Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› nejm-study-design
nejm-study-design
GitHub用于在撰写NEJM临床论文前，确认研究设计的严谨性及强制性的前瞻性试验注册。检查方案、统计计划及设计要素，确保符合ICMJE政策，避免注册缺失等致命缺陷。
触发场景
撰写临床试验或观察性研究报告时
需确认试验是否在入组前完成注册
需核实研究设计（随机化、盲法、终点）的严谨性
缺乏最终版研究方案或统计分析计划
安装
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill nejm-study-design -g -y
SKILL.md
Frontmatter
{
"name": "nejm-study-design",
"description": "Use to confirm study-design rigor and the mandatory prospective trial registration, protocol, and statistical analysis plan before writing up a clinical study for NEJM. Surfaces registration problems early — they cannot be fixed retroactively."
}
Study Design & Registration (nejm-study-design)
When to trigger
- A trial is being written up and you must confirm it was registered before enrollment.
- There is no finalized protocol or statistical analysis plan to submit.
- The design (randomization, blinding, endpoints) needs to be stated rigorously for the Methods.
- An observational study needs its design (cohort / case-control / cross-sectional) named and justified.
The non-negotiable: prospective trial registration
Per ICMJE policy (which NEJM enforces), a clinical trial must be prospectively registered in a public registry — ClinicalTrials.gov or a WHO ICTRP primary registry — before the first patient is enrolled. This is a deal-breaker, not a formatting detail.
- The trial registration number (e.g., an NCT number) is reported in the abstract and Methods.
- Registration must predate enrollment; retrospective registration is generally disqualifying for an Original Article.
- The registered primary outcome must match the reported primary outcome — discrepancies are a top reviewer flag.
If a trial was never prospectively registered, raise it now. It changes venue and framing and cannot be repaired by writing.
Protocol and statistical analysis plan (SAP)
For trials, NEJM expects the full trial protocol and the statistical analysis plan to be submitted (typically as a supplement) and made available to reviewers.
- The SAP must be pre-specified — finalized (and dated) before unblinding / database lock.
- Any deviation from the registered protocol or SAP must be disclosed and explained in the manuscript.
- A CONSORT-aligned trial write-up depends on these documents existing (see
nejm-reporting).
Design rigor checklist (trials)
- Randomization — method (e.g., permuted blocks, stratification factors) stated.
- Allocation concealment — mechanism described (central randomization, sealed envelopes).
- Blinding — who was blinded (participants, clinicians, outcome assessors, analysts); if open-label, justify and address bias.
- Primary endpoint — single, pre-specified, patient-important, with a clear definition and time point.
- Secondary endpoints — pre-specified and ordered; multiplicity handled (see
nejm-statistics). - Sample size / power — assumptions, effect size, alpha, power stated; not post hoc.
- Analysis population — intention-to-treat as primary; per-protocol as sensitivity.
- Interim analyses / stopping rules — DSMB, pre-specified boundaries (e.g., O'Brien-Fleming) if used.
Observational studies
- Name the design: cohort (prospective/retrospective), case-control, cross-sectional, or nested designs.
- Pre-specify the primary exposure and outcome; define confounders and the adjustment strategy a priori.
- Address selection bias, information bias, and confounding explicitly; use cautious, non-causal language.
- Report per STROBE (see
nejm-reporting); registration of observational protocols is encouraged though not mandated.
Operating pass for New England Journal of Medicine
Use this as a second-pass capability check. First lock the clinical question, population, endpoint, effect size, safety signal, and practice implication; then test whether the manuscript addresses clinical-medicine reviewers who expect practice-changing evidence, patient relevance, safety, and exact reporting discipline.
- Primary move: Return a claim-evidence-risk ledger; every recommendation must point to a manuscript location or missing artifact.
- Decision ledger: return
claim / evidence / blocker / next editrows so the next pass can patch the manuscript directly. - Neighbor test: compare against JAMA for broad clinical medicine, Lancet for global-health/public-health reach, specialty journals for narrower disease domains; if the neighboring outlet has the stronger audience claim, recommend re-routing before polishing.
- Verification floor: before submission-ready advice, re-open
resources/official-source-map.mdfor volatile rules and name the one unresolved fact that could change the recommendation.
Output format
【Study type】 RCT / other interventional / cohort / case-control / cross-sectional / SR-MA
【Registration】 registry + number + registered BEFORE enrollment? yes/no/UNREGISTERED-FLAG
【Registered vs reported primary outcome】 match? yes/no
【Protocol + SAP】 available, pre-specified, dated? yes/no
【Design rigor gaps】 randomization / concealment / blinding / endpoints / power / ITT
【Protocol/SAP deviations to disclose】 [...]
【Next】 nejm-reporting
Anti-patterns
- Do not write up an unregistered trial as an Original Article without flagging it to the user.
- Do not present a post-hoc analysis plan as if it were pre-specified.
- Do not report a primary outcome that differs from the registered one without disclosing the change.
- Do not treat per-protocol as the primary analysis for a superiority trial — ITT is primary.
版本历史
- 1839142 当前 2026-07-05 14:06
