Biostatistics & Population Health

Reporting guidelines: CONSORT, STROBE, PRISMA

Clinical Overview and When to Suspect Inadequate Reporting

— CONSORT (Consolidated Standards of Reporting Trials): randomized controlled trials (RCTs)

— STROBE (Strengthening the Reporting of Observational Studies in Epidemiology): cohort, case-control, cross-sectional

— PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses): systematic reviews and meta-analyses

— Trial publication that fails to describe allocation concealment, blinding, or loss to follow-up → CONSORT gap

— Observational study lacking confounder list, source population, or selection methodology → STROBE gap

— Meta-analysis missing search strategy, eligibility criteria, or flow diagram of records screened → PRISMA gap

Board pearl: Reporting guidelines ≠ methodological quality tools. CONSORT tells you what to report; risk-of-bias tools (Cochrane RoB 2, Newcastle-Ottawa, ROBINS-I) tell you whether the methods were sound. Step 3 distractors often swap these — if the stem asks "how to assess bias," pick a bias tool, not CONSORT/STROBE/PRISMA.

Reporting guidelines are structured checklists that standardize how health research is written, ensuring transparency, reproducibility, and critical appraisal — they do not dictate study design but dictate disclosure.

The three dominant frameworks tested on Step 3 biostatistics/EBM blocks:

When to "suspect" a reporting deficiency in a vignette:

Step 3 typically frames these as "Which element is missing?" or "Which guideline applies?" questions, often paired with critical appraisal of a manuscript abstract.

Clinical relevance: poorly reported research drives inappropriate guideline adoption, wasted resources, and patient harm — the EQUATOR Network (Enhancing the QUAlity and Transparency Of health Research) hosts >500 reporting guidelines as a centralized repository.

Each framework includes a checklist plus a flow diagram (CONSORT and PRISMA) tracking participants/studies through phases.

Presentation Patterns and Key History — Recognizing Study Type First

— Words: "randomly assigned," "1:1 allocation," "double-blind," "placebo-controlled," "intention-to-treat"

— Two or more arms with a prospective intervention

— Includes cluster RCTs, crossover trials, non-inferiority trials (each has a CONSORT extension)

— Cohort: "followed over 10 years," "incidence of MI in smokers vs nonsmokers," exposure precedes outcome

— Case-control: "patients with pancreatic cancer were compared with matched controls regarding prior alcohol use" — outcome already exists, look backward

— Cross-sectional: "prevalence of hypertension in a 2024 NHANES sample" — single time point

— "Searched PubMed, Embase, Cochrane through March 2024"

— "Pooled odds ratio across 14 studies"

— Forest plots, funnel plots, heterogeneity (I²) statistics

— SPIRIT (trial protocols), CARE (case reports), SRQR/COREQ (qualitative), TRIPOD (prediction models), STARD (diagnostic accuracy), CHEERS (economic evaluation), AGREE (clinical practice guidelines).

Key distinction: A retrospective chart review of patients given drug A vs drug B is observational (STROBE), not a trial — randomization is the dividing line. If clinicians chose the treatment based on clinical judgment, allocation was nonrandom → confounding by indication → STROBE applies, and the paper needs adjustment methods reported.

Before selecting a reporting guideline, you must classify the study from the vignette stem. This is the single most common Step 3 trap.

RCT signals → CONSORT:

Observational signals → STROBE:

Systematic review / meta-analysis → PRISMA:

Key historical context: CONSORT first published 1996 (updated 2010, with CONSORT 2025 revision), STROBE 2007, PRISMA 2009 (updated PRISMA 2020, 27 items + new flow diagram).

Extensions you may encounter:

Structural Anatomy of Each Guideline — The "Exam" of the Manuscript

— Title/abstract (identify as randomized), background/objectives, trial design, participants (eligibility, settings), interventions (with sufficient detail to replicate), outcomes (primary/secondary, any changes after trial began), sample size calculation, randomization (sequence generation, allocation concealment, implementation), blinding, statistical methods, participant flow (CONSORT diagram), recruitment, baseline data, numbers analyzed, outcomes/estimation, ancillary analyses, harms, limitations, generalizability, interpretation, registration, protocol, funding.

— Item 6: eligibility criteria and sources of participants

— Item 7: clearly define exposures, outcomes, confounders, effect modifiers

— Item 12: statistical methods including how missing data and confounding were addressed

— Item 16: report adjusted estimates alongside unadjusted, with confidence intervals

— Eligibility criteria (PICO), information sources, full search strategies, study selection process, data collection, risk-of-bias assessment, effect measures, synthesis methods (including heterogeneity), reporting bias assessment (funnel plot, Egger test), certainty assessment (often GRADE), registration (PROSPERO).

Board pearl: The flow diagram is the single most testable visual. CONSORT diagram tracks participants (assessed → randomized → allocated → followed → analyzed). PRISMA diagram tracks records (identified → screened → eligible → included). STROBE has no required flow diagram, though many journals request one.

Each guideline mirrors the IMRaD structure (Introduction, Methods, Results, and Discussion) plus title/abstract and other information.

CONSORT 2010 — 25 items across:

STROBE — 22 items, with 18 common across designs and 4 design-specific (cohort, case-control, cross-sectional). Hot spots:

PRISMA 2020 — 27 items + abstract checklist (12 items) + flow diagram:

Each guideline has an Explanation and Elaboration (E&E) companion paper giving rationale and examples — these are the source of many test items.

Diagnostic Workup — Reading a Manuscript for Compliance

— RCT not labeled "randomized" → CONSORT item 1a violation

— Systematic review without "systematic review" or "meta-analysis" in title → PRISMA gap

— Observational study not specifying design (cohort vs case-control) → STROBE item 1

— CONSORT: was allocation concealment described separately from sequence generation? "Sealed opaque envelopes" or central web-based allocation is the gold standard wording.

— STROBE: are confounders prespecified, and is the source population defined (e.g., "all adults enrolled in Kaiser Permanente Northern California 2015–2020")?

— PRISMA: is the search strategy reproducible — at least one full electronic search included as supplement?

— CONSORT: numbers randomized vs analyzed (look for ITT vs per-protocol), harms reported separately (CONSORT Harms extension)

— STROBE: unadjusted and adjusted estimates, with confidence intervals; reasons for nonparticipation

— PRISMA: characteristics of included studies, risk-of-bias summary, forest plots with pooled estimate and I², subgroup/sensitivity analyses

— All three require explicit limitations and generalizability (external validity) statements.

— CONSORT and PRISMA require registration (ClinicalTrials.gov, PROSPERO).

Step 3 management: When a vignette asks "what is the most important missing element," prioritize items that affect bias interpretation — allocation concealment, blinding, loss to follow-up (CONSORT); confounder adjustment (STROBE); search comprehensiveness and risk-of-bias assessment (PRISMA). These drive validity, not just completeness.

Approach a study like a chart: scan title/abstract first, then methods, then results, looking for guideline-required elements.

Title/abstract red flags:

Methods section essentials by guideline:

Results section essentials:

Discussion section:

Advanced or Confirmatory Studies — Guideline Extensions

— CONSORT-Cluster: cluster randomized trials (e.g., randomizing clinics, not patients) — must report number of clusters, intracluster correlation coefficient (ICC)

— CONSORT-Noninferiority/Equivalence: prespecified margin, choice of analysis population (per-protocol often co-primary)

— CONSORT-Harms, CONSORT-PRO (patient-reported outcomes), CONSORT-AI (artificial intelligence interventions, 2020), CONSORT-SPI (social/psychological interventions), CONSORT-Pilot

— STROBE-ME (molecular epidemiology), STROBE-MR (Mendelian randomization), STROBE-Vet (veterinary, not Step 3), RECORD (routinely collected health data — claims, EHR-based studies — critical for modern observational research)

— RECORD-PE (pharmacoepidemiology using routinely collected data)

— PRISMA-NMA (network meta-analyses, indirect comparisons), PRISMA-IPD (individual patient data), PRISMA-DTA (diagnostic test accuracy reviews), PRISMA-ScR (scoping reviews), PRISMA-Harms, PRISMA-Equity, PRISMA-Children, PRISMA-Abstracts, PRISMA-S (search reporting).

— TRIPOD/TRIPOD+AI: prediction models (e.g., risk calculators)

— STARD 2015: diagnostic accuracy primary studies

— CHEERS 2022: cost-effectiveness analyses

— SQUIRE 2.0: quality improvement reports (very Step 3-relevant — QI projects fall here, not CONSORT)

Key distinction: A registry-based RCT still uses CONSORT (it's randomized); a registry-based comparative effectiveness study uses STROBE + RECORD (observational). The data source doesn't dictate the guideline — the design does.

Many study types have a specific CONSORT/STROBE/PRISMA extension — recognizing the extension is increasingly tested.

CONSORT extensions (selected):

STROBE extensions:

PRISMA extensions:

Allied non-CONSORT/STROBE/PRISMA frameworks worth distinguishing:

Risk Stratification — Choosing the Right Guideline in 10 Seconds

— Step 1: Did the investigators assign the intervention? Yes → trial; No → observational or review.

— Step 2 (trial branch): Was assignment random? Yes → CONSORT (+ extension as needed). No → quasi-experimental → TREND statement (rarely tested, but a known distractor).

— Step 3 (observational branch): Is it a synthesis of multiple existing studies? Yes → PRISMA. No → STROBE.

— Step 4 (specialty designs): Diagnostic accuracy → STARD; Prediction model → TRIPOD; Case report → CARE; QI project → SQUIRE; Economic evaluation → CHEERS; Guideline → AGREE/RIGHT.

— "Investigators reviewed the EHR to compare outcomes" → STROBE/RECORD, not CONSORT

— "Researchers pooled 23 trials of statins" → PRISMA, not CONSORT (the trials used CONSORT; the review uses PRISMA)

— "Patients were assigned by alternation (every other patient)" → quasi-randomized; methodologists still often apply CONSORT, but allocation concealment is impossible — high bias risk

— Missing registration → suspicion of selective outcome reporting

— Missing allocation concealment / blinding → high risk of selection and detection bias

— Missing flow diagram → unable to assess attrition bias

— Missing funding/conflict disclosure → integrity concern

Board pearl: If a Step 3 stem describes researchers changing the primary outcome after data lock and not disclosing it, this violates CONSORT item 6b (changes to outcomes) and is a form of outcome reporting bias. Trial registration (ClinicalTrials.gov, mandated by ICMJE since 2005) is the safeguard.

Algorithmic approach for the exam:

Common Step 3 vignette switches that change the answer:

Risk hierarchy of reporting deficiencies (which is "worst"):

Pharmacotherapy Analog — Item-Level Deep Dive for CONSORT

— Item 8a — Sequence generation: must describe the method (e.g., "computer-generated random number sequence, block size 4") — not just "randomized."

— Item 9 — Allocation concealment mechanism: central randomization via web/phone, or sequentially numbered opaque sealed envelopes (SNOSE). This prevents the enroller from knowing the next assignment.

— Item 11a — Blinding: who was blinded (participants, providers, outcome assessors, data analysts) and how. Open-label trials must justify why blinding wasn't possible (e.g., surgery vs medication).

— Item 12 — Statistical methods: prespecified primary analysis, handling of missing data (multiple imputation > complete case), subgroup analyses prespecified vs post hoc.

— Item 13 — Participant flow (the CONSORT diagram): enrolled → randomized → allocated to each arm → received intervention → lost to follow-up → analyzed. Numbers must reconcile.

— Item 16 — Numbers analyzed: intention-to-treat (ITT) is the primary analysis for superiority trials; per-protocol is co-primary for non-inferiority.

— Item 17 — Outcomes and estimation: effect size with 95% confidence interval, not just p-value.

— Item 19 — Harms: all serious adverse events, withdrawals due to harms, with absolute numbers.

— Item 23 — Registration, Item 24 — Protocol, Item 25 — Funding/role of sponsor.

— Reporting p < 0.05 without confidence intervals

— Conflating randomization with allocation concealment

— Reporting only per-protocol results in a superiority trial (introduces attrition bias)

Step 3 management: When a question describes loss to follow-up of >20% with only per-protocol analysis reported, the correct critique is attrition bias compounded by violation of intention-to-treat principle — a CONSORT item 16 deficiency. Recommend sensitivity analysis with worst-case imputation.

CONSORT 2010 critical items most tested:

Common manuscript failure modes:

Procedures Analog — Item-Level Deep Dive for STROBE and PRISMA

— Item 5 — Setting: dates of recruitment, exposure, follow-up, data collection — anchors temporality and external validity.

— Item 6 — Participants: for cohort, eligibility + methods of selection + follow-up; case-control, source/methods of case ascertainment + matching criteria; cross-sectional, eligibility + methods of selection.

— Item 7 — Variables: clearly define outcomes, exposures, predictors, potential confounders, effect modifiers — with diagnostic criteria if applicable.

— Item 9 — Bias: describe efforts to address potential sources (selection bias, information bias, recall bias in case-control).

— Item 12 — Statistical methods: how confounding was addressed (multivariable regression, propensity scores, instrumental variables, stratification, matching), subgroup interactions, missing data handling, sensitivity analyses.

— Item 13–14 — Participants and descriptive data: numbers at each stage, reasons for nonparticipation, follow-up time.

— Item 16 — Main results: report both unadjusted and confounder-adjusted estimates with 95% CIs; categorize continuous variables only with justification.

— Item 5 — Eligibility criteria (PICO + study design)

— Item 7 — Search strategy: present full strategy for at least one database in the supplement

— Items 11–12 — Risk-of-bias assessment of included studies (RoB 2 for RCTs, ROBINS-I for non-randomized)

— Item 13 — Synthesis methods: rationale for meta-analysis vs narrative synthesis, model choice (fixed vs random effects), heterogeneity (I², τ²), subgroup/sensitivity analyses

— Item 14 — Reporting bias: funnel plot, Egger test if ≥10 studies

— Item 22 — Certainty: typically GRADE assessment per outcome

— Item 24 — Registration: PROSPERO ID

Board pearl: Heterogeneity reporting is required by PRISMA but the interpretation is methodological — I² >50% = substantial heterogeneity, warranting subgroup or sensitivity analysis or reconsideration of pooling. A meta-analysis pooling wildly heterogeneous studies without discussion is a PRISMA item 13d violation.

STROBE critical items:

PRISMA 2020 critical items:

Special Populations — Vulnerable Subjects and Equity Reporting

— Report participant characteristics across PROGRESS-Plus dimensions: Place of residence, Race/ethnicity, Occupation, Gender/sex, Religion, Education, Socioeconomic status, Social capital — Plus age, disability, sexual orientation.

— Disaggregate outcomes when feasible to detect differential effects.

— RCTs in children: include age strata, developmental considerations, assent vs consent, parental/guardian role.

— Pregnancy trials: gestational age at enrollment, fetal/neonatal outcomes, follow-up duration of offspring.

— Both require CONSORT compliance plus journal-specific pediatric reporting standards.

— STROBE requires reporting of age distribution, multimorbidity, frailty indices when relevant to confounding.

— Trials frequently exclude older adults; CONSORT item 21 (generalizability) should address this limitation explicitly.

— Pharmacology trials should report eGFR/Child-Pugh stratification in baseline tables (CONSORT item 15).

— Subgroup analyses must be prespecified to avoid spurious findings.

— NIH and FDA increasingly require race/ethnicity reporting in clinical trials; CONSORT 2025 strengthens this requirement.

— Missing minority representation should be acknowledged as a limitation affecting external validity.

Key distinction: A trial that enrolls 95% white participants and reports efficacy in "all patients" is not technically a CONSORT violation if demographics are reported — but the generalizability statement (item 21) must acknowledge limited external validity. Step 3 may frame this as a disparities/equity question, expecting the answer to invoke equity reporting.

Reporting guidelines increasingly require demographic and equity disclosures — high yield in modern Step 3 testing.

CONSORT-Equity 2017 extension and PRISMA-Equity 2012:

Pediatric and pregnancy reporting:

Elderly and frailty:

Renal/hepatic impairment subgroups:

Underrepresented groups:

Special Populations — Patient-Reported Outcomes, AI, and Pragmatic Trials

— Identify PRO as primary/secondary in abstract

— Describe PRO hypothesis, instrument validation, mode of administration, scoring, handling of missing PRO data

— Highly relevant for oncology, rheumatology, mental health trials

— Describe algorithm version, input data handling, output interpretation, integration into care pathway, human-AI interaction, error case analysis, performance across subgroups

— Step 3 increasingly includes vignettes on diagnostic AI tools — CONSORT-AI applies to interventional trials of these tools.

— Describe why the trial is pragmatic along the PRECIS-2 dimensions (eligibility, recruitment, setting, organization, flexibility of delivery/adherence, follow-up, primary outcome, primary analysis)

— Pragmatic trials approximate real-world care; explanatory trials maximize internal validity.

— Define population using database codes (ICD, CPT), validate codes, address data linkage, cleaning, and missingness specific to claims/EHR data

— Increasingly important as real-world evidence (RWE) drives FDA decisions.

Board pearl: A vignette describing an EHR-based comparative effectiveness study of two diabetes drugs should invoke STROBE + RECORD, not CONSORT. Tip-off phrases: "extracted from the Optum/Marketscan/Medicare claims database," "ICD-10 codes used to identify cases," "propensity-score matching."

CONSORT-PRO (2013): for trials using patient-reported outcomes (PROs) as primary or key secondary endpoints

CONSORT-AI and SPIRIT-AI (2020): for trials of AI-based interventions

CONSORT extension for pragmatic trials (2008):

CONSORT extension for nonpharmacologic treatments: surgery, behavioral interventions, devices — must describe operator experience, learning curve, standardization, fidelity.

STROBE-MR (Mendelian randomization): genetic variants as instrumental variables — must report on three core assumptions (relevance, independence, exclusion restriction).

RECORD (2015): routinely collected health data

Complications — Consequences of Poor Reporting

— Wasteful research: Chalmers and Glasziou estimated ~85% of biomedical research is avoidably wasted, much due to poor reporting precluding use.

— Spurious meta-analytic conclusions: if primary studies underreport methods, downstream PRISMA syntheses overstate certainty.

— Guideline misadoption: practice guidelines built on incompletely reported trials may codify ineffective or harmful interventions (e.g., historical examples in hormone therapy, antiarrhythmic prophylaxis).

— Selective outcome reporting (SOR): primary outcome silently swapped to a significant secondary outcome — detectable by comparing publication to registered protocol (CONSORT item 23).

— Publication bias: "negative" trials remain unpublished — detectable by funnel plot asymmetry, Egger test, trim-and-fill in meta-analyses.

— Spin: discussion sections overinterpret non-significant primary outcomes by emphasizing secondary endpoints.

— Salami slicing/duplicate publication: detected by careful citation and registration cross-checks.

— ICMJE requires prospective registration as a condition of publication.

— FDAAA 801 mandates results reporting on ClinicalTrials.gov within 1 year of trial completion; noncompliance is publicly tracked and can result in civil monetary penalties.

— Retractions for fabrication, falsification, or duplicate publication are tracked by Retraction Watch.

Step 3 management: If a manuscript shows a primary endpoint differing from its ClinicalTrials.gov registration without explicit justification, the appropriate critique is selective outcome reporting, and the recommendation is to request the original protocol and statistical analysis plan before clinical adoption.

Direct clinical harms from inadequate reporting:

Specific bias patterns linked to reporting failures:

Regulatory consequences:

Reproducibility crisis: poorly reported preclinical and clinical research underlies the broader reproducibility concern in biomedicine.

When to Escalate — From Reporting Critique to Action

— Missing exact randomization block size, unclear baseline imbalances — write to authors via journal correspondence.

— Unblinded outcome assessment of a subjective outcome (e.g., pain), missing per-protocol sensitivity analysis, incomplete confounder adjustment in STROBE → downgrade certainty in GRADE; reflect in clinical decisions with caveats.

— No allocation concealment in unblinded trial with subjective outcomes, undisclosed conflicts of interest with sponsor analysis control, post hoc primary outcome switch without justification, meta-analysis pooling studies of incompatible designs.

— Evidence of data fabrication or falsification → notify the journal editor and, for U.S. federally funded research, the Office of Research Integrity (ORI).

— Plagiarism or duplicate publication → editor + COPE (Committee on Publication Ethics) flowcharts.

— IRB ensures ethical conduct prospectively but is not the venue for post-publication reporting concerns.

— Research integrity officer (RIO) at the home institution receives misconduct allegations.

— PubPeer and post-publication peer review platforms have catalyzed many investigations.

CCS pearl: In a CCS-flavored research-integrity scenario, the order of operations is: (1) document the concern objectively, (2) discuss with the principal investigator if appropriate, (3) report to the institutional RIO or department chair, (4) escalate to ORI if federally funded — never go directly to media or social platforms first.

Reporting deficiencies should trigger graded responses analogous to clinical escalation:

Minor deficiency (request clarification):

Moderate deficiency (cautious interpretation):

Major deficiency (do not apply to practice):

Critical deficiency (consider reporting/retraction):

Institutional roles:

Clinician role: when a guideline change is proposed based on a single trial, scrutinize CONSORT compliance and registry concordance before adoption; advocate via specialty society channels.

Key Differentials — Same-Category Reporting Tools

— CONSORT = trial reports (the publication)

— SPIRIT 2013 = trial protocols (before the trial starts)

— Both are managed by the same group; SPIRIT covers items investigators must specify prospectively (sample size justification, analysis plan, data management, monitoring committees).

— PRISMA originated for systematic reviews of RCTs but PRISMA 2020 covers any systematic review (including observational).

— MOOSE (2000) specifically for meta-analyses of observational studies — still cited but largely subsumed by PRISMA.

— PRISMA = reporting checklist (did the authors describe what they did?)

— AMSTAR-2 and ROBIS = methodological quality / risk-of-bias assessment of systematic reviews (did they do it well?). Step 3 distractor pairs.

— STROBE = reporting standard for observational studies

— NOS and ROBINS-I = risk-of-bias tools for observational studies, used within a PRISMA-guided systematic review to appraise included studies.

— CONSORT = reporting

— RoB 2 (2019) = current standard for RCT risk-of-bias appraisal in Cochrane reviews (domain-based: randomization, deviations, missing outcome data, measurement, selection of reported result). Jadad scale is older, simpler, largely retired.

Key distinction: "How should authors report their trial?" → CONSORT. "How should we assess bias in this trial for our systematic review?" → RoB 2. Step 3 frequently swaps these terms; reading the verb in the stem is decisive.

Within the "reporting guideline" category, distinguish closely related instruments:

CONSORT vs SPIRIT:

PRISMA vs MOOSE:

PRISMA vs AMSTAR-2 / ROBIS:

STROBE vs Newcastle-Ottawa Scale (NOS) / ROBINS-I:

CONSORT vs Cochrane RoB 2 / Jadad:

Key Differentials — Other-Category Frameworks

— Rates certainty of evidence (high/moderate/low/very low) and strength of recommendation (strong/weak).

— Downgraded by risk of bias, inconsistency, indirectness, imprecision, publication bias; upgraded by large effect, dose-response, plausible confounding biasing against the observed effect.

— Used by virtually all major guideline organizations (WHO, ACP, NICE, AHA/ACC).

— Traditional pyramid: meta-analyses/SRs → RCTs → cohort → case-control → case series → expert opinion.

— Modern view emphasizes quality of individual study (a well-done cohort may exceed a flawed RCT).

Board pearl: A vignette about a new sepsis prediction algorithm built from EHR data should invoke TRIPOD+AI, not CONSORT or STROBE. If that algorithm is then tested in an RCT comparing alert-on vs alert-off arms, the trial uses CONSORT-AI.

EBM / evidence synthesis frameworks distinct from reporting guidelines:

GRADE (Grading of Recommendations, Assessment, Development and Evaluations):

AGREE II / RIGHT: clinical practice guideline appraisal (AGREE) and reporting (RIGHT) — for the guideline document itself, not the underlying research.

PICO/PICOTS framework: structures clinical questions (Population, Intervention, Comparator, Outcome, Timing, Setting) — used to define eligibility in PRISMA reviews.

Hierarchies of evidence:

Quality improvement reporting (SQUIRE 2.0): for QI projects — distinct from research; emphasizes context, intervention, study of intervention, measures, ethical considerations, results, discussion. Hospital QI projects published in journals should use SQUIRE.

Case report reporting (CARE): structured case report template — distinct from observational studies.

TRIPOD / TRIPOD+AI (2024): prediction model studies — derivation and validation of risk scores (e.g., new cardiovascular risk calculators). Increasingly tested.

Secondary Prevention — Embedding Reporting Standards Into Practice

— Register every prospective study before enrolling the first participant (ClinicalTrials.gov for trials, PROSPERO for systematic reviews, OSF/Registry of Studies for observational research where feasible).

— Write the protocol using SPIRIT (for trials) or a predefined protocol/SAP for observational studies.

— Maintain a CONSORT/STROBE/PRISMA checklist as a living document during the project; many journals require submission of the completed checklist with the manuscript.

— Use the checklist to structure your review; flag missing items as required revisions.

— Cross-check registered outcomes against reported outcomes for selective reporting.

— Verify flow diagram numbers reconcile internally and with the text.

— Adopt a structured critical appraisal habit: use JAMA Users' Guides, CASP checklists, or guideline-specific summaries.

— Maintain skepticism for unregistered, single-center, industry-sponsored trials with surrogate primary outcomes.

— Step 3 EBM blocks reward recognizing the guideline-design pairing and the bias most likely missed.

— Practice by reading abstracts and naming the applicable guideline + one likely deficiency.

— Many top journals (JAMA, NEJM, BMJ, Lancet, Annals) require completed reporting checklists at submission and post adverse-event reporting compliance for trials.

Step 3 management: When asked "What is the single best step to improve transparency of this planned trial?" the answer is almost always prospective registration on ClinicalTrials.gov before enrolling participants, paired with publication of the protocol.

For clinician-investigators, prospective adoption prevents reporting failures:

For peer reviewers and editors:

For clinician-readers:

For trainees:

For institutions and journals:

Follow-Up, Monitoring Parameters, and Educational Counseling

— CONSORT 2010 is the dominant version; CONSORT 2025 is the major revision (be aware of expanded equity, registration, harms, and statistical methods items).

— STROBE 2007 remains current; updates ongoing.

— PRISMA 2020 is current; replaced PRISMA 2009. Key changes: 27 items (vs 27 in 2009 but reorganized), revised flow diagram with separate database/register vs other sources, more emphasis on risk of bias and certainty of evidence.

— Authors should complete the checklist with page numbers referencing where each item is addressed.

— Many journals request the checklist as a supplementary file (e.g., Research Square, BMJ, PLOS).

— Emphasize that guidelines are minimum standards, not ceilings — exceptional studies will exceed them.

— Reporting compliance is necessary but not sufficient for high-quality evidence; well-reported but poorly designed studies remain low quality.

— Many CME-accredited journal clubs use CONSORT/STROBE/PRISMA as the critical appraisal scaffold.

— Statistical reviewers increasingly required for trials and meta-analyses at major journals.

— Lay summaries and plain-language abstracts are now required by many funders (NIH, Wellcome) — extension of transparency principles.

Board pearl: When in doubt about which version of a guideline applies, the Step 3 answer favors the most recently updated version named in the question; if unspecified, default to CONSORT 2010, STROBE 2007, PRISMA 2020.

Versioning awareness: reporting guidelines are updated periodically; recognize the current version for Step 3:

Track via EQUATOR Network (equator-network.org) — single hub for >500 guidelines, extensions, translations.

Monitoring at the manuscript level:

Counseling colleagues and trainees:

Continuing professional development:

Public engagement:

Ethical, Legal, and Patient Safety Considerations

— Informed consent disclosure: CONSORT and STROBE require reporting that ethical approval was obtained and consent procedures described. If a trial enrolls patients unable to consent (e.g., emergency research under FDA exception from informed consent, 21 CFR 50.24), this must be explicitly reported with community consultation procedures.

— Vulnerable populations: prisoners, children, pregnant individuals, and cognitively impaired adults have additional federal protections (45 CFR 46 Subparts B, C, D). Reporting should clarify safeguards used.

— ICMJE form discloses financial and nonfinancial conflicts.

— CONSORT item 25 and PRISMA item 26 require funding source and role of funders in design, analysis, decision to publish — industry-sponsored trials with sponsor-controlled analysis warrant heightened scrutiny.

— ICMJE 4-criterion definition (substantial contribution, drafting/revising, final approval, accountability).

— Ghost authorship (uncredited industry writers) and gift authorship are ethical violations.

— Harms underreporting is pervasive; CONSORT-Harms (2004) and PRISMA-Harms exist to counter this. Step 3 may show a trial reporting only efficacy without serious adverse events — flag as a CONSORT item 19 violation.

— Adverse event signals in registry data require timely communication (FDA MedWatch).

— ICMJE since 2018 requires a data sharing statement in trial publications.

— Patient privacy (HIPAA in U.S., GDPR in EU) constrains data sharing — must be balanced with transparency.

Step 3 management: A drug trial showing efficacy but omitting that 4 participants died in the treatment arm and 1 in placebo without statistical comment is a CONSORT-Harms violation and an ethical safety reporting failure — flag to the journal editor and consider FDA notification if the drug is marketed.

Research ethics intersect directly with reporting:

Conflicts of interest (COI):

Authorship integrity:

Patient safety in reporting:

Data sharing and privacy:

Transition-of-care analog: When trial participants transition to routine clinical care at study end, post-trial access to effective interventions is a Declaration of Helsinki obligation; reporting should describe post-trial care.

High-Yield Associations and Rapid-Fire Clinical Facts

Board pearl: If you must memorize one cross-table for Step 3 — RCT/CONSORT, observational/STROBE, systematic review/PRISMA — and pair each with its risk-of-bias partner (RoB 2, ROBINS-I, AMSTAR-2). That single matrix answers most questions on this topic.

CONSORT 2010 = 25 items + flow diagram; CONSORT 2025 = updated, expanded.

STROBE 2007 = 22 items (18 common + 4 design-specific); no required flow diagram.

PRISMA 2020 = 27 items + abstract checklist + flow diagram tracking records.

SPIRIT 2013 = 33 items, for trial protocols.

PROSPERO registers systematic reviews; ClinicalTrials.gov registers trials.

EQUATOR Network is the central hub for reporting guidelines.

GRADE ≠ reporting guideline; it rates certainty of evidence.

RoB 2 (RCTs) and ROBINS-I (non-randomized) are risk-of-bias tools, not reporting tools.

AMSTAR-2 = quality appraisal of systematic reviews (methodology); PRISMA = reporting of systematic reviews (transparency).

TRIPOD/TRIPOD+AI = prediction models; STARD = diagnostic accuracy; CHEERS = economic evaluation; SQUIRE 2.0 = QI; CARE = case reports.

CONSORT-AI and SPIRIT-AI address AI interventions (2020).

RECORD extends STROBE for routinely collected health data.

PRISMA-NMA for network meta-analysis; PRISMA-IPD for individual patient data; PRISMA-DTA for diagnostic test accuracy; PRISMA-ScR for scoping reviews.

Intention-to-treat is the default analysis for superiority trials; per-protocol is co-primary in non-inferiority.

Allocation concealment prevents enroller foreknowledge; blinding prevents post-allocation knowledge.

Funnel plot asymmetry / Egger test screens for publication bias (≥10 studies).

I² >50% indicates substantial heterogeneity.

ICMJE mandates prospective trial registration as a publication prerequisite.

FDAAA 801 mandates results reporting on ClinicalTrials.gov.

Declaration of Helsinki governs ethics of human-subjects research globally.

Board Question Stem Patterns

— Stem describes a multicenter double-blind RCT and asks which reporting framework applies. Answer: CONSORT.

— Stem describes a Cochrane-style pooled analysis of 12 trials. Answer: PRISMA.

— Stem describes a 10-year cohort of postmenopausal women followed for incident MI. Answer: STROBE.

— Trial doesn't describe how the allocation sequence was concealed from the enrolling clinician. Answer: allocation concealment (CONSORT item 9).

— Observational study doesn't list confounders or adjustment methods. Answer: confounder reporting (STROBE item 7/12).

— Meta-analysis doesn't include a search strategy. Answer: PRISMA item 7.

— "Which tool should reviewers use to assess methodological quality of the included RCTs?" Answer: Cochrane RoB 2 — not CONSORT.

— Published primary outcome differs from ClinicalTrials.gov registration. Answer: selective outcome reporting; check protocol/SAP.

— Superiority trial with 18% loss to follow-up reports per-protocol only. Answer: intention-to-treat should be primary; this is a CONSORT item 16 deficiency introducing attrition bias.

— Trial enrolls 92% non-Hispanic white participants and generalizes broadly. Answer: limited external validity; CONSORT item 21 / equity extension.

— I² = 78%. Answer: substantial heterogeneity; explore via subgroup/sensitivity analyses; consider whether pooling is appropriate (PRISMA item 13d).

— Trial of a sepsis prediction algorithm. Answer: CONSORT-AI.

Step 3 management: When two answer choices both seem to fit (e.g., "CONSORT" and "Cochrane RoB 2"), reread the verb: report → CONSORT; assess bias / appraise quality → RoB 2. Verb discrimination is the high-yield decoder.

Pattern 1 — Identify the guideline:

Pattern 2 — Identify the missing element:

Pattern 3 — Distinguish reporting from bias assessment:

Pattern 4 — Selective outcome reporting:

Pattern 5 — Choosing analysis population:

Pattern 6 — Equity / generalizability:

Pattern 7 — Heterogeneity in meta-analysis:

Pattern 8 — AI tool study:

One-Line Recap

CONSORT reports randomized trials, STROBE reports observational studies, and PRISMA reports systematic reviews/meta-analyses — together they form the transparency backbone of evidence-based medicine, complemented by but distinct from risk-of-bias tools (RoB 2, ROBINS-I, AMSTAR-2) and certainty-of-evidence frameworks (GRADE).

— CONSORT 2010 (25 items + participant flow diagram) for RCTs; key items: sequence generation, allocation concealment, blinding, ITT analysis, harms, registration. Extensions for cluster, non-inferiority, pragmatic, harms, PRO, AI, equity.

— STROBE 2007 (22 items) for cohort, case-control, cross-sectional; key items: source population, exposure/outcome/confounder definitions, adjusted estimates with 95% CIs. RECORD extends STROBE for routinely collected EHR/claims data.

— PRISMA 2020 (27 items + abstract checklist + records flow diagram) for systematic reviews and meta-analyses; key items: PICO eligibility, full search strategy, risk-of-bias assessment of included studies, heterogeneity (I²), publication bias, GRADE certainty, PROSPERO registration. Extensions for NMA, IPD, DTA, scoping reviews.

— Distinguish reporting from quality: Reporting guidelines say what to disclose; RoB 2, ROBINS-I, AMSTAR-2, GRADE assess how good the underlying methods/evidence are. On Step 3, the verb in the stem ("report" vs "assess bias" vs "rate certainty") chooses the correct tool. EQUATOR Network is the master hub; ICMJE + FDAAA 801 mandate prospective registration; ClinicalTrials.gov and PROSPERO are the registries. Prospective registration + adherence to the relevant checklist + transparent disclosure of conflicts, funding, harms, and equity considerations is the durable answer to almost any Step 3 EBM-integrity stem.

Board pearl: Memorize the triad — CONSORT-trial, STROBE-observational, PRISMA-review — and the cross-walk to RoB 2 / ROBINS-I / AMSTAR-2 for bias appraisal. That single mental table answers the overwhelming majority of reporting-guideline questions encountered on the USMLE Step 3.

Tight recap bullets: