Peer review round 18: LivingGLP1 protocol 0.18.0 (reviewer grok-4.7)

1. Round 17 comments

I raised no comments on version 0.17.0. The revisions made for the other reviewers leave the earlier absolute effect rules in place. Each adopted change is judged below. Two of them leave a small inconsistency, listed in section 2.

C17-1. Resolved. A Kaplan-Meier cumulative incidence is used only when (0 < F < 1) and the standard error of (g(F) = \log{-\log(1-F)}) is finite and positive. A two sided interval ((L, U)) of known level contributes ({g(U)-g(L)}/(2 z_{1-\alpha/2})) only when (0 < L < U < 1). Otherwise the standard error comes from the Greenwood variance by the delta method. If neither source qualifies, the Poisson likelihood is next and the binomial likelihood follows it. Limits that include 0 or 1 no longer produce an infinite standard error. For limits 0.08 and 0.125, the standard error is 0.120 at 95 percent and 0.143 at 90 percent. The normal approximation remains centered at (g(F)), which equals (\theta_a = \log \Lambda_a(t)).

C17-2. Resolved. The one arm restatement now covers only the piecewise integration and exponential approximation routes. With one covering arm, (\tau_k) is not estimated and each draw of the reference log hazards is from (\mathrm{N}(y_a, V_a)). The next sentence states that this rule does not apply to a reported Aalen-Johansen cumulative incidence used directly as (F_{\mathrm{ref}}). That exclusion is required, because a direct Aalen-Johansen value does not supply (y_a) and (V_a). The source order for a subdistribution (F_{\mathrm{ref}}) is unchanged: reported Aalen-Johansen cumulative incidence, otherwise piecewise integration, otherwise the exponential approximation, otherwise the relative only rule. Scenario values for a reported Aalen-Johansen source remain the covering arms' own cumulative incidences in the opening paragraph of Absolute effects.

F17-0. Resolved. The round 17 response records a partial adoption of the round 16 scenario proposal. The text matches that record: every usable arm, including a Kaplan-Meier arm, takes its scenario risk from its own likelihood under PR02 alone.

F17-1. Resolved. Person time may be one period or the sum of reported finer intervals. Under the stated constant hazard, the summed event count and the summed person time are the Poisson input (e^{\theta_a} E_a / t), the same construction the exponential approximation already uses. An arm that reports both count inputs still enters only the Poisson likelihood.

F17-2. Resolved. With one usable arm the reference risk is the posterior median of (1 - \exp(-e^{\theta_a})), heterogeneity is not estimated, and PR02 applies. The transform is strictly increasing, so that median equals the risk at the posterior median of (\theta_a). The applies_to field of PR02 now names a single usable arm's (\theta_a) and each usable arm's (\theta_a) for the scenario risks. The row still parses as seven fields. The rationale still places that prior on (\log \Lambda(t)) for a hazard based outcome without a competing event, which is the scale of (\theta_a) in this branch. Fixed time outcomes do not use (\theta_a).

F17-3. Partly resolved. The operative clause now requires status known for every participant randomized to that arm, and (f_a) must reach the analyzed horizon. The parenthetical still calls this "the rule of @sec-outcomes" and keeps that section's 95 percent rule for the sensitivity analysis. In Event outcomes and harms, the hazard ratio analysis requires status known for every randomized participant, and its 95 percent rule requires that figure in every arm. Those trial wide conditions are a different rule. The parenthetical can put a complete reference arm out of the baseline because of missing status in another arm, which is the defect this comment removed. See minor item 1.

F17-4. Resolved by deferral. The response leaves scenario inputs for fixed time outcomes to the pre-analysis amendment. The reason is sound. The gap predates version 0.17.0, the reviewer allowed the deferral, and a new rule in this round would have gone past the comments. The opening paragraph still says only "trial reference risks" for that branch. The deferral is recorded as F17-4.

F17-5. Resolved by deferral. Zero count scenario values stay prior dependent under the accepted likelihood. Labeling them, and showing them under the sensitivity prior of PR02, is deferred with F17-4. That is a display rule, and the likelihood itself is unchanged. The reason is sound.

F17-6. Partly resolved. The defining passage now names the relative only rule: when a valid absolute conversion is unavailable, absolute effects and absolute importance categories are marked unavailable, magnitude dependent certainty uses the relative range 1/1.25 to 1.25, and rankings that need absolute effects are omitted. The confidence level is the general quantile in C17-1. The Amendments paragraph points at comments marked deferred to amendment. An exact match on that action misses rows whose action only contains the phrase, including F16-note, which is the competing risk count likelihood deferral this pointer was meant to keep visible. See minor item 2.

2. New issues

No major errors. The relative effect models, estimands, and competing risk likelihoods are unchanged.

Minor

1. Absolute effects, protocol.qmd, the binomial (f_a) clause (lines 1041 to 1044). The arm level requirement and the citation of the trial wide hazard ratio rule cannot both be applied. Replace "status is known for every participant randomized to that arm (the rule of @sec-outcomes, with the 95 percent rule only in its sensitivity analysis)" with: "status is known for every participant randomized to that arm. Known status uses the definition in @sec-outcomes. This primary reference arm likelihood does not use the trial wide complete status requirement of the hazard ratio analysis. A sensitivity analysis instead takes (f_a) as the longest follow-up at which status is known for at least 95 percent of participants randomized to that arm; unknown participants stay in the denominator and count as event free."

2. Amendments paragraph, protocol.qmd (lines 84 to 86). The new index uses the exact action deferred to amendment. That action is the full value for F12-note, F15-note, F17-4, and F17-5. It is not the full value for F16-note (revised; part deferred to amendment) or F14-note (noted; deferred to amendment). F16-note is where count likelihoods for the competing risk routes are deferred. Replace "those marked deferred to amendment" with: "those whose action contains deferred to amendment, including noted; deferred to amendment and revised; part deferred to amendment."

3. Checks and registration

Against the version 0.17.0 snapshot, the only authoritative edits are in protocol.qmd (version lines, the Kaplan-Meier standard error, the Poisson sum, arm level (f_a), the one arm reference risk, the competing risk one arm restatement, the relative only name, and the amendments pointer), in the applies_to field of PR02, and in amendments.csv (A017, dated 2026-10-13, version 0.18.0, nine comments, before results). The nine comments are the nine round 17 rows in comments.csv. outcomes.csv still has 77 outcomes. The other authoritative tables are unchanged. concepts.yaml remains version 1.2.0. R/search_strategy.R still builds the PubMed, Europe PMC (excluding patents), OpenAlex, and ClinicalTrials.gov queries from the class terms, the agent terms, the mechanism block, and the PubMed vocabulary. The Chinese terms remain only in concepts.yaml. The renderer was read and was not executed. The Quarto file was read and was not rendered. The iGlarLixi switching row remains a template (is_template yes).

Team names, affiliations, funding, competing interests, confirmation of Embase access, PRESS review by a human information specialist, verification of the Chinese terms by a Chinese-reading team member, and confirmation and locking of provisional thresholds by clinical advisors and patient partners remain administrative pre-registration conditions. The checklist also lists an OpenAlex API key and a logged harvest of all four API sources before the baseline search. Those administrative items remain pre-registration conditions. This verdict follows the two minor text fixes above. With those sentences corrected, I would approve the methods for registration once the administrative conditions are completed.

VERDICT: MINOR REVISION