Peer review round 18: LivingGLP1 protocol 0.18.0 (reviewer fable-5.1)
What I verified against the 0.17.0 snapshot: the revision changes protocol.qmd (subtitle; the round and version counts at lines 43 and 241; the Amendments pointer at lines 83 to 86; the reference arm model at lines 1028 to 1058; the competing risk restatement at lines 1068 to 1073; the naming of the relative only rule at lines 1083 to 1087), the applies_to cell of PR02 in priors.csv, amendments.csv (A017, whose count of 9 comments matches the nine round 17 rows of comments.csv; verdicts.csv carries the three round 17 verdicts) and the README status line. The eleven other authoritative tables, concepts.yaml (1.2.0) and R/search_strategy.R are byte identical to the reviewed snapshot; protocol.html carries the 0.18.0 subtitle and every new passage. Numerical checks: the new standard error {g(U) − g(L)}/(2 z) gives 0.1201 for the interval 0.08 to 0.125 at 95 percent (the value of my round 16 and 17 checks) and 0.1431 at 90 percent; g(0) is undefined, so the condition 0 < L < U < 1 excludes the interval that C17-1 used as its counterexample; the Greenwood route by the delta method gives 0.2109 on the g scale for F = 0.10 with standard error 0.02, finite and positive as required.
1. Round 17 comments
| Item | Status | Reason |
|---|---|---|
| F17-0 | Resolved | The response opens with a correction recording the partial adoption of F16-2, and the comments.csv row carries the action corrected. |
| F17-1 | Resolved | Line 1038 adopts "over one period, or summed over reported finer intervals" verbatim, matching the exponential approximation at lines 666 to 667. |
| F17-2 | Resolved | Lines 1050 to 1052 name the reference risk with one usable arm as the posterior median of 1 − exp(−e^{θ_a}), and applies_to of PR02 now names the single arm and scenario uses in my words. |
| F17-3 | Resolved | Line 1042 reads "every participant randomized to that arm". The parenthetical that follows is the residual (item 1 below). |
| F17-4 | Partly resolved | Deferred to the pre-analysis amendment. The stated reason, that adding a rule "would extend the revision beyond the comments of this round", does not hold on its own terms, since F17-4 was a comment of this round. The deferral is nonetheless acceptable: my round 17 review allowed settlement in that amendment, the matter changes no estimand or relative effect model, and it is now traceable through the Amendments pointer. The better reason, which the record should carry, is that the fixed time scenario inputs belong with the deferred reference arm likelihood work of F12 and F16 in the same amendment. |
| F17-5 | Partly resolved | Deferred with F17-4; the labeling and sensitivity prior display are amendment level by my own round 17 judgment, and the pairing with F17-4 is sensible. |
| F17-6 | Resolved | (a) Lines 1085 to 1086 name the relative only rule at its definition; (b) the confidence level is general at lines 1031 to 1033; (c) lines 84 to 86 point to the deferred matters. Item 2 below is a residual of (c). |
2. New items in 0.18.0
No major items. None touches a relative effect model, an estimand or a relative estimate.
Minor
Cross reference after the arm level rule (lines 1042 to 1044). The adopted sentence makes f_a arm level, but its parenthetical still cites "the rule of @sec-outcomes", whose primary form at lines 505 to 507 reads trial wide ("status is known for every randomized participant at t"). The sentence now cites a rule it no longer matches. Replace "(the rule of @sec-outcomes, with the 95 percent rule only in its sensitivity analysis)" with "(the rule of @sec-outcomes restricted to that arm, with its 95 percent rule only in the sensitivity analysis)".
Scope of the Amendments pointer (lines 84 to 86). The pointer says matters "marked
deferred to amendment", but theactionvalues incomments.csvinclude "noted; deferred to amendment" (F14-note) and "revised; part deferred to amendment" (F16-note); a strict reading of "marked" misses two of the six deferred rows (F12-note, F14-note, F15-note, F16-note, F17-4, F17-5). Replace with "those whoseactionfield inprotocol/peer-review/comments.csvcontainsdeferred to amendment".Not new, for the record. The carve out at lines 1072 to 1073 makes visible that a reported Aalen-Johansen value used directly as F_ref has no stated one arm or pooling rule across several covering arms. This is the deferred F12-note matter (reported values off the horizon and their reference model), not a defect of the revision, and it should stay where it sits.
3. Judgment on revisions made for the other reviewers
C17-1: the adopted rule is right and the precedence is closed. A Kaplan-Meier value needs 0 < F < 1 and a finite positive standard error on g(F); an interval of unknown level or with L = 0 falls to the Greenwood variance, an arm with neither falls to the Poisson, then the binomial, and an arm with no qualifying input is listed and excluded (lines 1045 to 1048). The checks above confirm the coverage adjustment and the exclusion of the infinite limit. The response notes the overlap with F17-6b correctly.
C17-2: the restriction of the one arm restatement to the piecewise integration and exponential approximation routes is right, because a reported Aalen-Johansen cumulative incidence need not identify both cause specific hazards, so y_a and V_a may not exist for such an arm; the sentence at lines 1072 to 1073 says so without reopening the deferred horizon matter.
Conclusion
All seven of my round 17 items are addressed: five adopted verbatim or in substance, two deferred to the pre-analysis amendment within the scope I allowed, one with a reason that is weakly stated but an outcome that is acceptable. The two residual items above are one clause editorial corrections with no effect on any estimand, model or registration level decision, so by the standard of rounds 12 to 17 they are amendment level, to be settled in the next revision or in the pre-analysis amendment. The administrative items (team names, affiliations, funding, competing interests, Embase access, PRESS review, verification of the Chinese terms, locking of provisional thresholds) remain on the checklist as pre-registration conditions and do not bear on this verdict. I would approve the methods for registration once those conditions are completed.
VERDICT: ACCEPT