Methods need minor revision for three rule inconsistencies. Administrative pre-registration conditions do not determine this verdict.
| Round 6 comment | Status | Assessment |
|---|---|---|
| C6-1 | Partly resolved | Constant cause specific hazards, exponential approximation, treatment multipliers, and integration for time varying hazards are now explicit. Final fallback sentence contradicts the integration option and needs narrower scope. |
| C6-2 | Resolved | Proportional subdistribution hazards are explicit. Reported assessments and curves are examined; material nonproportionality triggers separate horizon-specific reporting and prevents constant-HR conversion across horizons. |
| C6-3 | Resolved | Formula defines (Q) correctly, removes incidence-rate and hazard interpretations, and specifies certainty thresholds on (Q). Newly introduced offset wording needs correction below. |
| C6-4 | Resolved | Known status includes documented competing death without preceding target event. Earlier administrative censoring leaves status unknown; competing deaths remain in randomized denominators. |
| C6-5 | Resolved | General bias-direction claim removed. Numerical examples explicitly assume independent missingness at baseline. Recalculation confirms approximately 1.935 and 1.265. |
No C6 proposal was declined. Retaining a separately labeled cause specific network before descriptive (Q), as explained in responses F6-1 and G6-2, is defensible provided conclusions retain that hazard interpretation.
No major finding. Remaining and newly introduced minor findings:
C7-1. Absolute-effect fallback overrides valid conversion options. protocol/protocol.qmd:561, “Competing risks,” especially lines 564 to 566.
Requiring relative effects only whenever constant hazards cannot be supported contradicts the immediately preceding integration option. Read literally, it also imposes a constant-hazard requirement on subdistribution conversion. General cause specific integration allows time varying baseline hazards; the exponential formula is its special case. Haque and Cortese, equations 6 and 7.
Replace the final fallback sentence with:
“For cause specific hazard ratios, report relative effects only when neither integration using the required time varying baseline hazards nor an exponential approximation using both baseline rates under a supported constant hazard assumption is possible. For subdistribution hazard ratios, conversion requires reference cumulative incidence at the analyzed horizon and supported proportional subdistribution hazards; constant cause specific hazards are not required. When the applicable conversion cannot be supported, follow the relative only rule of @sec-absolute.”
C7-2. New count-routing restriction conflicts with all-cause mortality classification. protocol/protocol.qmd:465, “Event outcomes and harms,” compared with protocol/protocol.qmd:527, “Competing risks,” and protocol/outcomes.csv:20, row O20.
Counts for every time to event outcome are directed exclusively to subdistribution networks. O20 is a time to event outcome explicitly assigned hazard type “not applicable.” These instructions give incompatible destinations for complete-status mortality counts. The blanket statement that cause specific networks are primary for time to event outcomes also needs this exception.
Replace the count-routing sentence and qualify the primary-network rule with:
“For time to event outcomes with competing events, complete-status counts enter only the subdistribution network and are never pooled with cause specific estimates. For outcomes without competing events, including O20, reported hazard ratios and eligible complete-status counts use hazard type ‘not applicable’ under proportional hazards, with one input per trial. This network is primary for those outcomes.”
C7-3. Recorded-harms offset is written as (t), although the defined quantity requires (\log(t)). protocol/protocol.qmd:500, “Recorded harms analysis.”
Revised sentence specifies nominal observation period (t) “as offset.” This conflicts with the (Q) formula and the log-follow-up specification elsewhere. For recorded proportions 0.10 and 0.19 over 1 and 2 years, respectively, (Q=1); using (t) as offset gives an exponentiated contrast of approximately 0.736.
Replace the first sentence with:
“All trials' counts enter the same complementary log log model with offset (\log(t)), where (t) is the nominal observation period in years.”
All requested files inspected. Read-only checks found no identifier mismatches, overlapping applicable windows, missing reference-SD or Turner mappings, nonrandomized-harms coverage mismatches, or registry candidate count discrepancies. Search renderer inspected statically; live search recall and R execution were not assessed. No files edited, created, or deleted by this review.
VERDICT: MINOR REVISION