A Translational Research-Program for Compound Race Pathology: Five Framework-Distinct Trial-Design Templates
Paper 8C · Pødenphant Lund, T. (2026l) · Preprint · Live on Zenodo
If a disease is driven by several processes at once, a single-target drug is mathematically bounded against it. The Compound Race Pathology framework (Paper 8B) makes that argument formally. This paper takes the argument downstream: it turns it into five specific clinical-trial-design templates, each with a pre-specified prediction and a stated condition under which the framework would be wrong. It is a hypothesis offered to translational researchers, not clinical advice.
Clinical caution. This paper is a research hypothesis and an invitation to test it. It is not a diagnostic manual, not a treatment guideline, and it does not tell any individual what to do. Every biological and mechanistic claim is a prediction to be tested, not an established fact. The author holds no clinical credentials in the disease-domains the protocols span, and no clinical collaborator is yet attached. See What this paper is, and is not.
| DOI (concept) | 10.5281/zenodo.20059871 |
| Status | Preprint, hypothesis-and-invitation; generates no new empirical data |
| Cite-letter | 2026l |
| Scope | 5 trial-design templates across 3 disease-domains |
| Author | Tomas Pødenphant Lund [ORCID] |
TL;DR
The Compound Race Pathology framework (Paper 8B) proposes that several major chronic diseases (cancer, autoimmune conditions, post-viral fatigue, treatment-resistant depression) share a compound multi-scale mechanism: multiple bounded-probabilistic decisions (in the friction-theory (Paper 1) sense) run at several substrate scales at once, in a configuration no single target can fully reverse. This paper translates that framework into five concrete clinical-research-design proposals.
The formal claim it inherits. Under a metabolic-control-analysis extension of the RACE architecture, the achievable effect-size of a single-target intervention is bounded by the control coefficient of the targeted process with respect to overall pathology. The margin from combining interventions is proportional to the sum of control coefficients across the modified scales. Compound disease has distributed control coefficients, so single-target therapy is structurally under-powered against it. This is a quantitative bound, not a metaphor.
Three operational implications follow, and each protocol is built from them:
- Combination-design is primary — factorial designs as the default for compound disease, not a fallback after monotherapy fails.
- Substrate-vector stratification is structural — upfront multi-axis profiling before first-line treatment, because response is conditional on which process dominates in a given patient.
- Hysteresis-fighting interventions form a distinct class — sorted into Kategori A (durable substrate-replacement), B (substrate-modification needing a consolidation protocol), and C (state-perturbation needing continuous maintenance).
The five protocols (each maps to one or more of Paper 8B's R1–R10 falsifiability conditions): CAR-T plus tolerance-substrate-stabilisation for autoimmune disease; a long COVID multi-target factorial trial; ketamine plus structured CBT consolidation for treatment-resistant depression; upfront multi-axis substrate-vector profiling for TRD with a proposed 6-axis biomarker panel; and psilocybin plus substrate-vector-matched integration for TRD.
What it claims, and what it does not. Each protocol carries an empirical anchor, an immediate implementation path, a prediction target, and a falsification criterion, and each could be initiated within roughly 1–3 years on existing infrastructure. The paper generates no new data. It is hypothesis-and-invitation throughout: the framework supplies the target, rationale, and falsification criteria; domain experts would operationalise the specifics.
The central construct: compound multi-scale mechanism
The framework reads each target disease as a compound race: not one failing process but several, coupled across scales. Cancer is a compound of DNA-damage, immune-evasion, and microenvironment processes. Autoimmune disease couples microbiome, cortisol, inflammation, and recent-infection processes. Post-viral fatigue couples mitochondrial, autonomic, immune, HPA, and central-sensitisation processes. Treatment-resistant depression, when chronic, couples a substrate component, a hysteresis (trace) component, and a social component.
The consequence is the metabolic-control-analysis bound stated in the TL;DR. When control over the disease outcome is distributed across many coupled processes, no single one carries a large enough control coefficient for a single-target drug to move the outcome far. This is the framework's structural explanation for why single-target trials in these conditions so often disappoint, and why it predicts combination effect-size scaling with the summed control coefficients of whatever is modified.
Outcome-cluster versus route-identity
A second construct does much of the translational work. Clinical labels such as "treatment-resistant depression", "long COVID", or "systemic lupus erythematosus" name outcome-clusters: presentations similar enough at the symptom level to receive the same label. They do not name route-identity: the specific processes that produced that outcome in a given patient.
The same outcome-cluster is reachable by multiple routes. TRD can arrive via an inflammation route, an HPA-axis route, a microbiome route, or a hysteretic-trace route, each producing the same SSRI-non-response presentation through computationally distinct processes. This is why single-axis stratification can fail: multiple routes converge on the same single-axis marker, so the marker does not discriminate routes. The framework reads the PREDDICT 2024 null (hsCRP-stratified celecoxib augmentation showed no modification of response) this way, and flags the honest alternative reading directly: the same null is equally consistent with there being no inflammation-driven subgroup to find. The two readings separate only in a multi-axis trial where the inflammation axis adds predictive value in combination that it does not add alone.
The operational move is to treat multi-axis profiling as route-identification rather than stratification: not "which subgroup is this patient in?" but "which of several converging processes produced this outcome for this patient?". The paper notes that this same diagnostic-imprecision pattern recurs across substrates (behaviour, LLM inference, and others), but is explicit that the clinical claim does not rest on that parallel; the cross-substrate observation is offered as a recurring pattern, not a load-bearing foundation.
The Kategori A/B/C taxonomy
Wholesale substrate-perturbation interventions (HSCT, FMT, CAR-T, ketamine plus CBT, psilocybin plus integration, HBOT) are sorted into three classes by what they do to the substrate:
- Kategori A — durable substrate-replacement. Examples: HSCT for severe relapsing-remitting MS, FMT for recurrent C. difficile, CAR-T for severe autoimmune disease.
- Kategori B — substrate-modification with a consolidation requirement. Examples: psilocybin plus integration, ketamine plus CBT, HBOT for long COVID. Durability depends on the quality of the consolidation protocol.
- Kategori C — state-perturbation requiring continuous maintenance. Examples: single-dose ketamine, esketamine maintenance, ECT, sleep deprivation.
Within Kategori A, a reversible-versus-fibrotic distinction sub-classifies into A1 (cellular, reversible, drug-free remission realistic), A2 (mixed), and A3 (fibrotic, halt-of-progression only). The paper is careful here: this A1/A2/A3 gradient is read inductively from the CASTLE 2026 CAR-T outcomes (9/10 SLE in DORIS remission, 4/5 inflammatory myopathy with ACR major response, 9/9 systemic sclerosis meeting the no-progression endpoint at 24 weeks). Because it was fitted to those outcomes, it is a post-hoc reading, not a confirmed prediction; the genuinely forward test is R5 on a larger follow-up trial with longer-term relapse data.
The five protocols
Each protocol is framework-distinct (it follows from compound multi-scale mechanism, not from descriptive frameworks such as allostatic load, critical-transitions theory, or network medicine), and each is near-term implementable. Listed here in the paper's order, by funder and regulatory readiness.
- CAR-T plus tolerance-substrate-stabilisation (autoimmune). After CD19-targeted CAR-T, stabilise the tolerance substrate during B-cell repopulation (low-dose IL-2 per the LUPIL-2 protocol, with TREC/KREC, FOXP3-methylation, and TCR-repertoire monitoring). Mechanistically anchored to the Grenov et al. 2025 bioRxiv preprint (relapse via newly generated B cells from precursor pools), which is flagged as a preprint and on which the R10 rationale is stated as contingent. Tests R8 (durability gradient) and R10 (tolerance-substrate predictability). Primary engagement candidate: the Erlangen Schett group.
- Long COVID multi-target factorial. A Bayesian adaptive-platform trial allocating 3–6 matched components per substrate-vector subgroup, with hierarchical Wang-group (symptom-cluster) × Klein (biological-substrate) stratification as route-identification on two axes at once. Empirically anchored by the RECOVER-AUTONOMIC 2026 interaction effect (p = 0.004: ivabradine improved symptoms only when paired with multi-disciplinary care). Tests R1, R2, and R6.
- Ketamine plus structured CBT consolidation (TRD). A four-arm design converting a Kategori C intervention (single-dose ketamine) into a Kategori B one by adding structured post-perturbation consolidation in the BDNF plasticity window. Pre-registered against CBT-ENDURE (NCT04760652, readout 2026–2027). Tests R7.
- TRD upfront multi-axis substrate-vector profiling. Profile six axes before first-line treatment and route accordingly, against sequential-SSRI augmentation. Tests R3. The regulatory precedent is precision oncology's shift from sequential single-agent to upfront biomarker-stratified treatment.
- Psilocybin plus substrate-vector-matched integration (TRD). A post-NDA investigator-initiated supplementary trial comparing matched against standard integration, with sustained-response at 26 weeks as the primary endpoint. Tests R10. The paper is explicit that R10 remains untested for psilocybin: bimodal response alone is not evidence of a baseline-predictable subgroup.
The proposed 6-axis biomarker panel (and its precondition)
For the TRD profiling protocol, the paper proposes a specific 6-axis substrate-vector panel: (1) inflammation (hsCRP plus IL-6), (2) HPA-axis (diurnal cortisol slope), (3) microbiome (stool Shannon diversity plus key taxa), (4a) rumination (Ruminative Response Scale), (4b) calibration-tolerance (IU-12 plus MCQ-30, a candidate exploratory axis), (5) behavioural activation (BADS), and (6) sleep substrate (a 7-day wearable baseline). Target cost is roughly $400–1,500 at diagnosis, using routine clinical-laboratory assays.
The panel is presented as a specification, not a validated instrument. The paper states analytical validation as an explicit precondition before any cut-off is used for routing: inter-laboratory reliability (ICC > 0.75 per axis), assay precision (coefficient of variation < 15%), test-retest reliability, and cross-platform harmonisation. The harmonisation requirement is sharpest for the microbiome axis, where Shannon diversity is not portable across 16S, metagenomic, and metatranscriptomic pipelines, so the panel should fix one pipeline. Several cut-offs (the hsCRP/IL-6 and cortisol thresholds, and the axis-4b cut-offs validated in anxiety and OCD rather than TRD) are candidate thresholds imported from adjacent literature, not settled diagnostic boundaries.
Predictions and falsification criteria
The framework's falsifiability set (R1–R10, specified in Paper 8B) is what makes the program testable. Each protocol carries a stated condition under which it would count against the framework:
- R1, R2, R6 (long COVID). If multi-target combination produces no margin over the best single-target component and over the Palmer-Sorger best-of-N population-level benchmark in a matched-versus-unmatched factorial design, these conditions are challenged for long COVID.
- R3 (TRD profiling). If upfront 6-axis substrate-vector-stratified first-line treatment produces no margin over sequential SSRI-augmentation in an adequately-powered RCT, R3 is falsified for TRD/MDD. If multi-axis stratification adds no predictive value beyond single-axis, the multi-axis-distinct claim falls.
- R7 (ketamine plus CBT). If the CBT-ENDURE primary endpoint is not significant, or no durability margin appears across the matched / unmatched / standard / alone comparison, R7 is challenged for this protocol class.
- R10 (psilocybin). If prospective baseline-biomarker stratification fails to predict the sustained-response subgroup above chance in an adequately-powered RCT, R10 is falsified for psilocybin TRD.
- R8, R10 (CAR-T plus tolerance). If post-CAR-T tolerance-substrate-stabilisation produces no margin over CAR-T alone in sustained drug-free remission, R8 and R10 are unsupported for CAR-T autoimmune; the SLE > IIM > SSc gradient is the direct R8 test.
A separate, deliberately fenced prediction sits outside the R1–R10 core: the calibrated-retrieval test, transferred from the friction-theory inference work (Paper 1 §5.8.7). It predicts that calibration-aware intervention preserves the "recognition-commit slope" while correctness-only feedback flattens it, testable as a mediator outcome in a calibration-tolerant CBT sub-arm. The paper labels this its most speculative content: the recognition-commit slope has no validated clinical measure, the axis-4b cut-offs are TRD-unvalidated, and a null here does not bear on the R7 or R10 tests.
The trial-design templates also carry an honest power caveat applied across all five: the main-effect contrast is powered as primary, but the framework-distinct subgroup × intervention interaction contrasts (R3, R6, R10) need roughly four-fold larger samples, so at the stated per-trial n they are exploratory and hypothesis-generating, to be confirmed by pooled or adaptive-enrichment analysis across the programme rather than within a single trial. There is no defined clinical observable for "signal-share", so it is used as an interpretive frame, not claimed as a directly operationalised endpoint.
What this paper is, and is not
What it is: a translational operationalisation of one framework into five framework-distinct trial-design templates, each with an empirical anchor (peer-reviewed where not flagged), a pre-specified prediction target, sample-size methodology, and a falsification criterion; plus a proposed 6-axis TRD biomarker panel (validation stated as a precondition) and a candidate set of reporting items for substrate-stratified compound-mechanism trials.
What it is not: a clinical guideline, a diagnostic manual, or a decision-algorithm for clinicians. It makes no patient-level recommendations.
Honest limitations, as the manuscript states them. The author is an independent researcher without clinical credentials in these disease-domains, without institutional affiliation, and without a clinical consortium; this is a genuine venue-fit constraint, and domain-expert collaboration is a precondition for moving any protocol from design to conduct. The paper synthesises existing empirical anchors plus framework-prediction extensions and generates no new empirical data. The biological and mechanistic claims are predictions, not proven facts: the A1/A2/A3 gradient is read inductively from CASTLE 2026 rather than confirmed; R10 is untested for psilocybin; the cross-axis coupling prediction has no currently-existing dataset to test it against; the calibrated-retrieval prediction is explicitly fenced as speculative; and one load-bearing citation (Grenov et al. 2025) is a preprint, flagged inline, with its R10 rationale contingent on peer review. The MCA bound and the signal-budget framing are deliberate simplifications offered as interpretive structure, not as fully operationalised clinical measurements.
Connections to other papers in the series
- Paper 1 (Friction Theory) — the friction-theory, metabolic-control-analysis, and R-set foundation this program draws on; the calibrated-retrieval prediction transfers from its §5.8.7.
- Paper 8B (Compound Race Pathology) — the framework foundation. Paper 8C is its translational operationalisation; the Kategori A/B/C taxonomy and R1–R10 set are recapped here so the program is self-contained.
- Paper 8 (Pressure, Hysteresis, and Experience) — the psychiatric-clinical companion. Paper 8 covers TRD as a clinical phenomenon under the diagnosis-indexed framework; Paper 8C covers it as specific trial-design templates. The boundary is mapped explicitly: Paper 8 covers what, Paper 8C covers how and who.
- Paper 8D (Treating the Base, Not the Top) — the diagnosis-independent, below-threshold-cofactor angle, referenced as the preventive-window counterpart to upfront profiling.
- Learning — the consolidation logic behind the Kategori B protocols (post-perturbation consolidation extending durability) is the same trace-and-consolidation account that runs through the learning work.
Read the paper
The full paper is on Zenodo (concept DOI 10.5281/zenodo.20059871):