Peer review round 9: LivingGLP1 protocol 0.9.0 (reviewer grok-4.7)
1. Round 8 comments
G8-1. Resolved. Known status counts a competing death as event free. Complete status counts are subdistribution inputs only when the outcome has a competing event, and they are not applicable inputs for O20. The 95 percent known status sensitivity is an analysis of the subdistribution network, or of the not applicable network for O20, and never of a cause specific network.
G8-2. Resolved. For a time to event outcome with a competing event, the cause specific network is primary. For O20 the not applicable network is the only network and is primary. For binary harms the subdistribution network is primary. Conclusions use the O20 network when it meets the connectivity rule, and descriptive reporting otherwise.
G8-3. Resolved. Subdistribution F_ref is now one ordered quantity: the reported Aalen-Johansen cumulative incidence, otherwise piecewise integration of the reference hazards, otherwise the exponential approximation under supported constant hazards, otherwise relative effects only. Constant cause specific hazards are required only when F_ref comes from that exponential approximation. Cause specific conversions are separate: the reference risk shown is the reference node cumulative incidence computed by the same route as the treated risk. The piecewise step is a sound expansion of the order requested in round 8. A clash that remains in Absolute effects is a new issue below.
G8-4. Resolved. A contrast of one minus Kaplan-Meier estimates, with competing deaths censored, enters the cause specific network as a normal likelihood on log(-log S) at a common time. The contrast is the difference in log cumulative cause specific hazards. Its standard error is the delta method standard error from the reported confidence intervals or the Greenwood variance, and it is used only when one minus S is strictly between 0 and 1. The contribution rule lists that contrast with its delta method variance and no continuity correction. It is never an absolute or reference risk.
G8-5. Resolved. The integral now uses event free survival, in which both the target event and competing death are events. Survival free of both is the exponential of the negative integral of the sum of the two cause specific hazards, and the target cumulative incidence is the integral of that survival times the target hazard. On the treated arm the target hazard is multiplied by its cause specific hazard ratio, and the competing hazard is multiplied by the competing event hazard ratio when one is reported. Otherwise the competing hazard is unchanged, and that no effect assumption is stated. With constant hazards of 0.2 and 0.1 per year, the cumulative incidence at 5 years is 0.518 under event free survival and 0.787 if survival treats only death as the event.
G8-6. Resolved. The response declines a flagged median fallback, and the reason is sound. The number randomized times a median follow-up is not time at risk, and a flag does not correct the rate. Person time is the total time until the first of the target event, the competing event, censoring, or the analysis cutoff, over the same interval as the events. The number randomized times mean follow-up is allowed only when that mean is exactly this time at risk. Median follow-up, treatment exposure, and general study follow-up are not substitutes. When valid person time is unavailable, rates are not estimated by division; another supported route or the relative only rule applies.
2. New issues
No major errors. The survival function in the integral is the correct event free survival. The new defects are inconsistencies between sections.
Minor 1. Absolute effects applies the subdistribution baseline order to cause specific hazard ratios. File: protocol/protocol.qmd, Absolute effects, Hazard based outcomes. After stating that a single event rate is never used for an outcome with competing events, the paragraph gives one sequence for both cause specific and subdistribution reference risks: reported Aalen-Johansen cumulative incidence, then piecewise integration, then the exponential approximation, "in the order given there." That sequence is the subdistribution order in Competing risks. The cause specific bullet requires the reference risk shown to be computed by the same route as the treated risk: piecewise integration, otherwise the exponential approximation. A reported cumulative incidence is outside that route. Pairing a reported Aalen-Johansen baseline with an integrated treated risk puts the two arms on different baselines.
Replace the competing events sentence in Absolute effects with: "For an outcome with competing events a single event rate is never used. For a cause specific hazard ratio the reference risk shown is computed by the same route as the treated risk: piecewise integration of the reference hazards, otherwise the exponential competing risks approximation under supported constant hazards, otherwise the relative only rule. For a subdistribution hazard ratio, F_ref is the reported Aalen-Johansen cumulative incidence, otherwise that piecewise integration, otherwise the exponential approximation under supported constant hazards, otherwise the relative only rule."
Minor 2. A digitized cumulative incidence curve is treated as if it supplied both cause specific hazards. File: protocol/protocol.qmd, Competing risks, reference hazards, and Data items, outcome level. Reference hazards are piecewise constant "on the intervals of digitized reference arm curves... when such curves exist," and extraction includes "digitized reference arm cumulative incidence curves where published." The slope of a target cumulative incidence is event free survival times the target hazard, so that curve does not identify both cause specific hazards. "When such curves exist" makes this the first absolute risk route whenever a curve was published. Pooling "per interval on the log scale with the reference risk model" also leaves out the variance for each arm and interval.
Replace the reference hazard sentence in Competing risks with: "Piecewise constant cause specific hazards are used only when, on each interval, the covering reference arms provide events and person time at risk for both the target event and the competing event, or provide the target cumulative incidence together with the competing event cumulative incidence or with event free survival. A target cumulative incidence curve alone is not sufficient. Each interval's log hazards are pooled with a random effects model on the log scale, using the median of the predictive distribution, the same summary used for reference risks; each arm interval needs a variance. If those inputs are absent, use constant rates λ1 and λ2 only when constant hazards are supported; otherwise apply the relative only rule."
In Data items, extract digitized reference arm curves for the target cumulative incidence and for the competing event, or for event free survival, where published. Record a target cumulative incidence curve alone only as a cumulative incidence, not as the source of both hazards.
Minor 3. An O20 Kaplan-Meier contrast has no stated network. File: protocol/protocol.qmd, Event outcomes, and the cause specific parenthesis in Competing risks. When a trial has more than one eligible input, the new rule takes the reported hazard ratio, then the cumulative incidence or Kaplan-Meier contrast, then the arm counts. The O20 sentence names only reported hazard ratios and complete status counts as not applicable inputs. The Kaplan-Meier contrast is defined inside the cause specific parenthesis. When O20 has no reported hazard ratio, that contrast can be read as a cause specific input, which conflicts with O20 having only its not applicable network.
In the O20 sentence, name reported hazard ratios, Kaplan-Meier contrasts, and complete status counts as not applicable inputs to that one network. State that a Kaplan-Meier contrast for an outcome with a competing event remains cause specific.
3. Checks and registration
The authoritative tables match the reviewed version 0.8.0 snapshot except protocol/amendments.csv row A008, which logs this revision. search/strategies/concepts.yaml version 1.2.0 is identical to that snapshot. R/search_strategy.R builds the API queries from the class and agent terms. The Chinese terms remain a separate key for manual CNKI and Wanfang searches. The renderer was read and was not executed. The iGlarLixi switching row remains a template and stays out of the networks. One input per trial is compatible with sending each hazard type to its own network, because hazard type is part of the network key.
Team names, affiliations, postal address, 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 pre-registration conditions on the administrative checklist. This verdict is based on the methods. The three inconsistencies above should be aligned with the cause specific and subdistribution rules already written in Competing risks. With those sentences aligned, the methods are acceptable for registration once the administrative conditions are completed.
VERDICT: MINOR REVISION