RTX-VAX: Transparent Vaccine-Response Readiness Stratification for Non-Live Vaccination During or After Rituximab Therapy in Rheumatic and Autoimmune Disease

Abstract

Vaccination planning around rituximab remains a common clinical problem in rheumatology and autoimmune care. Clinicians often know that vaccine responses are reduced after B-cell depletion, but bedside decisions still require a practical judgment: vaccinate now, delay to a better window, or proceed despite expected blunting because infection risk is immediate. RTX-VAX is an executable Python skill that converts this decision into a transparent, auditable readiness score for non-live vaccines. The model weights time since last rituximab infusion, time until the next cycle, B-cell reconstitution when available, baseline IgG reserve, prednisone exposure, concomitant methotrexate, age, prior documented vaccine nonresponse, and urgency context. It outputs a 0-100 vaccine-readiness score, expected humoral-response band, timing-window classification, modifiable barriers, and a recommendation statement. RTX-VAX is intentionally simple, dependency-free, and reviewer-runnable. It is not a validated probability model and does not replace clinician judgment, but it addresses a real clinical coordination problem with transparent logic aligned to ACR/EULAR vaccination guidance and contemporary rituximab immunogenicity literature.

Clinical methodology

Problem being solved

Rituximab-treated patients frequently need influenza, pneumococcal, COVID-19, hepatitis B, or recombinant zoster vaccination. The core clinical tension is between immunogenicity optimization and infection-prevention urgency. Current practice is often inconsistent because timing decisions are discussed narratively rather than computed explicitly.

Design principles

1. Guideline-concordant timing first. Time since the last rituximab infusion and time before the next infusion are the strongest practical anchors for the score.
2. Biologic plausibility. B-cell reconstitution and IgG reserve are incorporated because they help explain why two patients with the same calendar timing may not have the same response potential.
3. Modifiable treatment burden. Prednisone and methotrexate remain clinically relevant response suppressors and are made visible in the output.
4. Urgency matters. A perfect vaccine window is not always possible; influenza season or local outbreak pressure may justify proceeding in a poor-response window.
5. Executable transparency. Pure Python, no dependencies, no hidden weights, no black box.

Output interpretation

RTX-VAX does not claim to estimate exact seroconversion probability. Instead, it returns a readiness band for practical planning:

• very low expected humoral response
• low expected humoral response
• intermediate expected humoral response
• favorable expected humoral response
• high expected humoral response

This is meant to support timing conversations, not replace them.

Why this score exists

Rituximab is used across rheumatoid arthritis, ANCA-associated vasculitis, connective tissue disease, and other autoimmune conditions. These patients often face high infection burden precisely when vaccine response is least reliable. A transparent tool helps clinicians document why vaccination was advanced, deferred, or performed in a suboptimal window.

Demo output

Running python3 skills/rtx-vax/rtx_vax.py prints three scenarios:

1. RA early post-rituximab poor window → low score, multiple barriers, defer-if-safe recommendation.
2. AAV maintenance favorable window → favorable score and proceed-now recommendation.
3. Urgent influenza before next cycle → suboptimal score but vaccinate-now recommendation because delay is riskier.

Limitations

• Not a fitted or prospectively validated prediction model.
• Designed for adult autoimmune/rheumatic populations and non-live vaccines.
• Humoral response is only one dimension of vaccine benefit; cellular immunity is incompletely captured.
• Disease-specific factors, prior infection history, and local epidemiology are not fully modeled.
• B-cell counts and IgG values are optional; without them the model is less individualized.

Authors

Dr. Erick Zamora-Tehozol (ORCID: 0000-0002-7888-3961), DNAI, RheumaAI

References

1. Bass AR, Chakravarty E, Akl EA, et al. 2022 American College of Rheumatology Guideline for Vaccinations in Patients With Rheumatic and Musculoskeletal Diseases. Arthritis Rheumatol. 2023;75(3):449-464. DOI: 10.1002/acr.25045
2. Furer V, Rondaan C, Heijstek MW, et al. 2019 update of EULAR recommendations for vaccination in adult patients with autoimmune inflammatory rheumatic diseases. Ann Rheum Dis. 2020;79(1):39-52. DOI: 10.1136/annrheumdis-2019-215882
3. van Assen S, Holvast A, Benne CA, et al. Humoral responses after influenza vaccination are severely reduced in patients with rheumatoid arthritis treated with rituximab. Arthritis Rheum. 2010;62(1):75-81. DOI: 10.1002/art.25033
4. Schultz K, Jannat-Khah D, Spiera R. B Cell Reconstitution is Associated With COVID-19 Booster Vaccine Responsiveness in Patients Previously Seronegative Treated With Rituximab. J Rheumatol. 2023;50(3):420-424. DOI: 10.3899/jrheum.220475
5. Hua C, Barnetche T, Combe B, Morel J. Effect of methotrexate, anti-tumor necrosis factor α, and rituximab on the immune response to influenza and pneumococcal vaccines in patients with rheumatoid arthritis: a systematic review and meta-analysis. Arthritis Care Res (Hoboken). 2014;66(7):1016-1026. DOI: 10.1002/acr.22246

skill_md

#!/usr/bin/env python3
"""
RTX-VAX: Transparent vaccine-response readiness stratification for non-live
vaccination during or after rituximab therapy in rheumatic and autoimmune disease.

Authors: Dr. Erick Zamora-Tehozol (ORCID: 0000-0002-7888-3961), DNAI, RheumaAI
"""
from __future__ import annotations

from dataclasses import dataclass
from typing import List, Optional


@dataclass
class RTXVaxAssessment:
    label: str
    vaccine_type: str = "non-live"
    months_since_last_rituximab: float = 0.0
    weeks_until_next_rituximab: Optional[float] = None
    cd19_cells_per_ul: Optional[float] = None
    igg_mg_dl: Optional[float] = None
    prednisone_mg_day: float = 0.0
    methotrexate: bool = False
    age_years: int = 50
    prior_documented_nonresponse: bool = False
    urgent_vaccination_context: bool = False
    active_major_flare: bool = False


def clamp(value: float, low: float, high: float) -> float:
    return max(low, min(high, value))


def timing_points(months_since_last_rituximab: float) -> int:
    if months_since_last_rituximab < 3:
        return 0
    if months_since_last_rituximab < 6:
        return 15
    if months_since_last_rituximab < 9:
        return 30
    return 35


def next_dose_points(weeks_until_next_rituximab: Optional[float]) -> int:
    if weeks_until_next_rituximab is None:
        return 0
    if weeks_until_next_rituximab < 2:
        return -15
    if weeks_until_next_rituximab < 4:
        return -10
    if weeks_until_next_rituximab < 8:
        return -5
    return 0


def bcell_points(cd19_cells_per_ul: Optional[float]) -> int:
    if cd19_cells_per_ul is None:
        return 0
    if cd19_cells_per_ul <= 0:
        return -5
    if cd19_cells_per_ul <= 10:
        return 8
    if cd19_cells_per_ul <= 50:
        return 18
    return 25


def igg_points(igg_mg_dl: Optional[float]) -> int:
    if igg_mg_dl is None:
        return 0
    if igg_mg_dl < 500:
        return -12
    if igg_mg_dl < 700:
        return -4
    return 6


def prednisone_points(prednisone_mg_day: float) -> int:
    if prednisone_mg_day >= 20:
        return -12
    if prednisone_mg_day >= 10:
        return -8
    if prednisone_mg_day > 0:
        return -3
    return 0


def age_points(age_years: int) -> int:
    if age_years >= 75:
        return -6
    if age_years >= 65:
        return -3
    return 0


def score_assessment(a: RTXVaxAssessment) -> dict:
    components = {
        "base": 35,
        "time_since_last_rituximab": timing_points(a.months_since_last_rituximab),
        "time_until_next_rituximab": next_dose_points(a.weeks_until_next_rituximab),
        "b_cell_reconstitution": bcell_points(a.cd19_cells_per_ul),
        "baseline_igg": igg_points(a.igg_mg_dl),
        "prednisone": prednisone_points(a.prednisone_mg_day),
        "methotrexate": -8 if a.methotrexate else 0,
        "age": age_points(a.age_years),
        "prior_nonresponse": -10 if a.prior_documented_nonresponse else 0,
        "active_major_flare": -6 if a.active_major_flare else 0,
        "urgency_credit": 4 if a.urgent_vaccination_context else 0,
    }
    raw_score = sum(components.values())
    score = round(clamp(raw_score, 0, 100), 1)

    if score < 20:
        band = "Very low expected humoral response"
    elif score < 40:
        band = "Low expected humoral response"
    elif score < 60:
        band = "Intermediate expected humoral response"
    elif score < 80:
        band = "Favorable expected humoral response"
    else:
        band = "High expected humoral response"

    favorable_window = (
        a.months_since_last_rituximab >= 6
        and (a.weeks_until_next_rituximab is None or a.weeks_until_next_rituximab >= 4)
    )
    if favorable_window:
        window = "Favorable"
    elif a.months_since_last_rituximab >= 4:
        window = "Suboptimal but usable"
    else:
        window = "Poor"

    barriers: List[str] = []
    if a.months_since_last_rituximab < 6:
        barriers.append("Last rituximab infusion was <6 months ago")
    if a.weeks_until_next_rituximab is not None and a.weeks_until_next_rituximab < 4:
        barriers.append("Next rituximab dose is <4 weeks away")
    if a.cd19_cells_per_ul is not None and a.cd19_cells_per_ul <= 10:
        barriers.append("Minimal or absent B-cell reconstitution")
    if a.igg_mg_dl is not None and a.igg_mg_dl < 700:
        barriers.append("Low baseline IgG reserve")
    if a.prednisone_mg_day >= 10:
        barriers.append("Prednisone dose may blunt vaccine response")
    if a.methotrexate:
        barriers.append("Concomitant methotrexate may reduce immunogenicity")
    if a.prior_documented_nonresponse:
        barriers.append("Prior documented vaccine nonresponse")

    if favorable_window and score >= 60:
        recommendation = "Proceed now: timing window is favorable for non-live vaccination."
    elif a.urgent_vaccination_context and a.vaccine_type.lower() == "influenza":
        recommendation = "Vaccinate now despite a suboptimal window because influenza timing is urgent; counsel that response may be blunted."
    elif score >= 40:
        recommendation = "Vaccination is reasonable now, but response may improve if timed farther from rituximab and at least 4 weeks before the next cycle."
    else:
        recommendation = "If clinically safe, delay vaccination until better B-cell recovery / rituximab timing; if risk of waiting is high, vaccinate with counseling about reduced response."

    return {
        "label": a.label,
        "score": score,
        "response_band": band,
        "window": window,
        "components": components,
        "modifiable_barriers": barriers,
        "recommendation": recommendation,
    }


def demo() -> List[dict]:
    scenarios = [
        RTXVaxAssessment(
            label="RA early post-rituximab poor window",
            vaccine_type="pneumococcal",
            months_since_last_rituximab=2.0,
            weeks_until_next_rituximab=12,
            cd19_cells_per_ul=0,
            igg_mg_dl=540,
            prednisone_mg_day=12.5,
            methotrexate=True,
            age_years=61,
            prior_documented_nonresponse=True,
        ),
        RTXVaxAssessment(
            label="AAV maintenance favorable window",
            vaccine_type="COVID-19 booster",
            months_since_last_rituximab=7.5,
            weeks_until_next_rituximab=8,
            cd19_cells_per_ul=68,
            igg_mg_dl=910,
            prednisone_mg_day=0,
            methotrexate=False,
            age_years=56,
            prior_documented_nonresponse=False,
        ),
        RTXVaxAssessment(
            label="Urgent influenza before next cycle",
            vaccine_type="influenza",
            months_since_last_rituximab=4.5,
            weeks_until_next_rituximab=2,
            cd19_cells_per_ul=7,
            igg_mg_dl=760,
            prednisone_mg_day=5,
            methotrexate=False,
            age_years=70,
            urgent_vaccination_context=True,
        ),
    ]
    return [score_assessment(s) for s in scenarios]


if __name__ == "__main__":
    print("=" * 74)
    print("RTX-VAX: Rituximab Vaccine-Response Readiness Stratification")
    print("=" * 74)
    for result in demo():
        print(f"\n{result['label']}")
        print(f"  Score: {result['score']}")
        print(f"  Band: {result['response_band']}")
        print(f"  Window: {result['window']}")
        print(f"  Recommendation: {result['recommendation']}")
        if result["modifiable_barriers"]:
            print("  Barriers:")
            for barrier in result["modifiable_barriers"]:
                print(f"    - {barrier}")
    print("\nReferences:")
    print("  1. Bass AR et al. Arthritis Rheumatol. 2023;75(3):449-464. DOI: 10.1002/acr.25045")
    print("  2. Furer V et al. Ann Rheum Dis. 2020;79(1):39-52. DOI: 10.1136/annrheumdis-2019-215882")
    print("  3. van Assen S et al. Arthritis Rheum. 2010;62(1):75-81. DOI: 10.1002/art.25033")
    print("  4. Schultz K et al. J Rheumatol. 2023;50(3):420-424. DOI: 10.3899/jrheum.220475")
    print("=" * 74)

SKILL.md

name: rtx-vax description: Transparent vaccine-response readiness stratification for non-live vaccination during or after rituximab therapy in rheumatic and autoimmune disease.

RTX-VAX

RTX-VAX estimates how likely a patient on rituximab is to mount a useful non-live vaccine response and whether the current timing window is favorable, suboptimal, or poor.

Clinical problem

Rituximab reduces vaccine immunogenicity, but clinicians still have to decide whether to vaccinate now, wait for a better window, or proceed because delay is riskier than imperfect response. In practice this decision depends on:

• time since the last rituximab infusion
• time until the next infusion
• evidence of B-cell reconstitution when available
• background immunosuppression
• baseline immunoglobulin reserve
• whether the vaccine is urgent (for example influenza circulation or outbreak context)

A transparent score helps standardize these discussions.

What it outputs

• vaccine-readiness score (0-100)
• response band: very low, low, intermediate, favorable, high
• window classification: poor, suboptimal, favorable
• guideline-aligned timing interpretation
• modifiable barriers that may improve response
• practical recommendation summary

Intended use

• adult rheumatic and autoimmune disease care
• non-live vaccines such as influenza, pneumococcal, COVID-19, hepatitis B, recombinant zoster
• pre-vaccination planning around rituximab cycles

Run

python3 rtx_vax.py

Clinical methodology

RTX-VAX is a transparent weighted heuristic, not a fitted regression model. It prioritizes factors repeatedly associated with reduced immunogenicity in the literature:

1. Time since last rituximab — strongest practical timing variable.
2. Time until next rituximab — short remaining window lowers expected response.
3. B-cell reconstitution when measured — recovery improves humoral response.
4. IgG reserve — low baseline IgG suggests weaker humoral reserve and higher infection concern.
5. Prednisone and concomitant methotrexate — both can blunt vaccine response.
6. Age and prior documented vaccine nonresponse — modest adverse modifiers.
7. Urgency override — for influenza/outbreak settings, vaccinating now may still be preferable even in a poor response window.

References

1. Bass AR, Chakravarty E, Akl EA, et al. 2022 American College of Rheumatology Guideline for Vaccinations in Patients With Rheumatic and Musculoskeletal Diseases. Arthritis Rheumatol. 2023;75(3):449-464. DOI: 10.1002/acr.25045
2. Furer V, Rondaan C, Heijstek MW, et al. 2019 update of EULAR recommendations for vaccination in adult patients with autoimmune inflammatory rheumatic diseases. Ann Rheum Dis. 2020;79(1):39-52. DOI: 10.1136/annrheumdis-2019-215882
3. van Assen S, Holvast A, Benne CA, et al. Humoral responses after influenza vaccination are severely reduced in patients with rheumatoid arthritis treated with rituximab. Arthritis Rheum. 2010;62(1):75-81. DOI: 10.1002/art.25033
4. Schultz K, Jannat-Khah D, Spiera R. B Cell Reconstitution is Associated With COVID-19 Booster Vaccine Responsiveness in Patients Previously Seronegative Treated With Rituximab. J Rheumatol. 2023;50(3):420-424. DOI: 10.3899/jrheum.220475
5. Hua C, Barnetche T, Combe B, Morel J. Effect of methotrexate, anti-tumor necrosis factor α, and rituximab on the immune response to influenza and pneumococcal vaccines in patients with rheumatoid arthritis: a systematic review and meta-analysis. Arthritis Care Res (Hoboken). 2014;66(7):1016-1026. DOI: 10.1002/acr.22246

Limitations

• This is not a validated probability model and should not be used as a sole treatment or vaccination decision-maker.
• It is intended for non-live vaccines; live-vaccine decisions require separate guideline review.
• B-cell and IgG data are optional but improve interpretability; if unavailable, the score is less individualized.
• Cellular immunity may persist despite poor humoral response, so a low score does not prove zero clinical benefit.
• Disease flare risk, local epidemiology, and patient preferences still matter.