imm-biopsy: working specification
Inferring tissue B-cell depletion from blood and biopsy measurements
1. Question and scope
Determine what a peripheral blood measurement or a biopsy from tissue A can establish about B cells in tissue B in the same patient. The intended use is interpretation of pharmacodynamic measurements and design of a tissue-sampling program for B-cell-depleting therapies in autoimmune disease.
The initial evidence comes from the supplied brief. Its study-level findings remain provisional until the checks in References are completed. The brief is a targeted literature selection; a systematic search has not been performed.
Out of scope: individual treatment recommendations, a ranking of therapies from unmatched studies, and a claim that one negative biopsy establishes whole-body eradication of B-lineage cells.
The project should answer:
- How well do blood counts track local tissue burden, depletion and recovery?
- Given a negative biopsy in one tissue, how often are cells still detected in another tissue sampled in the same patient at a comparable time?
- How much disagreement is attributable to spatial sampling within a tissue, assay limitations, phenotype definitions or different sampling times?
- Which residual or recovering populations are associated with persistent disease, relapse or clinical benefit?
- What additional sampling would most reduce uncertainty for a specified development decision?
2. Measurement definitions
A negative result means no cells meeting the assay’s definition were detected in the material examined. Report the phenotype, count or density, denominator, assay threshold, anatomical location and time. For example: “No CD19-positive cells detected in the analyzed lymph-node sections at week 8,” followed by the analyzed area and detection rule when available.
CD19 and CD20 are cluster-of-differentiation surface markers used to identify overlapping B-cell populations. Keep marker-defined populations separate from functional claims about pathogenic cells. Extract naïve and memory B cells, plasmablasts and plasma cells separately where the study permits. A plasmablast is a proliferating antibody-secreting cell; a plasma cell is a differentiated antibody-secreting cell. Record the actual marker panel used to distinguish them.
| Quantity | Required definition | Interpretation limit |
|---|---|---|
| Blood burden | Absolute cells per volume, or a reported fraction with its denominator | A percentage can change when other cell populations change |
| Tissue burden | Cells per analyzed area, volume or mass; or fraction of evaluable recovered cells | Histology preserves location; a cell suspension represents recovered cells |
| Local depletion | Change from pretreatment burden using comparable sampling and assays | Without baseline tissue data, percentage depletion is not estimable |
| Nondetection | Observed zero with sample amount, detection rule and assay sensitivity | Zero is an observation, not proof of biological absence |
| Recovery | Reappearance over time of a specified phenotype | A single later sample cannot separate survival from regeneration |
| Disease-relevant depletion | Change in a population linked to disease activity or pathogenic function | Marker depletion and clinical benefit are separate outcomes |
For a measurable baseline, define percentage reduction as \(100(1-B_{\mathrm{post}}/B_{\mathrm{pre}})\), retaining the original units. If the post-treatment value is censored below a threshold, carry that censoring into the result. Do not substitute zero and report an exact 100% reduction. A pretreatment zero makes this percentage undefined.
3. Evidence designs
Classify each observation by the inference its sampling supports.
| Sampling design | Question it can address | Restriction |
|---|---|---|
| Blood plus one tissue in one patient | Does blood track that tissue? | Does not establish agreement between two tissues |
| Two tissues in one patient at matched times | Is depletion concordant across those tissues? | Does not describe unsampled organs |
| Serial sampling of one tissue | How does local burden change over time? | Does not establish synchronous changes elsewhere |
| Separate regions or cores in one organ | How reproducible is local sampling? | Sections from one core are not independent spatial replicates |
| Different organs from different patients | Has depletion been observed in several tissue types? | Cannot estimate within-patient cross-tissue agreement |
“Matched” requires documented dates relative to treatment, and a justified time window specified before analysis. Store the actual interval between samples. Keep marrow aspirates, marrow core biopsies and other cell suspensions distinct.
The initial brief identifies the following evidence to check:
- Minopoulou 2024: a same-patient node–marrow discordance observation, reported in a conference abstract.
- Albach 2026: paired marrow and synovial observations in a mechanistic subset of the COMPARE trial.
- Teng 2007: blood, marrow and synovium before and after rituximab; evaluable matched counts need confirmation.
- Tur 2024/2025: serial lymph-node samples, with other organs contributed by additional patients.
- Feng 2025: blood and marrow cell populations during recovery after dual-target therapy.
4. Sources of variability
Represent the following separately when extracting evidence and considering a model. Their relative contributions have not been estimated by this project.
| Source | Information to collect | Question for analysis |
|---|---|---|
| Within-tissue sampling | Site, core count, core size, section spacing, analyzed area and cellular aggregates | Could the sampled region miss residual cells? |
| Between-tissue biology | Organ, disease involvement, baseline burden and tissue architecture | Are differences consistent across patients? |
| Cell population | Marker panel, phenotype, target expression and clonality when available | Are the assays measuring the same population? |
| Assay and processing | Staining, flow cytometry or sequencing; recovery, viability and thresholds | Could measurement differences explain discordance? |
| Time | Treatment, conditioning, biopsy and blood-sampling dates | Could the samples represent different phases of depletion or recovery? |
| Patient and treatment | Disease, prior therapy, regimen and conditioning | To which patients and treatments could the finding transfer? |
| Selection into biopsy | Clinical indication, research sampling, consent and missing repeat samples | Were persistent disease or accessible tissues preferentially sampled? |
Tissue exposure, effector-cell availability and local survival conditions are candidate biological explanations. Record measured evidence separately from a mechanism proposed by an author or by this project.
5. Extraction requirements
Use one record per patient, specimen, time, assay and phenotype when individual information is available. Link paired observations by patient identifier. Retain aggregate-only reports as aggregate records; do not invent individual observations from a study total.
| Field group | Required fields |
|---|---|
| Provenance | Study identifier, linked reports, source URL, page/figure/table, extraction date, verification status |
| Denominators | Total enrollment, tissue subset, evaluable patients and paired patients for each comparison |
| Clinical context | Disease, involved organs, prior B-cell therapy, regimen, target, conditioning and concomitant treatment |
| Sample | Patient identifier if reported, tissue, region, collection method, date, relation to treatment and baseline availability |
| Assay | Platform, marker panel, gating/staining definition, sample amount, recovery and detection/quantification limits |
| Result | Raw count, density or fraction, units, denominator, nondetection/censoring and uncertainty |
| Pairing | Other specimen identifier, same-patient status, timing difference and phenotype comparability |
| Clinical outcome | Disease measure, response or relapse definition, date and relationship to sampling |
| Missingness | Not reported, not collected, failed assay or unavailable source; reason when known |
Keep CD19-only chimeric antigen receptor T-cell (CAR-T) therapy, dual-target CAR-T therapy and anti-CD20 antibodies separate. Record lymphodepleting conditioning and treatment sequence. Add T-cell engagers if a subsequent search identifies relevant human tissue measurements.
Check for repeated publication of the same patients before counting evidence across reports. Total enrollment must never stand in for a missing paired sample count.
6. Analysis plan
Start with an evidence map by tissue pair, treatment, phenotype, assay and time. The first deliverable is a table of which comparisons are actually observed, with missing comparisons explicit. Draw patient-level blood and tissue trajectories only where repeated measurements are available.
For a prespecified tissue pair A and B, phenotype and time window, tabulate both detected, A detected/B negative, A negative/B detected and both negative. A directly relevant quantity is
\[ \widehat{P}(B\text{ detected}\mid A\text{ negative}) = \frac{n(A\text{ negative},B\text{ detected})} {n(A\text{ negative},B\text{ evaluable})}. \]
This estimates observed discordance among the sampled patients under the two assays. Its complement describes observed nondetection in B; it does not estimate absence of cells throughout B or throughout the body. If the denominator is zero, report the quantity as not estimable. Report the numerator, denominator and an interval with the estimate. Repeated visits from one patient require an analysis that accounts for within-patient dependence.
Assess quantitative association separately from categorical agreement. A correlation between blood and tissue values does not by itself establish that a negative blood sample predicts a negative tissue sample. Preserve assay units and thresholds; avoid combining incomparable densities and cell fractions.
A later hierarchical model could separate variation between patients, tissues and replicate samples, while representing values below detection limits. Estimate only components supported by the sampling design. Without replicate cores, within-tissue variation and measurement error may remain inseparable; without paired tissues, cross-tissue dependence cannot be learned from these studies alone. Any simulation using assumed dependence must label it as an assumption and vary it in sensitivity analysis.
The supplied brief provides no validated numerical probability that a negative node predicts depletion in every disease-relevant tissue. Leave that quantity unestimated unless new evidence supports a narrower, explicitly defined claim.
7. Work sequence and acceptance criteria
- Verify the five source entries against the specific passages and figures listed in the reading queue. Preserve unresolved counts and timing as missing.
- Extend the search to peripheral blood–tissue comparisons, paired human tissues and repeated cores. Record search terms, databases, dates and exclusions before describing the review as systematic.
- Build the extraction table and evidence map. Each finding must have a source location, status and correct paired denominator.
- Assess whether any tissue pair has enough comparable observations for quantitative analysis. If it does not, report the observations and the evidence gap without a pooled probability.
- Draft a sampling proposal for a specified disease and regimen, stating the tissues, cell populations, timing, assays and development decision it serves.
The working synthesis is ready for review when every claim distinguishes local measurement from extrapolation, every negative result retains its measurement limits, and study conclusions remain separate from proposed mechanisms.
8. Decisions still open
- Which disease, treatment and development decision should the first sampling proposal address?
- Which tissue is feasible as the routinely sampled site, and which disease-relevant tissue would it need to represent?
- Is the desired endpoint initial depletion, duration below a threshold, recovery of particular populations, or a relationship to clinical benefit?
- Which paired data and assay-validation materials can be obtained beyond the published reports?
Document responsibilities
| Document | Reader | Kind | Length |
|---|---|---|---|
| Project index | Someone finding the project | How-to | Short |
| Working specification | Someone conducting the review and planning analysis | Reference | Enough to define measurements, extraction and decisions |
| References | Someone verifying a claim | Reference | One consistent entry per source |
| Supplied evidence brief | Someone checking the starting context | Source record | Preserved as received |