A cryopreservation result is only as credible as its denominator. A post-thaw viability of 85% means little until the reader knows how many vials, donors, and independent freezing runs sit behind it. It also matters whether the vials were frozen together in one controlled-rate run, and whether the person who counted live cells knew which arm they were reading. This chapter covers the reporting layer built on top of the Phase 3 interfaces of WIA-CRYO-010. The data are first retrieved through the REST API, then analysed, then reported against the guideline that fits the study design.
Five families of guidance are used, each answering a different question:
| Guideline | Study type it governs | Typical cryopreservation use |
|---|---|---|
| STROBE | Observational studies (cohort, case-control, cross-sectional) | Retrospective analysis of banked material, such as storage duration against post-thaw quality |
| CONSORT | Randomised trials | Comparing cryoprotectant formulations or cooling protocols with random allocation |
| MIATA | T-cell assays | Functional readouts on cryopreserved peripheral blood mononuclear cells |
| MIQE 2.0 | Quantitative PCR | Gene-expression readouts on thawed cells or tissue |
| MIAPE 3.0 | Proteomics | Mass-spectrometry readouts on preserved samples |
The browser-based simulator in the reference implementation tree (cryo-research/simulator/) is the practical entry point. It lets a reader exercise the request and response contract described below before connecting a laboratory system. Treat it as a teaching tool rather than a production analysis environment.
Most real-world cryobiology data are observational. A biobank holds samples frozen over many years by different operators, in different containers, with different cryoprotectant lots. Questions such as "does storage time reduce recovery?" are therefore cohort-style questions, and STROBE applies. Its checklist has 22 items. It asks for the study design, the setting and dates, the eligibility criteria, the variables and their measurement, the efforts made to address bias, how the study size was reached, how quantitative variables were handled, the statistical methods including confounding control and missing data, and a participant flow description.
The API supports this directly. The list endpoint accepts filters and returns a total count alongside the page of results:
GET /experiments?cell_type=hepatocyte&viability_min=80&limit=50&offset=0
{
"results": [...],
"total": 145,
"page": 1,
"pageSize": 50,
"links": {
"self": "/experiments?page=1",
"next": "/experiments?page=2"
}
}The total value is the number that belongs in a STROBE flow description: how many records matched the query before any analytic exclusions. Two habits make the number defensible.
viability_min=80 is a selection criterion. If it is applied at retrieval, the analysed population is truncated on the outcome. Any model of viability fitted to that population is then biased, because only the successful runs remain. Filter on exposure and eligibility variables, and analyse the outcome afterwards.total to the final analysis set.Confounding deserves particular attention in cryopreservation. Donor, processing delay before freezing, freezing batch, freezer position, and storage temperature excursions all travel together with storage time. A STROBE-conformant report states which of these were measured, which were adjusted for, and which remain uncontrolled.
When a laboratory or clinic compares two freezing media or two cooling profiles on purpose, CONSORT governs. The elements that matter most here are the unit of randomisation, the allocation method and its concealment, blinding, and the participant flow diagram. Cryobiology adds specific traps:
The experiment record is the natural place to hold the protocol arm and the principal investigator. The error contract also helps with completeness, because a submission that omits a mandatory element is refused rather than silently stored:
{
"error": {
"code": "VALIDATION_ERROR",
"message": "Required field 'principalInvestigator' missing",
"details": {...}
}
}For clinical procedures such as the freezing of clinical specimens, the standard points to FHIR Specimen, Procedure, and Observation resources. A trial report can then trace each outcome observation back to the specimen and the procedure that produced it.
MIATA (Minimal Information About T cell Assays) exists because T-cell results, such as ELISpot counts, intracellular cytokine staining, and tetramer binding, vary strongly with how cells were handled before the assay. Cryopreserved peripheral blood mononuclear cells are the common material, so the cryopreservation history is part of the assay description. A MIATA-style report covers:
Two statistical points recur. First, replicate wells are technical replicates; they quantify pipetting and counting noise and say nothing about donor-to-donor variation. Second, a positivity threshold chosen after seeing the data inflates false-positive claims, so the rule belongs in the analysis plan. The experiment's viability and recovery results are exactly the numbers a MIATA report needs from the preservation side. A registered webhook for viability.low_alert gives the laboratory a prompt when a thawed lot falls below the level it has chosen to accept. The report should say whether such lots were excluded and on what pre-stated criterion.
MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) concerns how a qPCR result was obtained. For work on preserved material, the questions begin before the instrument: how was the sample frozen, stored, and thawed, how was RNA extracted, and how was its integrity assessed? Degradation changes Cq values and can mimic biology. The revised MIQE 2.0 guidance continues to emphasise transparent reporting of the following:
The statistical rule that matters most is the separation of replicate levels. Triplicate reactions from one extract are technical; extracts from independent freezing runs or donors are biological. Confidence intervals and tests should be built on the biological level. Reporting a tight interval from technical triplicates is one of the most common ways qPCR papers overstate precision.
MIAPE (Minimum Information About a Proteomics Experiment) covers sample preparation, separation, mass-spectrometric acquisition, and the downstream identification and quantification steps. A proteomics comparison of fresh versus cryopreserved cells depends on details that are easy to omit: lysis conditions, digestion protocol, instrument and acquisition mode, the search engine and database, the false discovery rate applied, and the quantification approach. Cryopreservation-specific information includes the freezing and thaw conditions and the interval between thaw and lysis, because protein state can change during that interval.
Statistically, proteomics is a many-hypotheses problem. Thousands of proteins are tested at once, so the report must say how multiplicity was controlled and whether the threshold applies to identification, to differential abundance, or to both. Missing values are common and not random, since low-abundance proteins drop out more often. The handling of those gaps must be stated rather than hidden inside an imputation default.
The API is not a repository for raw spectra. It carries the experiment-level record. A good practice is to cite the experiment identifier, for instance exp-2025-001, in the proteomics methods section. The accession of the deposited dataset should be kept next to that identifier in the study's own documentation, so a reader can move in either direction.
Across all five guidelines the same analytical decisions recur. Standardising them means writing them down before data collection and recording them with the analysis.
Process variables can also be derived from the monitoring channel. The WebSocket protocol streams temperature updates in a fixed message shape:
{
"type": "temperature_update",
"experimentId": "exp-2025-001",
"timestamp": "2025-01-21T10:15:23.456Z",
"data": {
"probe": "T001",
"temperature": -125.3,
"unit": "°C"
}
}A client subscribes to a channel of the form experiment.exp-2025-001.temperature. From a series of such messages, an analyst can estimate a cooling rate as the slope of temperature against time over a stated window. The report should state the probe, the window, and the sampling interval, because rate estimates depend on all three. Millisecond timestamps are only meaningful if the clocks that produced them were synchronised, which is worth confirming before any rate-dependent claim.
The working checklist for this chapter has 22 items, the same length as the STROBE checklist. It is organised by domain rather than reproduced item by item, because the five guidelines overlap heavily. What matters operationally is that each domain resolves to something retrievable from the system. A checklist item that cannot be tied to a field, an export, or a document tends to be answered from memory.
| Reporting domain | What the report must state | Where it comes from |
|---|---|---|
| Population and selection | Query criteria and the count before exclusions | Filter parameters and total from GET /experiments |
| Responsibility | Who ran and is accountable for the study | Mandatory principalInvestigator field |
| Process conditions | Cooling, storage, and thaw parameters | Temperature channel and experiment record |
| Outcomes | Viability and recovery, with time points | Experiment results; CSV export |
| Clinical linkage | Specimen, procedure, and observation provenance | FHIR Specimen, Procedure, Observation |
| Data integrity | Evidence that records met the schema | JSON Schema validation; refusal with 422 |
| Analysis | Methods, multiplicity, missing-data handling, software | Analysis plan and code, outside the API |
Two properties of the interface support the data-integrity row. Every submission is validated against JSON Schema before it is accepted, and validation failures come back as a 422 response, so a report can state that stored records conformed to the schema. Deletion is a soft delete that archives the experiment, so an analysis can still be reproduced after a record is retired from active use.
For implementers, one administrative point applies. Systems that expose this interface declare an OpenAPI contract and a conformance tier, and the reporting package should cite that declaration so a reviewer knows what level of assurance stands behind the data.
In Korea, the Ministry of Food and Drug Safety (MFDS, 식품의약품안전처) reviews applications for medicinal products and, where cell-based or tissue-based products are involved, those governed by the Advanced Regenerative Medicine and Advanced Biopharmaceuticals Act. A dossier typically needs a description of the analysis plan, the statistical methods, the results, and the quality evidence behind the data. The current form names, section numbers, and required attachments change over time, so confirm them with MFDS guidance and avoid copying a mapping from this book as final.
What can be stated reliably is the shape of the crosswalk. Each element a submission asks for should be pointed to a concrete artifact:
| Submission element (conceptual) | Supporting artifact |
|---|---|
| Study design and randomisation description | CONSORT or STROBE narrative plus the experiment records |
| Source data for tables and figures | CSV export of the experiment, with header row and UTF-8 encoding |
| Complete dataset for a reviewer | Bulk export in JSON-LD, gzip-compressed |
| Assay-level methods (T-cell, qPCR, proteomics) | MIATA, MIQE 2.0, or MIAPE 3.0 reporting section |
| Data integrity statement | Schema validation outcome and the archived (soft-deleted) record history |
The two export routes behave differently. A single experiment is exported with GET /experiments/exp-2025-001/export?format=csv. A multi-experiment package is requested asynchronously:
POST /bulk-export
{
"experimentIds": ["exp-2025-001", "exp-2025-002"],
"format": "json-ld",
"compression": "gzip"
}The bulk call returns a job ID, so the submitting team should plan for the asynchronous wait and keep the job record with the dossier working files.
Several practical features of the interface line up with how Korean laboratories and hospitals work.
Z). Korea Standard Time is UTC+9, so a freezing run logged in local time must be converted deliberately. A nine-hour offset misread inside a cooling-rate analysis is an easy way to corrupt a time series.X-RateLimit-Remaining and X-RateLimit-Reset report the budget. Paging through a large study record by record wastes that budget. Use cursor-based pagination (cursor and limit, with nextCursor and hasMore in the response) or a bulk export instead.Cursor pagination also has a statistical benefit. If new experiments are submitted while an analyst walks through the list by offset, records can shift between pages and be skipped or read twice. A cursor keeps the position stable, which protects the count that the report later cites.
This chapter tied statistical reporting to the data the system actually holds. The main points are:
total gives the starting denominator, schema validation with 422 refusals supports the integrity statement, and the temperature stream allows process variables to be derived. The CSV and bulk exports carry the data to reviewers.Reporting assumes that experiment records already move reliably between the systems that create and consume them. Laboratory instruments, hospital information systems, and long-term archives each keep their own versions of a specimen's history. The next chapter turns to that integration layer and examines how records are exchanged, reconciled, and kept consistent across those systems so that the numbers reported here can be traced back to their source.