Sequential Review Gate¶
Workflow — instantiates Hypothesis Testing Frame
Re-evaluates evidence at predefined milestones while controlling how interim findings change action.
A Sequential Review Gate lets evidence be examined repeatedly as it accumulates — at pre-scheduled interim milestones — without letting those repeated looks corrupt the verdict. Its defining problem is that every extra peek at growing data is another chance to cross a threshold by luck, so naively "checking as we go" inflates false conclusions. The mechanism solves this by pre-specifying, before any data arrive, when the looks happen and how demanding each interim bar must be, spending a fixed error budget across the schedule so the whole sequence stays honest. What makes this THIS mechanism is that it governs the temporal multiplicity of repeated looks and the action each gate may take — continue, stop, or adapt — rather than analyzing a single readout.
Example¶
A clinical trial testing a new drug against a serious illness cannot ethically wait until the very end to look at the data: if the drug is clearly harming patients, or clearly saving them, the trial should stop early. So an independent Data Safety Monitoring Board reviews the accumulating results at three pre-planned interim points and one final analysis. The catch is statistical: four looks at the same growing dataset give four chances to see a fluke cross the significance line, which would badly inflate the false-positive rate if each look used the ordinary bar.
Before the first patient is enrolled, the trial charter fixes the schedule and the interim boundaries using an alpha-spending rule.[n1] Early looks demand a very extreme result to stop — the drug must look overwhelmingly good or bad — while the final look uses very nearly the ordinary bar, and the boundaries are set so the total false-positive budget across all four looks stays at the target. At the second interim, a strong safety signal crosses the pre-set early-stopping boundary; the board triggers a halt. Because the schedule and boundaries were frozen in advance, the early stop is credible rather than a case of looking until the answer looked good.
How it works¶
- Pre-schedule the looks. The number and timing of interim analyses are fixed before data collection, not added opportunistically.
- Budget the error across looks. A spending rule allocates the total false-positive allowance across the schedule, so repeated peeks do not accumulate risk.
- Make early bars extreme. Interim boundaries demand stronger evidence than the final one, so an early stop reflects a genuinely decisive signal.
- Gate the action. Each milestone can only continue, stop for benefit or harm, or adapt — within pre-agreed rules, not free improvisation.
Tuning parameters¶
- Number and timing of looks — more interims allow earlier stopping but spread the error budget thinner, raising each interim's bar.
- Spending shape — how much error to spend early versus late; conservative-early schedules protect against premature stops but rarely stop early.
- Stopping rules — separate boundaries for efficacy, harm, and futility; each encodes a different reason to halt.
- Adaptation scope — what a gate may change (sample size, allocation) versus what stays fixed; wider adaptation adds flexibility and analytic complexity.
When it helps, and when it misleads¶
Its strength is that it makes monitoring legitimate: it allows the ethical and economic benefits of stopping early — for a clear win, a clear harm, or obvious futility — while preserving the error control that naive repeated testing destroys. Freezing the schedule and boundaries in advance is what separates principled interim monitoring from data-dredging.
Its failure mode is the peeking problem when the discipline is absent: looking repeatedly and acting on the first threshold crossing, with no spending rule, so a chance excursion gets mistaken for a real effect. A subtler misuse is adding unplanned looks or shifting boundaries mid-stream after seeing interim trends. Early stops also tend to overstate effect size, since trials that halt early catch the metric at a random high. The guarding discipline is to pre-register the look schedule and spending function, hold interim boundaries fixed, and read an early-stopped result's magnitude with caution.
How it implements the components¶
multiplicity_context— it explicitly accounts for the repeated-look multiplicity, spending an error budget so many peeks do not inflate false positives.precommitment_record— the interim schedule, boundaries, and stopping rules are frozen before data collection begins.follow_up_trigger— each gate fires a defined action: continue, stop for benefit or harm, or adapt within pre-agreed rules.
It schedules and error-budgets the re-looks but does not set the per-look numeric cut or produce the status label at each gate — the boundary a look is compared against is a Decision Threshold Rule, so it does not implement evidence_threshold or decision_rule itself.
Related¶
- Instantiates: Hypothesis Testing Frame — the workflow realization that spreads evidence evaluation across scheduled milestones under error control.
- Consumes: Decision Threshold Rule supplies the boundary each interim look is compared against.
- Sibling mechanisms: A/B Test Interpretation Protocol · Decision Threshold Rule · Null Hypothesis Significance Test · Legal Burden-of-Proof Analog
Editorial Notes¶
Form Classification¶
Form family: Decision, Gate & Allocation
Rationale: Sequential Review Gate operates by makes continue, stop, accept, or reject dispositions at preregistered interim looks. That concrete deployed or enacted form is Decision, Gate & Allocation under the frozen taxonomy.
Nearest alternative: Assessment, Review & Assurance — Although Assessment, Review & Assurance can support this mechanism, the frozen evidence makes its operative form the act that makes continue, stop, accept, or reject dispositions at preregistered interim looks; the alternative is therefore secondary rather than defining.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Reviewing accumulating evidence only at planned milestones while controlling resulting action extends sequential experimental design.
Related originating lineages:
- Data Science & Analytics — Data science, analytics, and operational monitoring supplies a parallel or contributing lineage for the mechanism's defining operation: re-evaluates evidence at predefined milestones while controlling how interim findings change action.
- Law & Governance — Procedural gates define who may reconsider a commitment and on what record.
- Mathematics — Mathematical modeling, proof, and abstract-structure practice supplies a parallel or contributing lineage for the mechanism's defining operation: re-evaluates evidence at predefined milestones while controlling how interim findings change action.
- Medicine & Healthcare — Trial-monitoring practice provides the clearest safety-critical precedent for interim review.
- Organizational & Management Science — Stage-gate governance turns milestone evidence into bounded continuation, revision, or stop decisions.
Review resolution: The blind reviewers agree that statistics_experimental_design is the primary origin and differ only on alternate origin disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain cross_disciplinary_synthesis because the combined record shows material contributions from several lineages. The broader reach of multi_domain records portability separately from historical provenance, and encyclopedia_synthesis=true preserves the affirmative synthesis judgment where either reviewer identified one.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] Alpha-spending functions (Lan–DeMets, generalizing O'Brien–Fleming boundaries) allocate a study's total false-positive allowance across pre-planned interim analyses, so that evidence can be examined repeatedly as it accumulates while the overall error rate stays at the intended level. Early interim boundaries are deliberately stringent, requiring an extreme result to stop. ↩