When Every Directional Result Is Positive, the Evidence Needs Another Test

Every directionally reported microbiome countermeasure in the frozen synthesis appears beneficial. This uniform pattern does not establish efficacy. It places a duty on those who carry the evidence forward: examine the registry and reporting chain before the pattern informs program planning or company valuation.

The frozen evidence package covers 114 source rows resolving to 72 analysable studies from 2,434 unique records as of 21 August 2026. Twenty-four records involve a countermeasure. Thirteen state a direction of effect, and all 13 describe a benefit; the remaining 11 state no direction. Against a 0.5 directional null, an exact binomial test of the 13-of-13 pattern gives a p-value of 0.00024 and a 95 percent Clopper-Pearson interval from 0.753 to 1.000, assuming independent records (Gastronaut 2026 synthesis).

The calculation establishes unusual asymmetry within the coded set. It does not show that the interventions work, that their effects are comparable, or that the literature captures all completed research. NASA’s next question is how those results reached publication, followed by whether the record is complete enough to guide another test. Investors inherit a different responsibility: determining whether the apparent success rate survives the same accounting before it influences valuation.

What the synthesis found

Across the 13 directional records, the interventions and endpoints are unrelated. Their signs reveal a reporting pattern, not a common efficacy estimate. Keeping biological targets, exposure conditions, outcome definitions, and evidence tiers distinct preserves the service of each study to the question it addressed.

The tier results define the boundary. T4 simulation records contain nine directional benefits and no reported failures; T3 human-analog records contain two benefits and no failures. Neither tier contains a directional failure, so the tier-transfer odds ratio is undefined. The available cells do not support a favorable or unfavorable tier contrast.

The 11 nondirectional countermeasure records require separate handling. They are not null results; the synthesis did not code a direction. Classifying them as successes would inflate the pattern. Classifying them as failures would impose a result the source record never supplied. Their status deserves the care of full-text and registry review.

A related retrieval exercise identified 55 ClinicalTrials.gov records, reinforcing the need for an audit. These records have not yet entered a registry-to-publication audit. They remain an identification pool, not a denominator for efficacy, completion, or publication rates. Full-text review found a separate reporting issue: 15 of 22 human-tier papers stated direction in the paper when the abstract did not. Abstract-only screening would miss that information (Gastronaut 2026 synthesis).

Why reporting becomes the next test

Several processes could produce 13 directional benefits and no reported failures. The interventions may tend to help. Endpoint selection may follow data review; journals may receive or publish fewer null and unfavorable studies. Abstracts may feature selected outcomes from more complicated full texts. Related cohorts, laboratories, or protocols may generate multiple records, weakening the independence assumption behind the binomial calculation.

The synthesis cannot distinguish among these explanations. Publication bias is plausible, not proven. A registry-to-publication audit can test the evidence-production process and leave a clearer record for the work that follows.

Declared matching rules come first. The audit would link registry records and publications by intervention, population, dates, investigators, sites, sample size, and prespecified outcomes. It would keep every eligible record in view, including completed but unpublished studies, terminated studies, missing results, and publications that changed outcomes or timepoints. Continuity depends on preserving the whole accounting, not only the favorable portion.

The audit also distinguishes studies that never reported results from publications whose relevant endpoint was present but could not be coded directionally.

Cluster control keeps one cohort from becoming several independent confirmations. Before examining direction, the analysis would define its unit of inference; it would then report findings at both record and cluster level. The ledger would show the original outcome, reported outcome, direction, statistic, publication status, and reason for exclusion. That record serves the later test designer, who otherwise could inherit a signal whose denominator and independence remain uncertain.

NASA decision

NASA can place a preregistered registry-to-publication audit ahead of any use of the positivity pattern as countermeasure evidence. The protocol would declare matching criteria, primary outcomes, direction rules, missing-result handling, and cluster definitions. Its program value lies in preserving null and unfavorable results with the same care as favorable ones, while keeping simulation, analog, terrestrial human, and actual-flight evidence separate.

The process would not decide whether a countermeasure is effective. It would establish whether the evidence base is complete and traceable enough to justify the next test. This distinction protects both the work already done and the integrity of the evidence tier still to be earned.

Investor decision

For investors, an unqualified 13-of-13 success narrative retains unresolved denominator and independence questions. The diligence object differs from NASA’s test protocol: a result ledger that exposes denominator changes, related-study clusters, declared outcomes, deviations, nulls, and unresolved records.

Diligence milestones include completion of the audit, resolution of nondirectional records, independent reproduction of coding, and publication of declared decision rules. These evidence-governance milestones give future capital decisions a more faithful inheritance, without assigning a health premium or presuming a positive result.

ORCA relevance and claim boundary

Gastronaut is developing ORCA as a 2.0 m diameter standalone modular cultivation enclosure. ORCA is approximately TRL 3 to 4 and remains at the ground stage, with zero flight and lunar cycles. Its 0.1 g to 1.0 g research range and nominal 0.5 g to 0.65 g operation are design objectives, not demonstrated performance.

For this evidence question, ORCA’s design objective is publication-aware claim control. A proposed operating-history structure could link each test configuration to its biological material, declared endpoint, result, deviation, and disposition. The structure would be Gastronaut company material, not a NASA-validated asset. Its service would be to preserve the test record, without establishing efficacy or transferring a simulation or ground result into a flight claim.

Limitations

All counts and statistics in this report belong to the frozen synthesis of 114 source rows and 72 analysable studies. A later extraction contains 118 source rows and 75 candidate inclusions but remains unreconciled. Its denominator is not combined with the synthesis statistics here. The exact binomial result assumes independent records and a 0.5 directional null; both are analytical choices, not facts about the underlying biology. Heterogeneous interventions and endpoints prevent a pooled efficacy interpretation.

The synthesis records direction, not a common magnitude of effect. A beneficial direction across different endpoints does not show common clinical relevance, durability, or practical importance. Outcome-specific interpretation remains necessary after the reporting audit.

Constructive engagement request

Gastronaut proposes a joint audit with NASA reviewers or designated independent methodologists. First, the parties would declare registry-publication matching rules, outcome hierarchy, cluster treatment, and null-result capture. The resulting ledger would identify the questions that warrant prospective testing while giving investors a bounded diligence record. The purpose is specific: learn whether the positive pattern survives transparent accounting, then pass that answer forward without presuming what it will be.

References

Gastronaut. Gut Microbiome and Immune Function: Frozen Evidence Synthesis. Evidence version frozen 21 Aug. 2026. Research synthesis.