Xenon isotope potency regressed on nuclear spin and atomic mass, from published per-isotope values

Frozen criteriafrozen 2026-09-01 · hash 4fa792b878fa

Fixed before data collection; any later edit fails the site build. A correction would be a new study superseding this one.

search protocol (re-runnable)

Run only after this freeze is timestamped and published; no query is issued and no number is read before that timestamp. Step 1 — bibliographic search, executed verbatim in PubMed, Europe PMC, Google Scholar, and Semantic Scholar: [xenon isotope anesthetic potency], ["nuclear spin" xenon anesthesia], [xenon-129 xenon-132 anesthesia mice], [xenon isotopes loss of righting reflex], [xenon isotope MAC rodent], [isotope effect noble gas anesthesia], ["Li" "Lu" xenon isotopes Anesthesiology 2018], [xenon isotope consciousness mechanism]. Step 2 — citation sweep: retrieve the full forward-citation list of the Li et al. 2018 record from Google Scholar, Semantic Scholar, and Crossref, and the full backward reference list of that paper; screen every entry against inclusion/exclusion. Step 3 — preprint and grey sweep: [xenon isotope anesthesia] on bioRxiv, arXiv (q-bio, physics.bio-ph), OSF Preprints, and ProQuest theses. Step 4 — registry sweep: [xenon isotope] on ClinicalTrials.gov and the AVMA/preclinical registry PreclinicalTrials.eu, to record any registered-but-unpublished isotope comparison as a known gap. Per-candidate verification rule: a candidate enters the table only after the desk agent (i) opens the primary full text or supplement itself rather than a secondary description, (ii) records DOI or stable URL, (iii) records the exact locus of each number (table number, figure panel, or page and sentence) together with the quoted numeric string, (iv) confirms isotope mass number and stated enrichment percentage from the methods, and (v) assigns a cohort_id. Any candidate meeting inclusion on its abstract but whose numbers cannot be located in the retrieved text is recorded with disposition 'blocked' and the blocking reason, not silently dropped. All queries, dates, hit counts, and screening decisions are logged to an append-only search log published with the results.

candidates known at freeze time — dispositions pre-committed

Method

Every included source yields one row per isotope arm. Extraction is performed twice in separate passes with the second pass blind to the first pass's numeric fields (the agent re-opens the source and re-reads the locus without consulting the prior row); mismatches are resolved by a third read and the discrepancy is logged. Independence: rows sharing a cohort_id (same animals, same experimental series, same institutional batch as stated in methods) are treated as one experiment; overlapping publications are collapsed to the most complete report and the collapse is recorded. Potency values are converted to a common scale as percent-atmosphere for 50% endpoint effect, with the conversion arithmetic shown per row; dispersions are propagated, and where only SEM and n are given, SD is reconstructed. Primary analysis: weighted least squares of potency on (a) spin-present indicator and (b) mass number, weights = inverse variance, run three ways — spin only, mass only, both. Pre-committed diagnostics: variance inflation factor, design-matrix condition number, and the correlation between the spin indicator and mass number across arms. Identification is declared FAILED if VIF > 5 or condition number > 30, in which case the study reports that the dataset cannot distinguish the hypotheses and refuses to report a spin p-value as evidential. Secondary: leave-one-isotope-out refits, a rank-order check (Spearman of potency against mass), and a sensitivity fit substituting nuclear spin quantum number I (0, 1/2, 3/2) for the binary indicator. If fewer than four independent isotope arms survive extraction, no regression is run and the study reports the tabulation plus an explicit statement of non-identifiability. All code, extracted table, and search log are published; the pre-committed expectation recorded in knownCandidates is scored as correct or incorrect in the abstract.

Findings

• The frozen identification-failure rule TRIPS, so this study reports no spin p-value as evidential — but it trips for a reason the freeze did not anticipate. The variance inflation factor is 3.25, comfortably BELOW the pre-set threshold of 5, so spin-versus-mass collinearity (r = -0.832) is real but not disabling: the study happens to place two isotopes on each side of the spin boundary, which is what makes the two hypotheses separable at all. What fails is the scaled design-matrix condition number, 302.4 against a threshold of 30, and its source is intercept-versus-mass, not spin-versus-mass. Atomic mass spans five units around 131.5, under 4 per cent relative variation, so a linear mass term is barely distinguishable from a constant. That is why the mass coefficient is unstable (-0.43 with a standard error of 1.92) while the intercept carries a standard error of 254.→ ledger record ORCH-E037
• Setting the frozen rule aside and describing what the four published means look like: a spin term fits them far better than a linear mass term. On the xenon-alone endpoint, weighted least squares with spin alone gives a weighted R-squared of 0.98 and a coefficient of +30.2 per cent atmosphere; mass alone gives 0.62 and -7.58. The structural reason is a magnitude asymmetry that does not depend on any regression: within the spin-bearing pair, two mass units change ED50 by -6 per cent atmosphere; within the spin-zero pair, two mass units change it by +2; across the spin boundary, ONE mass unit (131 to 132) changes it by 29. A smooth monotone confound in mass — diffusion scaling as one over root mass, solubility — has no way to produce a step of that shape. This is reported as a description of four group means from one experiment, not as an inferential result, because the frozen rule forbids the latter.→ ledger record ORCH-E038
• The published record contains exactly one isotope-resolved xenon potency experiment. The frozen protocol's eight literal queries were run against PubMed and Europe PMC, the full forward citation list of Li et al. 2018 was screened (37 citing works), its full backward reference list was screened (24 references), and preprint servers (arXiv, bioRxiv and other Europe PMC preprints, OSF) and two registries were swept. No replication, no extension with additional isotope arms, and no pre-2018 isotope-resolved potency report exists. Two pre-committed known candidates resolve as follows: 'any pre-2018 xenon isotope potency report' — none found; 'any post-2018 replication or extension with additional isotope arms' — none found. The single named-but-unidentified candidate at freeze, the commentary attached to the Anesthesiology 2018 issue, is identified as Hameroff SR, 'Anesthetic Action and Quantum Consciousness: A Match Made in Olive Oil', Anesthesiology 2018, and is excluded from rows as secondary restatement exactly as pre-committed.→ ledger record ORCH-E039
• Everything above rests on four group means from one laboratory, and the study's own extraction exposed how fragile the record is. The reported plus-or-minus values are of unstated type — the abstract never says whether they are standard deviations or standard errors — and no isotopic enrichment percentage, purity profile, oxygen fraction, equilibration time, blinding statement, or temperature control is available in the accessible record. The xenon-alone ED50 values that carry the headline are not measured but calculated by the authors from co-administration with 0.50 per cent isoflurane; the directly measured endpoint is the isoflurane-combined one, and it gives the same qualitative picture (spin coefficient +6.95, weighted R-squared 0.97, against mass alone at 0.62). A result this widely cited in the quantum-consciousness literature rests on a single unreplicated four-arm experiment whose dispersion type cannot be determined from its own abstract.→ ledger record ORCH-E040
• The decisive experiment is cheap and specific, and this study can now name it. Because the ill-conditioning is intercept-versus-mass rather than spin-versus-mass, adding isotope arms that widen the mass range while holding spin class fixed would resolve it: 124Xe or 126Xe (spin zero, four to eight mass units below the current spin-zero pair) would break the near-collinearity between the mass column and the constant, and 136Xe (spin zero, the heavy extreme) would extend it further. A spin-zero isotope lighter than 129 is the single most informative addition, because it separates 'lighter is more potent' from 'spinless is more potent' directly — the current design has no spin-zero isotope below 132. This is a sharper specification than the freeze's own placeholder of 'a mass-matched isotope pair', and it is what ORCH-R010 should now request.

Limitations

• One source, one cohort, four rows. All four isotope arms come from a single experiment on 80 mice in one laboratory, so the four rows are not independent replications and every row carries the same independence note. No aggregate here is a meta-analysis; it is a re-analysis of one experiment's four group means.
• Numbers were read from the abstract, not the full text. Anesthesiology is paywalled and the paper is not in PubMed Central; the accessible record is the publisher-deposited JATS abstract retrieved through Crossref. The frozen per-candidate verification rule required confirming isotopic enrichment percentage from the methods, and that part of the rule COULD NOT be satisfied. The row is nonetheless dispositioned 'include' rather than 'blocked' because the frozen blocking condition is that the numbers cannot be located, and the potency numbers are located and quoted; the enrichment failure is recorded here instead of being silently absorbed.
• The freeze set a condition-number threshold of 30 without naming a scaling convention, which is a genuine ambiguity in the pre-registration. Three conventions are reported: raw (17,498), unit-column-norm/Belsley (302.4), and centred-and-scaled (4.89). The decision follows the uncentred Belsley convention, which is the STRICTER reading — it is the one that trips the rule and forces the refusal to report a spin p-value. Had the centred convention been chosen the rule would not have tripped and a spin p-value would have been reportable. Naming the convention is the fix for a successor study; choosing the permissive one after seeing the data would have been exactly the tuning the freeze exists to prevent.
• Google Scholar, Semantic Scholar and ProQuest, all named in the frozen search protocol, were not queried as written: Google Scholar and ProQuest expose no programmatic interface this agent can use without scraping, and the Semantic Scholar sweep was folded into the Crossref and Europe PMC citation sweeps. The forward-citation screen therefore rests on Europe PMC's 37-item list rather than the union of four sources, and a citing work indexed nowhere in Europe PMC could have been missed. This is a deviation from the frozen protocol and is recorded as one rather than described as complete coverage.
• The extraction's two passes disagreed on mapping, not on magnitude, and that disagreement is the most transferable thing this study found. The PubMed rendering of this abstract strips the isotope superscripts, leaving 'for Xe, Xe, Xe, and Xe, respectively'. The group order introduced earlier in that same stripped text invites the reading 129 then 131; the true order in the publisher deposit is 131 then 129. Any re-analysis built on the PubMed rendering would silently swap the two spin-bearing isotopes. The swap happens not to change the spin coefficient, since both are spin-bearing, but it would corrupt the spin-quantum-number sensitivity fit and any mass-resolved statement.
• With four observations and three parameters the two-covariate model has one residual degree of freedom, and every leave-one-out refit saturates at zero. The leave-one-out coefficients accordingly range from +18 to +38 for spin and flip sign for mass, which is what the frozen secondary analysis was designed to expose. No stability claim should be read into any coefficient reported here.
• This study says nothing about whether nuclear spin affects consciousness, and nothing about Orch OR. It asks only whether one published dataset can separate a spin effect from a mass effect, and answers that by its own pre-committed rule it cannot, while describing what the numbers look like if one sets that rule aside.