B cell reset in autoimmune diseases
Working specification
1. Questions
- How is B cell reset characterized in autoimmune diseases? Focus on B cell immunophenotypes, with molecular, serological and functional measurements where they add information.
- Does B cell reset predict long-term response? Establish whether a measurement made during reconstitution predicts subsequent remission, beyond what is already known from clinical response and treatment exposure.
The evidence review answers these questions from the available studies. The references record the source checks and access limitations. The present evidence supports candidate biomarkers; a universal predictive definition remains unestablished.
Out of scope. Disease dependence on B cells or plasma cells belongs to imm-bcell-plasma-cause, blood–tissue agreement to imm-biopsy, and depletion kinetics to imm-bcell-indresp-est. This project is a research synthesis and study proposal; it does not specify patient treatment.
| Document | Reader | Kind | Length |
|---|---|---|---|
| Index | Finding the documents | How-to | Short |
| Specification | Planning a biomarker study | Reference | Enough to define measurements and analysis |
| Evidence review | Interpreting reset claims | Reference | Definitions, evidence and limitations |
| References | Checking a claim or choosing a paper | Reference | Source-specific notes |
2. Operational Definitions
Use separate variables for four observations. Combining them into a single label would obscure which part was actually measured.
| Observation | Operational record |
|---|---|
| Depletion | Absolute B cell count, assay detection limit, nadir and duration below that limit |
| Reconstitution | First confirmed return above a prespecified assay threshold, followed by the absolute-count trajectory |
| Biological change | Returning subsets, antigen specificity or repertoire differ from baseline at a named time after treatment and after return |
| Durable clinical response | Disease-specific response or remission maintained for a stated duration, with all concomitant and rescue therapy recorded |
For this project, phenotypic reset means a sustained change in the composition of reconstituted B cells toward an explicitly defined reference pattern. This is a proposed descriptive definition, not a validated endpoint. The review describes the frequently reported transitional/naïve predominance and reduced memory/plasmablast compartments. A functional tolerance claim additionally requires evidence about self-reactivity; surface markers alone cannot establish it.
A biomarker used to predict remission must be defined without including future remission in its definition. Otherwise the predictor contains the outcome it is supposed to predict.
3. Evidence Standards
Include human longitudinal studies with B cell or related biomarker measurements and clinical follow-up. Use mechanistic studies to interpret markers, and reviews to locate definitions and primary studies. Keep chimeric antigen receptor (CAR) T cells, anti-CD20 antibodies, B cell maturation antigen (BCMA) targeting and other interventions identifiable. Conditioning chemotherapy and continued maintenance therapy are exposures that can affect both reconstitution and outcome.
For each study extract disease, treatment, conditioning, cohort size, phenotyping denominator, measurement times, subset gates, absolute counts, clinical endpoint, follow-up, relapse count when available, and whether the biomarker preceded relapse. Record absent information as unreported or not checked. Do not infer it from a figure caption.
Distinguish:
- Coexistence: a phenotype and remission occur in the same patients.
- Prognostic association: an earlier measurement distinguishes later outcomes.
- Validated prediction: performance is evaluated in a separate cohort with a fixed assay and rule.
- Causal mediation: an analysis supports the biological change as a mechanism of benefit. Association alone does not settle this.
Check for overlapping patients before combining publications. Small series and follow-up reports must not become independent replications by being counted twice. Do not pool remission percentages across incompatible disease endpoints.
4. Proposed Sampling and Assays
The schedule below is a design proposal, to be adapted to the intervention. It is not a schedule validated by the literature review.
| Time | Measurements | Purpose |
|---|---|---|
| Before treatment and, for CAR-T, before conditioning | Absolute counts; full subset panel; serum; stored cells; disease activity; prior therapy | Separate pretreatment disease biology from conditioning effects |
| Weeks 2 and 4, then monthly through month 6 | Sensitive absolute counts; drug/CAR exposure if available; clinical activity and safety | Establish depletion and detect return |
| First quantifiable return, confirmed at the next visit | Full subset panel; serum; stored cells for repertoire/functional work | Characterize the first returning population |
| Months 3, 6, 9, 12, 18 and 24 | Subsets when sufficient cells exist; disease-specific serology; clinical activity; medication use | Measure persistence and subsequent outcome |
| Suspected flare, before rescue when feasible | Same panel and disease assessment | Distinguish changes preceding flare from effects of rescue |
| Annually thereafter | Clinical outcome, medications and targeted immune measurements | Assess durability beyond two years |
Record both time since treatment and time since B cell return. At each visit, report cells/µL, the fraction of B cells in each subset, the number of acquired B cell events, sample handling and assay limits. A percentage based on too few cells is unavailable, not zero. Use a prespecified event-count requirement qualified by the assay laboratory.
A core panel should resolve transitional, mature naïve, unswitched memory, switched memory, double-negative cells and plasmablasts. The review’s measurement table gives candidate gates and their interpretation. Add disease-relevant activated subsets and repertoire assays where samples permit. Retain serum for immunoglobulins and disease-specific autoantibodies. Track immunoglobulin replacement, vaccination and infection dates when interpreting antibody levels.
5. Proposed Analysis
Prediction at a Fixed Landmark
Use month 6 as a candidate landmark, chosen before inspecting outcomes. Among patients alive and in the required clinical response state at month 6, test whether the month-6 phenotype predicts subsequent loss of response through month 24. Report initial nonresponders and earlier failures separately; this estimand describes durability among landmark responders.
Within that population, distinguish patients with sufficient B cell return for phenotyping from patients still depleted and patients with missing samples. Analyse a phenotypic predictor within the evaluable group and report that selection explicitly. Persistent depletion can be a separate predictor in the broader population; it cannot be assigned a favourable reset score.
Start with prespecified continuous candidates: absolute switched-memory count, absolute plasmablast count and mature-naïve fraction. Do not optimise a binary cut-off against the same small dataset used to report performance. Compare a clinical model containing disease activity and treatment history with a model adding the biomarker. Adjust for disease, prior depletion, conditioning and ongoing therapy only as the sample and event count permit. Sparse relapse data support descriptive estimates, not a large multivariable model.
For an adequately sized dataset, report discrimination, calibration and uncertainty, with validation in independent patients. A marker that separates risk groups but gives inaccurate absolute risks needs recalibration before use in decisions. Estimate sample size from the expected number of relapses and desired precision, rather than the number of flow-cytometry variables available.
Time Since Reconstitution
As a secondary analysis, align trajectories on first confirmed return and examine subsequent relapse. Starting follow-up at infusion while classifying patients by a later return would give the returning group guaranteed survival time. A landmark analysis or appropriately specified time-dependent model avoids that error. An analysis restricted to returners answers a conditional question and must retain that qualification.
Clinical Endpoints and Intercurrent Events
Define the disease-specific endpoint and medication rule before pooling data. For lupus, report the chosen remission definition and steroid/maintenance requirements separately. For sclerosis, distinguish prevention of progression from remission. For myositis, distinguish response from recovery of fixed muscle damage.
A candidate durability endpoint is time to adjudicated disease flare or protocol-defined disease-directed rescue. Report each component. Treat death as a competing event for relapse and also report a sensitivity endpoint that counts death as failure. Account for informative loss to follow-up and retreatment; silently censoring patients at rescue could overstate durability.
6. Deliverables and Remaining Scientific Questions
The document set now contains definitions, a measurement table, populated clinical evidence tables and a source register. No patient-level analysis has been performed and no cut-off has been validated here.
The unresolved scientific questions are:
- Does phenotype improve prediction beyond early clinical and serological response, within a disease and treatment class?
- Does disappearance of known autoreactive clones add more information than total memory or plasmablast counts?
- How long must a changed phenotype persist to distinguish ordinary early reconstitution from a durable change?
- Which blood measurements fail because pathogenic cells persist in tissue?
- Can protective responses recover while disease-specific autoreactivity stays suppressed?
These questions require prospective samples and clinical outcomes. They cannot be completed by extending the narrative alone.