Protocol Walkthrough

1. From the protocol workflow, open Generate SAP or consent. This requires the latest generated protocol, unchanged source versions and reconciled source evidence anchors; a bare starter protocol is not sufficient.
2. Choose the family, enter its scope/site/language, and complete each authoring field with a source/version/page or decision reference. SAPs require separate model, adjustment, contrast, diagnostics, sensitivity, population, derivation/visit-window, intercurrent-event, multiplicity and output specifications for every endpoint. Consent requires source-linked participant-facing wording for every protocol arm, visit and procedure, site-specific language and explicit applicability decisions for genetics, future use, representatives and registry disclosure. Upload supporting specifications into the study, then use each input’s source-passage picker to link an exact quote. Identical input/source text can be linked in one batch. Exact text identity does not establish clinical support.
3. For a synthetic protocol, Load fictional walkthrough inputs supplies clearly fictional examples. Preview every generated section, acknowledge the review, then create the draft and handoff. The resulting document opens in the existing editor, review and Word/PDF export workflow.
4. Download author inputs for safekeeping before leaving unfinished work. Restore that file later, or reuse retained inputs from a published draft. After an amendment, reuse a retained draft, inspect changes to protocol fields and the current prior document, and record a retain/revise/retire decision with a reason for every input. Choose Compare reviewer edits with regenerated draft. Compare original, current and proposed sections, then choose a conflict-free combination, keep reviewer content, or use regenerated content. Give a reason for every section; overlapping block edits require an explicit choice, and added sections must be kept or explicitly retired. Required sections cannot be retired. Bound evidence objects cannot be copied into a new document through this path. Publication rejects a comparison if someone edits the prior draft in the meantime. The new handoff is linked below the prior one. In Evidence review and amendment work, an independent reviewer records a supported, insufficient or contradicted decision and rationale for every input. Draft edits invalidate earlier support approvals. Unresolved findings block exports; missing current support decisions block release. Protocol and source revisions automatically queue amendment work: claim it, reconcile retained inputs, publish a replacement, and review that replacement. Accept reviewed replacement completes the queued work and its replacement handoff atomically; predecessor documents and their handoff decisions remain retained. Verified drafts with changed protocol or study-source baselines require reconciliation before new exports.

1. Open a returned Word section and choose Review returned edits. Supported Word 2013 comment threads show each reply, its parent comment, inherited anchor and Word resolved status. Even a thread marked resolved in Word requires an explicit addressed/not-applicable response to every comment and reply.
2. Submit the responses as a separate review suggestion. The original bytes, thread relationships and responses remain retained. Modern durable-ID metadata, rich comment objects and exporting responses back to Word remain unsupported; the importer refuses unsupported structures rather than silently discarding them.

1. Use Queue dataset preparation to retain sources and derivations in a background job. Follow Background jobs and reusable domain mappings for queued, running, succeeded, failed or cancelled state. Jobs are bounded to the current 5,000-row and 750 KB source/recipe limits; this is not a large-file import adapter. You may leave the page; you are notified when the job finishes or fails. The scheduled worker recovers interrupted work; failed jobs can be retried within three attempts, or cancelled before retention.
2. Complete a dataset recipe, then save a reusable firm mapping profile with its key, domain, purpose and scope. Profiles retain mappings, joins, window and missing-data rules without source CSV contents. Apply a saved version to another dataset and supply the required files. Profile changes do not alter existing datasets. Detach the profile before intentionally changing its recipe. ADaM BDS-style and SDTM VS-style starters suggest column conventions only; they do not certify conformance or infer clinical selections.
3. After independent approval of the retained dataset, choose Queue retained dataset execution. Logistic binary-outcome and Poisson count models support treatment-only or one-baseline adjustment. These numerical implementations are not qualified for regulatory use: they run only on a dataset marked synthetic, are refused on study data, and any result appendix carries the label “Not qualified for regulatory use”. Poisson assumes equal exposure; no robust covariance, repeated-measures or survival model is included. The job retains results with convergence diagnostics and model limitations.
4. Once execution succeeds, select its receipt in Attach retained execution results when generating or reconciling a SAP. GLM appendices show the log-scale contrast, Wald interval and p-value, exponentiated odds or expected-count ratio, baseline adjustment and dispersion. Dataset supersession or withdrawn approval still blocks use of stale results. Job completion does not constitute clinical review or SAP release.

1. Open Dataset mappings, derivations and versions. Load the synthetic mapping fixture or upload a rectangular CSV. Recognized column aliases are suggested; ambiguous names and treatment codes require explicit choices. Choose observed value, change from baseline or percent change, and an explicit unit scale. Complete provenance, data cut, coded-identifier attestation and rationale.
2. Preview mapped dataset shows counts and the first ten source-row derivations. Review and acknowledge the recipe, then Retain dataset version. A different active reviewer, QC member or administrator must use Inspect mapping and review status and approve the exact retained version with a rationale. The author cannot approve their own mapping. Then execute a retained dataset by choosing its endpoint, method, alpha and rationale; the server recomputes the rows from retained source and mapping. Attach its execution receipt to a SAP through the result appendix builder.
3. Prepare replacement loads a current version. Download source CSV if needed, upload revised CSV, review retained mappings and provide a replacement rationale. Supersession flags previous runs and SAP drafts, blocks stale exports, invalidates dependency/signoff fingerprints and queues amendment work. Previous versions remain downloadable. Reuse affected draft inputs, rerun current data and reconcile every result appendix before independent review.
4. Use Load synthetic joined fixture to try a primary visit file plus subject lookup. Add up to three lookup CSVs with unique keys and explicit aliases. Add composite key pairs for up to four key columns per side. Exact visit and population filters remain available. Load synthetic window and missing-data fixture demonstrates composite lookup, nearest-target selection and explicit exclusions. Preview retains joined source-row numbers, excluded primary rows and unused lookup rows. Every primary row must match; ambiguous joins fail. Bounds: 4 to 5,000 rows and 100 columns per CSV, 750 KB combined source/recipe; 50 dataset keys and 100 versions per key. Window rules select the nearest target within inclusive bounds by one to three identity columns, using integer study days or complete ISO dates relative to a reference date. Date offsets use day zero at the reference date. Choose reject, earlier or later for equidistant visits; same-day duplicates always fail. Window selection runs before missing-data rules and never falls back to another visit. Declare exact missing tokens, then reject, exclude or replace outcome/baseline with an explicit source-unit constant and rationale. Preview retains missing originals and replacement decisions. Constants require statistical review; this is not qualified multiple imputation. No many-to-many joins, inferred windows, carry-forward models or arbitrary formulas. Approval is an independent workflow decision, not clinical qualification. Rejection blocks dependent results and exports; replacement requires fresh approval. Standalone execution snapshots outside this version workflow do not track dataset supersession.

1. From Protocol → SAP and consent, expand Execute statistical models. Choose an endpoint and Welch, paired t or ANCOVA with treatment plus one baseline covariate. Load a synthetic dataset or prepare an imported dataset specification. For imported snapshots, complete name, provenance reference, data cut, deidentification attestation and rationale in the JSON. Import canonical CSV rows (id,group,outcome; optional baseline,pair), or execution JSON under 750 KB. Inspect every row and the method rationale.
2. Execute and retain result stores the input, numerical output, input digest, engine and protocol baseline. Download the reproducibility receipt from the latest 20 executions. Review assumptions, confidence intervals, p-values and ANCOVA residuals independently. Earlier results are marked stale after relevant protocol/source changes. Under Attach retained execution results, select current runs to generate separate SAP result appendices. Results are fetched from retained receipts on the server; the draft still needs independent methods, outputs and software review. Amendments must regenerate or retire old result appendices. The workspace does not qualify real datasets, execute arbitrary code, impute missing values or approve clinical conclusions.

1. Select US adult, UK adult or other/site-defined review explicitly. Supply the applicability determination and exact supporting site-source passages. US prompts separate HHS/FDA/HIPAA applicability; UK prompts distinguish participation consent from the basis for processing data. Pediatric, emergency, capacity and nation/state-specific questions require specialist review.
2. Complete the profile-specific participant summary, privacy and local language. The internal applicability determination stays in the retained input receipt and is not printed in the participant form. Changing profiles requires new profile-specific inputs; earlier drafts remain retained. These profiles are authoring prompts, not legal or ethics approval.

1. Open a generated document and choose Word section handoff. Prepare the current supported prose or rectangular-table section, download the retained copy and edit it in Word.
2. Return the file against its matching handoff. Choose Review returned edits. Accept or reject every inline insertion/deletion and give each a reason. For basic anchored Word comments, read the comment and quoted anchor, then record Addressed or Not applicable with a response rationale. Then create a separate review suggestion. Compare the full suggestion, including untracked edits, against its retained baseline before applying it. Word author names are unverified. Threaded/extended or unanchored comments, formatting revisions, moves, structural changes and altered protected evidence require manual reconciliation; the original file remains retained.

Use a separate test study and synthetic data only. You need a writer and a different reviewer in the same organization; an administrator is needed to publish a sponsor template. Draft exports are not clinical approval or signed records.

Download the synthetic exercise from the Protocol walkthrough panel or /examples/protocol-walkthrough-v1.zip. Extract it on your computer. The starter covers a bounded two-arm draft exercise; it is not a clinically qualified protocol or evidence of complete statistical/source coverage.

1. Create study SYN-P01. Upload only baseline/inputs. Check Protocol, SAP and Other file kinds and wait until all three documents are ready. Keep draft-design.json outside Sources.
2. If parsing fails, inspect the source error, retry from the file list or upload a corrected file. Ready means extraction finished, not that the extraction is complete or approved.
3. Use Create, choose Clinical trial protocol and review the creation source baseline. Create the draft, then open Protocol design, generation and amendments.

1. Open Import a protocol draft package and select baseline/draft-design.json. Map each named file to its ready uploaded version. Compare the import with the form, inspect the changes and explicitly choose the fields to apply.
2. Review the imported inputs, narrative content, source links and applicability. Imported applicability is pending. In M11 hierarchy, use Review applicability across authored headings to select content you reviewed and record your author decisions. Conditional headings and source checks still apply.
3. For a sponsor exercise, upload template/synthetic-sponsor-template.docx as Template before baseline generation. An administrator publishes a release in Trial schema, sponsor templates and study checks. Select that release, load Sponsor template body, start saved mapping, and include or exclude every parsed block with a rationale. Review mapped destinations and any placeholders.
4. For full source qualification, enable clinical/statistical specifications, calculations/coverage and retained source findings. These require explicit source comparisons, calculation assumptions, claim spans and review decisions. The sample does not fill or approve these automatically; unresolved contradictions remain blocking.

1. Save design version. Use the preview and the design-check list to resolve missing inputs; then generate the draft. A modified starter document must be reconciled or replaced before first structured generation.
2. Open Assign independent review. The author assigns a separate reviewer, who inspects the exact draft, records feedback and completes the review. Resolve comments/suggestions and close the round before an amendment.
3. Use Download draft Word and Download draft PDF. Open both files and inspect headings, table cells, references, page breaks, margins and missing or clipped content. If download fails, the panel shows the error and retains the design; refresh the saved protocol if another person changed it.

1. After the baseline is generated, upload amendment/inputs. These become new source versions. Start a protocol amendment with unused version 2.0 and a rationale.
2. For the synthetic exercise, import amendment/draft-design.json as a comparison and map the new protocol/SAP versions plus the original safety context. Select only intended changes; preserve any extra authored findings or template work and review applicability again. Alternatively edit the design manually. Reconcile the new source versions explicitly. Move final endpoint timing and the final visit from Day 28 to Day 35, set duration to 35 days, and review affected schedules, burden, safety/consent implications and source links. Record affected sections, descriptions and reasons in the amendment summary.
3. Save and inspect Amendment reconciliation. Choose and explain every required section disposition. Generate the reconciled amendment; prior content and decisions remain in history.
4. Assign affected SAP/source/document follow-ups to accountable owners. A separate reviewer records the six amendment-impact decisions and independent dependency completion. Changed evidence or roles can stale earlier decisions.
5. Repeat independent review and both draft downloads for version 2.0. QC and sign-off remain blocked until their actual requirements are satisfied; completing the exercise does not certify launch readiness.

1. Open Amendment work across studies from a protocol or family workspace to see registered revision-specific work across your firm. Filter by cross-study, assigned work, review-ready work or retained history. Owners accept and submit work; independent reviewers close it with rationale. Changed root sources, protocol content, target content or ancestor dependencies create new work and supersede old queue snapshots. Historical closures without a pinned root-source revision require fresh review. This queue covers registered dependencies and is bounded to 100 root protocols; it does not infer missing clinical relationships. Use Discover related study impacts and Scan related studies to find candidate documents from matching compounds, related indication terms, study-code references, identical extracted passages and prior regeneration. Related protocols open for review but cannot be registered as regeneration targets. Inspect the evidence; a candidate is not an applicability decision. Choose Review and register candidate, or create a downstream draft with sections in another synthetic study in the same firm. Register it in the protocol dependencies with an owner and a cross-study applicability rationale. Incoming work appears in that study's handoff queue.
2. Open Regenerate mapped downstream sections, then Load section comparison. Select protocol sections for the downstream sections you intend to replace, or choose Reuse available saved mappings after a later amendment. Removed source sections require explicit replacements. If downstream text was edited since regeneration, compare it and explicitly acknowledge replacement. Leave other sections at Keep existing content. Compare both texts, explain the mapping and apply. For whole-document accounting, enable Review every section in a whole-document regeneration plan. Give every unmapped section a retention reason. Apply whole-document plan saves all replacement and retention decisions together; inspect them in Whole-document plan history. This does not create missing family-specific content or certify scientific completeness. Check in mapped sections first; active editing locks and open reviews prevent replacement.
3. Open the downstream document to complete its family-specific authoring and review. The mapping retains a before/after receipt but does not generate a complete SAP or consent document. Bound evidence and scientific objects must be reconciled explicitly. A different reviewer must close the dependency; stale previews require reloading.
4. For running-matter qualification, download the first/even-page sponsor example from the walkthrough. Publish and map it as a template, export a multi-page draft and compare Word/PDF first-page, even-page and default headers/footers. Inspect fonts, margin clearance and page counters; For table-based running matter, use the sponsor running-table example. Its bounded columns, horizontal spans, borders and shading are supported; excessive header/footer height refuses PDF export. The formatted sponsor body-table example also preserves supported direct cell shading, alignment, borders and repeating headers through editing and Word export. Reload and reconcile older mappings if their source structure changes after parsing improvements. Complex multi-section layouts and images remain a separate qualification task.

Record the step, visible error, document/version, expected result and actual result. Share synthetic screenshots or files with the project team. Source truth, complex tables, complete sponsor fidelity and hosted recovery require separate acceptance; do not enter real participant data in this exercise.