016. Class I means benefit far outweighs risk (is recommended), Class IIa that benefit outweighs risk (is reasonable), Class IIb that benefit equals or marginally exceeds risk (may be considered), and Class III captures either no benefit or harm. Level A rests on high-quality evidence from more than one randomized trial or meta-analyses of high-quality trials, B-R on moderate-quality evidence from one or more randomized trials, B-NR on nonrandomized or observational studies, C-LD on limited data, and C-EO on expert opinion.

Guideline trustworthiness and methodology are appraised using AGREE II, Guidelines 2.0, and NASEM standards, which assess scope, stakeholder involvement, development rigour, conflict management, and external review Guidelines 2.0 (2014) · NASEM 2011 · AGREE II 2010. Recommendation strength reflects the balance of benefits and harms rather than citation volume alone Tricoci 2009.

1.3Registries and routinely collected data

Registries capture real-world practice, uncommon presentations, device use, and outcomes in anatomic subgroups typically excluded from randomized trials. Observational databases remain vulnerable to selection bias, unmeasured confounding, changing definitions, missing data, and variation in follow-up. Studies using electronic health records, claims data, and disease registries are reported according to the RECORD extension to STROBE RECORD 2015 · STROBE 2007. Registry analyses used to infer causal intervention effects require explicit bias assessment with ROBINS-I ROBINS-I 2016.

1.4Artificial intelligence and prediction models

Prediction scores, risk calculators, and diagnostic classifiers require transparent reporting, external validation, and calibration to the local target population. The TRIPOD statement governs the reporting of multivariable prediction models, while the TRIPOD+AI extension applies these standards to artificial-intelligence and machine-learning models TRIPOD+AI 2024 · TRIPOD 2015. Randomized trials of AI-supported interventions are assessed using the CONSORT-AI extension CONSORT-AI 2020 · CONSORT 2010.

1.5Clinical integration and living updates

Clinical integration matches the patient-specific uncertainty to the corresponding evidence tool. Follow-up plans and surveillance tests rely on diagnostic accuracy and prediction-model calibration proven in similar patient populations. Evidence application is fundamentally constrained by false-positive findings, low prior probability, inadequate statistical power, bias, and multiple testing Ioannidis 2005. Avoidable research waste stems from poor question prioritization, weak methodology, and incomplete reporting Chalmers 2014.

Living systematic reviews and guidelines continuously integrate emerging evidence, reacting to practice-changing trials, regulatory actions, and safety warnings AHA Digital Guidelines · 2024 · MacFarlane 2017. Updating methodologies require explicit triggers and defined thresholds to prevent unwarranted practice shifts from single studies. A topic enters living mode when three conditions hold together: it is a priority for decision-making, current certainty in the evidence is low or moderate, and new evidence is emerging fast enough to plausibly change the conclusion. Living mode then runs continual active surveillance with searches typically re-run monthly, against pre-specified criteria for when a newly identified study triggers re-synthesis and a guideline or textbook update.

1.6Areas of controversy

The translation of observational registry data into causal treatment effects remains methodologically controversial, despite formal frameworks such as ROBINS-I and RECORD ROBINS-I 2016 · RECORD 2015. The criteria for adopting artificial-intelligence models into routine vascular decision-making lack universal consensus, specifically regarding the required degree of external validation versus local calibration TRIPOD+AI 2024 · CONSORT-AI 2020. The threshold at which new evidence triggers a living guideline update requires balancing rapid integration against the risk of overreaction to isolated findings AHA Digital Guidelines · 2024 · MacFarlane 2017.