Recording distal cold and rewarming on the same timeline

Purpose, status, and non-use

Research on recurrent cold hands or feet can become uninterpretable before analysis begins. A single episode may generate a momentary cold-sensation rating, a skin-surface temperature at one named site, a flow or perfusion result from one named tissue compartment, and several possible summaries of rewarming. These observations sometimes align and sometimes diverge, while baseline, onset, exposure, first measurement, recovery, and clinical-course clocks are not interchangeable. The unresolved reader problem is therefore not how to choose one measure as the truth, but how to observe their relation without collapsing constructs, sites, compartments, or clocks. ^[1]

This methods note proposes a prospective, repeated, within-participant observation design for natural episodes. Measurements would be time-aligned but kept distinct; hand and foot procedures would remain separate. The next sections specify the population and episode hierarchy, acquisition, clocks, quality and missingness, analysis, validation, safety stops, and reuse package. An optional laboratory provocation module could improve control of exposure and clock origins, but it would retain a separate protocol label, analysis stratum, and claim set and would not substitute for a naturally occurring episode. This design responds to a bounded evidence gap identified through the 2026-08-16~17 PubMed retrieval and passage-verification route; it is not an observed or validated design effect. ^[1]

The proposal is a validation agenda, not a validated instrument. Protocol detail cannot establish feasibility, technical repeatability, bilingual wording validity, construct validity, replication, safety, or clinical usefulness; S7 states the staged tests and failure conditions. Until those phases succeed, the method must not be used to diagnose or rule out a condition, classify an individual, define normal rewarming, guide treatment or triage, reassure a participant, or serve as a self-test. ^[1]

Population and episode unit

The target population would be adults who report recurrent cold sensation localized to one or both hands, one or both feet, or both site families. Recruitment would be based on that reported episode pattern rather than an assumed diagnosis or a temperature, perfusion, or recovery threshold created by the study. Before recruitment, the protocol would fix the age range, setting, minimum episode frequency and recall window, eligible anatomy and laterality, relevant medications and exposures, named disease, injury, diabetes-foot, vascular, neurological, systemic, and skin contexts, and the practical ability to complete synchronized observations. The sampling frame and denominator would be reported, but a clinic, volunteer, or convenience sample would not be used to estimate prevalence. ^{[1, 2]}

The primary episode unit would be one prespecified distal-cold episode at one site family in one participant. Laterality and the episode and clock rules would be fixed before analysis. Timestamped samples would be nested within episodes, episodes within site family, and site families within participants. The protocol would prospectively set both a minimum target and a maximum accepted number of episodes per participant, retain every participant's episode count and observation opportunity, and carry within-participant dependence into analysis. Repeated episodes or samples would never be counted as independent people, and recruitment success would not be reported as a frequency of the underlying symptom in a wider population. ^[1]

Enrollment and reporting would retain hand-only, foot-only, and combined-site strata rather than pooling them by default. A participant reporting both site families could contribute to each only under its own prespecified episode definition and valid site-specific procedure; distal location alone would not make the observations exchangeable. Natural episodes would form the primary observation stream. Any optional provoked episode would have a distinct protocol label, analysis stratum, exposure and clock definition, and claim set. Better experimental control of a provoked episode would not validate transfer to a natural episode, and a natural episode's ecological relevance would not erase missed onset, absent true baseline, or severity-dependent capture. ^{[1, 2]}

Separate hand and foot procedures

Before observation begins, the acquisition manual would define a separate procedure for each eligible site family. For every hand or foot observation it would fix laterality, the named digit, the anatomical surface, a reproducible landmark or coordinate, region-of-interest dimensions, participant posture, limb support, operator handling, and the relation among the sensation prompt, temperature region, and any separately named flow or perfusion region. A moved, covered, injured, unavailable, or incorrectly positioned site would receive a prespecified deviation or missingness code; the operator would not replace it with the opposite side, a nearby region, another digit, or a whole-limb reading after seeing the data. These rules make the observation reproducible but do not identify an optimal site or device. ^{[1, 2]}

The hand procedure would prespecify the finger and surface—for example, a defined volar digit region only if that is the protocol's chosen target—together with side, wrist and forearm posture, hand support, and whether rings, gloves, drying, tape, probe contact, pressure, or repositioning can alter the region or its microclimate. The record would distinguish a local finger temperature region from a dorsal-hand, palm, or other perfusion region and would preserve the actual spatial offset between instruments. Dorsal-hand perfusion, palm temperature, finger-pad temperature, tactile report, and whole-hand sensation would remain separately named observations; none would silently stand in for another. The example sites are specifications to be justified and piloted, not evidence that one hand surface is best. ^[1]

The foot procedure would independently prespecify toe, plantar or dorsal surface, side, posture, limb support, the interval after weight bearing, and the timing and handling of footwear and sock removal, drying, tape, contact, pressure, or repositioning. It would retain a defined toe temperature region, a named local flow or perfusion compartment, arterial pressure, and whole-foot sensation as non-equivalent observations; great-toe conductance would not substitute for coldest-toe temperature, another toe, arterial pressure, or the participant's report. Hand and foot results would be reported separately. Cross-site agreement would be estimated only when the same participant contributes observations under the same prospective episode definition and construct schedule with valid site-specific procedures; even then, agreement would be an estimate rather than permission to pool or transfer a hand result to a foot. Controlled evidence of different hand and foot baselines, cooling, rewarming, and only limited cross-site association supports this caution but does not validate these procedures for natural episodes. ^{[1, 2]}

One timeline, four non-equivalent observations

Each episode would use one timestamp system while preserving four non-equivalent observations: a momentary report of felt cold, a site-matched skin-surface temperature, one prospectively chosen and explicitly named flow or perfusion compartment, and construct-specific rewarming trajectories. The flow or perfusion record would identify its instrument, tissue or depth target, anatomical region, units, acquisition duration, quality rule, and spatial relation to the temperature site; it would not be called generic circulation. At aligned collection times the protocol would retain each observation's actual timestamp and acquisition window. Alignment would permit comparison but would not imply that a participant's sensation, a surface-temperature value, and a device response are simultaneous, interchangeable, or governed by one latent clock. This method-site-compartment-task-clock bundle is the observation contract. ^[1]

A signed, version-identified decision record would be completed before the earlier of first enrolment or first acquisition, separately for hand and foot site families and for natural and provoked streams. It would fix episode eligibility, participant-noticed onset recording, first-feasible-sample delay, one primary construct-specific recovery origin, the recovery window, sampling schedule, observation limit, and one primary endpoint; a study could not begin with any required field blank or left as an unordered menu. A true pre-episode baseline may be unavailable during a natural episode; the record would retain noticed onset, the first feasible sample, and their delay rather than relabel the first value as baseline. Participant-reported, temperature-derived, and compartment-specific nadirs would remain distinct, and an unobserved nadir would remain missing. Alternative origins or endpoints would enter only a finite ordered sensitivity set locked in the same version. ^[1]

The same locked record would name the primary lag treatment, timestamp source and resolution, acquisition-window and pairing algorithm, numeric synchronization tolerance for each instrument pairing, tie rule, minimum valid-pair count, required temporal coverage, construct-specific quality criteria, and the consequence of failure at episode and participant levels. It would also fix the observation limit and ending-state rules before acquisition. Observed recovery, withdrawal, safety stop, device failure, loss of synchronization, and reaching the limit without recovery would remain distinct; a threshold not crossed by the limit would be right-censored rather than assigned the limit as a recovery time. A finite ordered sensitivity set could vary only the prospectively listed origin or endpoint, lag, tolerance, valid-pair threshold, first-sample delay, and censoring assumptions. Later additions would be versioned, timestamped exploratory deviations and could not replace the primary result. Verified protocols justify explicit clocks and sensitivity analysis, but establish no universal natural-episode interval, threshold, or normal recovery time. ^{[3, 4]}

Devices, sites, repeatability, and missingness

Every observation would carry enough metadata to reconstruct what the instrument sampled and when. The record would include manufacturer, model, device or serial ID, software and algorithm version; calibration method, date, traceability, uncertainty, drift check, and an in-session reference where feasible; and the anatomical landmark or coordinates, region size, spatial resolution, and relation to the paired site. Construct-specific fields would retain contact, tape, pressure or occlusion, emissivity, angle, distance, focus, acquisition duration, and response time as applicable. Ambient temperature, humidity and airflow, posture and acclimation, handling, synchronization method, timestamp resolution, first-sample delay, repositioning, drying, motion, operator identity, transformations, exclusions, and deviations would remain linked to the raw or minimally processed observation. Here spatial support means the device-specific pixel, field of view, probe contact, region, or aggregation meaning, not an unsupported numeric area. A calibrated instrument at one location would not make an adjacent measurement or a different tissue compartment its equivalent. ^{[1, 5]}

Technical comparison would be planned rather than inferred from disagreement. Where feasible, temperature devices would be compared at the same location, or at prospectively mapped locations with the mapping error retained, against an appropriate reference. Duplicate or repeated readings would be acquired in prespecified stable and changing phases when the instrument permits, with order and delay preserved; operator and device allocation would also be prespecified. Repeatability or agreement would be estimated with uncertainty separately by device, site, construct, conditioning, and hand or foot procedure. Between-episode stability within a participant would remain distinct from within-session repeatability, and neither would be called interchangeability, construct validity, or clinical validity. In one verified rewarming experiment, adjacent infrared and taped-contact readings disagreed in a time-varying way, but no independent same-location in-situ truth standard could assign the error to either system. Verified toe-pressure and plantar-muscle-perfusion examples likewise bound their repeatability to specific preconditioning, sites, devices, acquisition and processing rules; they do not validate this proposed method. ^{[5, 6, 7]}

Quality failure and absence would remain observable data states. Before the lock moment, the plan would fix reason codes, analysis-set membership, the primary missing-data model and assumptions, whether any imputation is permitted, and the sensitivity alternatives for no true baseline, late or missed synchronized samples, device or calibration failure, artifact, site relocation, timing error, contact microclimate, compartment mismatch, withdrawal, safety stop, episode end before setup, right-censored recovery, and not applicable. Missingness would be shown by construct, clock, site, episode, participant, severity, tolerance, and duration. A failed value would not be silently replaced by a nearby site, opposite side, device, or construct. The observation limit would not become an observed recovery time, and an unobserved threshold crossing would not be imputed outside the locked justified model. Complete-case analysis, if retained, would be a named sensitivity analysis rather than an unplanned replacement for the primary estimand. ^{[1, 7, 8]}

Participant-level analysis and endpoint sensitivity

The primary unit would remain the participant-level synchronized trajectory within one site family, with samples nested within episodes and episodes within participants. Before the earlier of first enrolment or first acquisition, one signed protocol and analysis-plan version would fix the primary relation as the within-participant association between momentary thermal-sensation score and site-matched skin-surface temperature at valid timestamped pairs. It would name the eligible population and stream, scale direction, transformation, model or association measure, effect scale, site-family-specific origin and window, primary endpoint, lag, synchronization and pairing rule, valid-pair threshold, episode-combination and unequal-episode rule, dependence handling, missing-data and censoring models, cross-participant summary, and uncertainty measure. Participant-specific estimates and heterogeneity would accompany any hierarchical summary; a population-average result would not be presented as every person's direction or strength. A named flow or perfusion relation would remain secondary and retain its instrument, compartment, site, units, and clock. ^{[1, 3, 9]}

Primary analyses would remain separate for hand and foot and for natural and provoked streams. The locked plan would state how eligible repeated episodes are combined within a participant, how unequal episode counts and within-participant dependence are handled, and which one site-family and stream-specific estimand is primary. Individual curves would show actual timestamps, missing samples, first-sample delay, observation limits, and quality flags; participant-level directions, magnitudes, and uncertainty would remain visible beside any cross-participant summary. Secondary construct relations and the multiplicity method would be enumerated in the locked version. Sensory context, medication, touch, disease, device, episode-context modifiers, and any later-created subgroup would remain exploratory unless separately powered and prospectively locked before the same start condition. ^{[1, 3]}

The locked sensitivity set would be finite and ordered, naming only justified alternatives for recovery origin or endpoint, lag, synchronization tolerance, valid-pair threshold, first-sample delay, complete-case restriction, missing-data assumption, and censoring assumption. An item could be marked prospectively inapplicable; the list would not imply that every endpoint is mandatory or estimable. The primary result could not be replaced by whichever site, clock, endpoint, device, lag, model, or subgroup looked strongest. Any later addition or change would preserve the original version, record its rationale and the timing of acquisition and data access, and be labelled exploratory. Opposing individual perfusion shifts in a verified hand study warn against aggregation and circular grouping but do not establish phenotypes; participants would not be typed by the sign of a response and then validated with that same response. Robustness, heterogeneity, replication, and validity would remain separate questions. ^{[1, 3, 4, 9]}

Validation, safety stops, and failure conditions

Protocol detail would not count as validation. Evaluation would proceed in ordered, separately reported phases: a feasibility and synchronization pilot; technical repeatability and same-location or prospectively mapped device-and-site agreement; content and cognitive evaluation of Korean and English sensory wording; construct-validity tests against predictions fixed in advance for both convergence and divergence; external replication in a separately recruited sample and setting; and clinical-utility evaluation only if the earlier measurement phases succeed. Each phase would retain separate hand and foot results and natural and provoked streams. Advancement would require prespecified completeness and synchronization targets without severity-dependent attrition, technical error small enough for the intended within-episode change, site-family-specific quality targets, replication of predicted agreements and divergences with participant-level uncertainty, and conclusions that are not reversed by reasonable prespecified recovery-origin or censoring analyses. Passing one phase would not validate a later phase or establish diagnosis, prognosis, treatment selection, a normal range, or clinical utility. ^{[1, 10]}

Operational safety stops would be governed before enrolment by the approved study protocol, ethics review, trained staff roles, site procedures, adverse-event plan, and jurisdiction-specific current authoritative sources. The examples below are illustrative and bounded by the sources that were accessible and current enough to verify; they are not an exhaustive red-flag or triage list, and an unlisted context must not be treated as safe by omission. Blocked NICE and unresolved-currentness sources cannot establish that other safety contexts are absent. This methods note supplies no universal stop, urgency, referral, reassurance, diagnosis, individualized advice, or substitute for qualified assessment; exact actions remain the responsibility of the applicable approved protocol, trained role, and separately governed pathway. Distal cold wording alone would not trigger a clinical classification. Instead, added context outside the observation question—such as acute or prolonged cold exposure with possible tissue injury, evolving skin or tissue change, trauma or neurovascular change, rapid multisystem illness, diabetes-foot tissue risk, or a lower-limb vascular course—would stop or defer research measurement only under the applicable prespecified rule and transfer responsibility to that separately governed pathway. The record would preserve which contextual rule fired, the construct and samples not collected, the time and ending reason, and any resulting censoring, without turning the protocol into individualized advice or reassurance. Exact actions cannot be copied across populations, anatomy, courses, jurisdictions, or source versions; blocked or currentness-unresolved guidance would not supply them. ^[1]

The method would fail its present purpose if sensory wording is not understood consistently; device or site error is as large as the change of interest; first-sample delay repeatedly misses the relevant episode phase; missingness or stopping depends materially on episode severity; named perfusion and temperature observations cannot be synchronized; conclusions require an unplanned endpoint, site, device, or pooled hand-foot analysis; external replication fails; or the procedure cannot operate safely without becoming a clinical pathway. Failure would be reported, not repaired by relabelling completeness, repeatability, agreement, robustness, or protocol adherence as validity. Severity-dependent capture failure would narrow the primary question or redesign acquisition; incompatible hand and foot devices, clocks, sites, or safety controls would split the protocol; failed same-location comparison would revise acquisition and quality rules; failed bilingual cognitive evaluation would retain separately validated items rather than forced equivalence; and a shift toward diagnostic accuracy or clinical utility would require a separately governed protocol with a justified reference standard and analysis plan. Version history would preserve every redesign. ^{[1, 10]}

Reuse package and unresolved questions

A reusable release would be a versioned methods package, not a free-standing checklist or an operational-data release. It would contain only blank templates, schemas, controlled vocabularies, synthetic examples, and protocol-level version or change records: the protocol and schedule; Korean and English item wording with translation and cognitive-testing history; separate hand and foot site maps and acquisition manuals; blank device-configuration and calibration templates; a timestamp-field data dictionary and reason-code schema; the prospective analysis plan and executable code; and blank deviation, adverse-event, and safety-stop forms with protocol-level change history. A protocol-level record may identify the changed component, version, effective date, responsible role, dependency, rationale, and compatibility status, but it must not contain a participant event, participant-linked timestamp, free text about a person, populated safety or deviation entry, or other sensitive operational detail. Every populated participant-linked or sensitive operational log—including deviation, adverse-event, safety-stop, calibration, device, timestamp, and free-text records—remains outside the package. Aggregation or de-identification does not make such a release automatic: any defined later release requires separate consent, ethics, privacy, governance, de-identification, disclosure-control, and access decisions. Source-linked design requirements would retain their population, site, device, compartment, protocol, clock, access, version, and inferential provenance. A changed item, site, device, clock, endpoint, or safety rule would trigger a new compatible or breaking protocol version, not silent replacement. No patient-level, private intake, or populated sensitive operational data belong in this reusable package. ^{[1, 10]}

Example files would contain synthetic records only until lawfully governed study data exist. They could test field types, timestamps, nesting, reason codes, missingness, censoring, protocol deviations, tables, and analysis execution, but they could not estimate a distribution, calibrate a threshold, validate an item or device, demonstrate feasibility or safety, reproduce a natural episode, or stand in for participant data. Synthetic values and scenarios would be conspicuously labelled, generated without copying a real person, excluded from evidentiary analyses, and shipped with deterministic generation instructions and expected test outputs. Any later real-data release would require separate consent, ethics, privacy, governance, de-identification, disclosure-control, and access decisions; this methods note grants none of them. ^{[1, 10]}

The package would also preserve unresolved questions rather than converting them into defaults: whether natural episodes can be captured early enough without severity-dependent missingness; which site-matched temperature and named flow or perfusion procedures are technically adequate for hands and for feet; whether Korean and English sensory items preserve the intended constructs; which recovery origins and endpoints remain stable under sensitivity analysis; whether predicted participant-level agreements and divergences replicate; and whether any later clinical use is safe and useful. Direct synchronized natural-episode evidence remains sparse. From 2026-08-16T16:06:11Z to 2026-08-17T03:40:34Z, thirteen prospectively frozen PubMed relevance-ranked query windows were screened (238 records), and no retained and passage-verified hand or foot dataset met all four synchronized natural-episode criteria within the bounded route and dates. The two site-family yields are independent; a useful hand dataset need not include feet, and a useful foot dataset need not include hands. This is not proof that unscreened ranks, other databases, inaccessible sources, or later publications contain no such data. Provoked, disease-specific, historical, 검증되지 않은 역사 자료 탐색, public-language, and synthetic materials cannot fill that gap by equivalence. Blocked NICE pages cannot support a claim, CG147 currentness remains unresolved, and the NHS Raynaud page cannot be described as newly reviewed unless its rendered review identity or an official editorial-status source materially changes. Exact safety actions and any clinical-use package therefore remain separately governed future work, not implied functions of these reuse materials. ^{[1, 10]}

Declarations

Author contributions (CRediT)

Yeonseung Choe: Conceptualization, Methodology, Clinical framing, Research-system design · Baekrokdam Research Commons (BRC): Investigation, Data curation, Writing – original draft, Evidence verification

Data availability

A machine-readable evidence ledger links each claim to its source, locator, permitted interpretation, and prohibited inference.

Funding

No external funding.

Competing interests

The authors and publisher share research infrastructure with Baekrokdam Korean Medicine Clinic. This relationship does not support any diagnosis, cause, or treatment-effect claim.

AI use disclosure

AI tools assisted retrieval, structuring, and bilingual drafting. Public claims were checked against source locators and evidence boundaries; BRC retains publication accountability.

Ethics

This work uses public literature and contains no individual patient data. Any proposed human-participant observation requires separate ethics review and consent.

References

1. Baekrokdam Research Commons. Cold hands and feet: claim, methods, and bounded-negative-yield evidence ledger [dataset]. 2026.
2. Norrbrand L, Kölegård R, Keramidas ME, Mekjavic IB, Eiken O. No association between hand and foot temperature responses during local cold stress and rewarming. European Journal of Applied Physiology. 2017;117(6):1141-1153. doi: 10.1007/s00421-017-3601-5. PMID: 28421275; PMCID: PMC5427170. Source
3. Stjernbrandt A, Björ B, Pettersson H, Lundström R, Liljelind I, Nilsson T, et al. Manifestations of cold sensitivity - a case series. International Journal of Circumpolar Health. 2020;79(1):1749001. doi: 10.1080/22423982.2020.1749001. PMID: 32264773; PMCID: PMC7178887. Source
4. Weller RS, Buczny J, Daanen HAM. Cold-induced vasodilation: A meta-analysis. Temperature. 2026;13(2):139-185. doi: 10.1080/23328940.2026.2646391. PMID: 42282897; PMCID: PMC13251525. Source
5. Maley MJ, Hunt AP, Bach AJ, Eglin CM, Costello JT. Infrared cameras overestimate skin temperature during rewarming from cold exposure. Journal of Thermal Biology. 2020;91:102614. doi: 10.1016/j.jtherbio.2020.102614. PMID: 32716864. Source
6. Havenith G, Lloyd AB. Counterpoint to ‘Infrared cameras overestimate skin temperature during rewarming from cold exposure’. Journal of Thermal Biology. 2020;92:102663. doi: 10.1016/j.jtherbio.2020.102663. PMID: 32888566. Source
7. Rosfors S, Modin A, Petrini J. Laser Doppler Measurements of Systolic Blood Pressure on the First and Second Toe in Patients with Peripheral Arterial Disease. International Journal of Angiology. 2022;31(1):52-55. doi: 10.1055/s-0041-1735237. PMID: 35221853; PMCID: PMC8881109. Source
8. Edalati M, Hastings MK, Muccigrosso D, Sorensen CJ, Hildebolt C, Zayed MA, et al. Intravenous contrast-free standardized exercise perfusion imaging in diabetic feet with ulcers. Journal of Magnetic Resonance Imaging. 2019;50(2):474-480. doi: 10.1002/jmri.26570. PMID: 30447040; PMCID: PMC6933046. Source
9. Eglin CM, Wright J, Maley MJ, Hollis S, Massey H, Montgomery H, et al. The peripheral vascular responses in non-freezing cold injury and matched controls. Experimental Physiology. 2023;108(3):420-437. doi: 10.1113/EP090721. PMID: 36807667; PMCID: PMC10103892. Source
10. Publication-object manifest