# Predictive Biomarkers for IVIG Response

## Clinical biomarkers predicting response to IVIG therapy

Predictive biomarkers for IVIG response are disease-specific: pre-treatment inflammatory markers (C-reactive protein, liver enzymes, bilirubin) predict resistance in Kawasaki disease, while post-treatment cellular immune markers (NK cells, regulatory T cells) predict response in chronic neurological autoimmune disorders.

## Abstract

Clinical biomarkers predicting IVIG response differ substantially by disease condition and reflect distinct pathophysiological mechanisms. In Kawasaki disease, pre-treatment inflammatory markers consistently predicted resistance: C-reactive protein (cut-offs 7-10 mg/dL), liver transaminases (AST ≥200 IU/L, ALT ≥80 IU/L), and total bilirubin (≥0.9 mg) demonstrated significant associations with non-response in multivariate analyses. Composite prediction scores combining multiple markers achieved 76-78% sensitivity and specificity. Post-treatment inflammatory markers measured 24 hours after infusion showed superior performance, with total leucocyte count >13.1×10⁹/L achieving 91% sensitivity and 89% specificity. In contrast, autoimmune neurological disorders demonstrated no predictive value for inflammatory markers. Instead, post-treatment cellular immune biomarkers proved informative: NK cell markers (CD56bright, KLRB1, KLRK1) measured 24 hours after IVIG initiation significantly discriminated CIDP responders from non-responders (p=0.007-0.027), while regulatory T cell frequency correlated with clinical recovery in Guillain-Barré syndrome. In dermatomyositis, higher baseline cutaneous disease activity predicted better response, whereas higher pulmonary activity predicted lower response. These divergent biomarker profiles reflect condition-specific IVIG mechanisms: acute inflammation suppression in Kawasaki disease versus immune cell modulation in chronic neurological disorders. Clinical consequences of non-response varied substantially, with Kawasaki disease non-responders experiencing coronary artery abnormalities at rates of 38-71% versus 1-5% in responders.

## Methods

We analyzed 10 sources from an initial pool of 200, using 8 screening criteria.

### Records from Elicit search

- n = 200
- Papers screened using: IVIG Therapy, Biomarker-Response Relationship, Biomarker Measurement, Response Criteria, Study Type, Human Subjects, Baseline Biomarker Assessment, Sample Size
- n = 200 Papers screened out
- n = 190 Papers included for extraction

## Screening

We screened in sources based on their abstracts that met these criteria:

- **IVIG Therapy**: Does this study involve patients treated with intravenous immunoglobulin (IVIG) therapy?
- **Biomarker-Response Relationship**: Does this study evaluate or assess the relationship between biomarkers and IVIG treatment response?
- **Biomarker Measurement**: Does this study measure or assess clinical biomarkers (laboratory, imaging, or clinical parameters)?
- **Response Criteria**: Does this study have clearly defined response criteria or outcome measures for IVIG therapy?
- **Study Type**: Is this an original research study (randomized controlled trial, cohort study, case-control study, cross-sectional study) or a systematic review/meta-analysis?
- **Human Subjects**: Does this study involve human subjects (not exclusively animal or in vitro studies)?
- **Baseline Biomarker Assessment**: Are biomarkers measured at baseline (before or at initiation of IVIG treatment) rather than only after treatment completion?
- **Sample Size**: Does this study include 10 or more patients (not a small case report or case series with fewer than 10 patients)?

## Data extraction

We asked a large language model to extract each data column below from each paper:

- **Medical Condition**:
  Extract the specific medical condition or indication for which IVIG therapy was administered.

- **Biomarker Details**:
  Extract all clinical biomarkers tested for predicting IVIG response, including:
  - Specific biomarker name/type
  - Timing of biomarker measurement relative to IVIG treatment
  - Measurement method or assay used
  - Units of measurement
  - Threshold values or cut-off points tested
  - Whether biomarker was tested individually or as part of a composite score

- **IVIG Response Definition**:
  Extract how IVIG treatment response was defined and measured, including:
  - Primary response criteria
  - Specific measurement tools or scales used
  - Time frame for assessing response
  - Definition of non-response or treatment failure
  - Whether response was measured as binary or continuous outcomes

- **Biomarker Performance**:
  Extract the predictive performance of each biomarker for IVIG response, including:
  - Statistical association with response
  - Sensitivity and specificity if reported
  - Positive and negative predictive values
  - Area under ROC curve if available
  - Whether biomarker was significant in univariate vs multivariate analysis

- **IVIG Protocol**:
  Extract details of the IVIG treatment protocol used, including:
  - IVIG dose
  - Administration schedule
  - Duration of treatment course
  - IVIG product or brand if specified
  - Any concurrent treatments or medications
  - Timing of IVIG initiation relative to disease onset

- **Study Population**:
  Extract characteristics of the study population relevant to biomarker performance, including:
  - Sample size and demographics
  - Disease severity or stage at enrollment
  - Previous treatments received
  - Inclusion and exclusion criteria
  - Setting
  - Geographic location or population type
  - Any population subgroups analyzed separately

- **Clinical Outcomes**:
  Extract the actual clinical outcomes observed in responders vs non-responders to IVIG.

## Results

### Characteristics of Included Studies

Ten studies investigating clinical biomarkers for predicting IVIG response were identified, spanning multiple autoimmune and inflammatory conditions. Five studies (50%) had full text available for detailed review.

| Study                         | Full text retrieved? | Medical Condition                    | Sample Size | IVIG Dose                              | Response Definition                                   |
|-------------------------------|----------------------|-------------------------------------|-------------|----------------------------------------|-----------------------------------------------------|
| Y. Levy et al., 1999          | No                   | Systemic lupus erythematosus       | 20          | 2 g/kg monthly over 5 days             | Binary; clinical improvement with decreased SLAM score |
| A. Irgens et al., 2012       | No                   | Myasthenia gravis with worsening weakness | 51      | 2 g/kg divided over 2 days             | Continuous; change in QMG Score at day 14         |
| C. Charles-Schoeman et al., 2025 | No                | Dermatomyositis                     | 95          | 2.0 g/kg every 4 weeks                 | Continuous; total improvement score (TIS)          |
| J. Hwang et al., 2010        | No                   | Kawasaki disease                     | 229 (206 responders, 23 non-responders)| Not specified                         | Binary; responder vs non-responder                  |
| M. Maddur et al., 2017       | Yes                  | Guillain-Barré syndrome             | 10          | 0.4 g/kg for 3-5 consecutive days     | Continuous; MRC and modified Rankin scores         |
| T. Sano et al., 2006         | No                   | Acute Kawasaki disease               | 112         | 2 g/kg within 2 days of onset         | Binary; responsive vs non-responsive                 |
| Anne K. Mausberg et al., 2020| Yes                  | Chronic inflammatory demyelinating polyneuropathy | 29 | Not specified, treatment up to 6 months | Binary; INCAT score decline ≥1 point over 6 months  |
| K. Egami et al., 2006        | No                   | Kawasaki disease                     | 320 (279 responders, 41 resistant)  | 2 g/kg within 9 days of illness       | Binary; resistance vs responder groups               |
| J. Abe et al., 2008          | No                   | Kawasaki disease                     | Not specified| High-dose (not specified)             | Binary; responsive vs nonresponsive                   |
| Mariko Fukunishi et al., 2000| No                   | Kawasaki disease                     | 82 (69 responsive, 13 non-responsive)| High-dose (not specified)             | Binary; defervescence within 5 days                 |

The studies represented diverse autoimmune conditions, with Kawasaki disease being the most extensively studied (5/10 studies). Sample sizes ranged from 10 to 320 patients. Most studies used standard high-dose IVIG protocols of 2 g/kg, though administration schedules varied. Response definitions were predominantly binary (responder vs non-responder), though some studies used continuous outcome measures.

### Predictive Biomarkers by Disease Category

#### Kawasaki Disease

Five studies examined biomarkers predicting IVIG resistance in Kawasaki disease, demonstrating substantial heterogeneity in both biomarkers identified and their measurement timing. Pre-treatment inflammatory markers emerged as consistent predictors across multiple studies.

| Study                         | Pre-treatment Biomarkers                               | Cut-off Values                            | Performance Metrics                                                                 | Post-treatment Biomarkers           |
|-------------------------------|-------------------------------------------------------|------------------------------------------|------------------------------------------------------------------------------------|-----------------------------------|
| T. Sano et al., 2006          | CRP, total bilirubin, AST, ALT, LDH, neutrophil count | CRP ≥7.0 mg, TB ≥0.9 mg, AST ≥200 IU/L | Multivariate p-values: CRP (p=0.009), TB (p<0.001), AST (p=0.002)              | Not assessed                       |
| K. Egami et al., 2006        | Age, illness days, platelet count, ALT, CRP           | Age <6 months, <4 illness days, platelet ≤30×10¹⁰/L, CRP ≥8 mg/dL, ALT ≥80 IU/L | Prediction score sensitivity 78%, specificity 76% | Not assessed               |
| Mariko Fukunishi et al., 2000| CRP, total bilirubin, LDH, gamma-glutamyltranspeptidase, hemoglobin | CRP >10 mg/dL, LDH >590 IU/L, hemoglobin <10 g/dL           | Multivariate p-values: CRP (p=.006), LDH (p=.035), TB (p=.046)                  | Not assessed                       |
| J. Hwang et al., 2010        | Neutrophil differential, CRP                           | Pre-treatment: neutrophil >51%         | Total leucocyte >13.1×10⁹/L (sens 91%, spec 89%), neutrophil >51% (sens 91%, spec 76%), total protein <72 g/L (sens 64%, spec 78%) |                                 |
| J. Abe et al., 2008          | Polycythemia rubra vera 1, G-CSF measured by mRNA and protein levels | Not specified                          | Higher in nonresponders                                                          | Not assessed                       |

C-reactive protein emerged as the most consistently identified pre-treatment biomarker, appearing in four of five Kawasaki studies. However, optimal CRP cut-off values varied considerably across studies: ≥7.0 mg in Sano et al., ≥8 mg/dL in Egami et al., and >10 mg/dL in Fukunishi et al.. Liver enzymes (AST, ALT) and total bilirubin also demonstrated predictive value in multiple studies.

Notably, J. Hwang et al. uniquely examined post-treatment biomarkers measured 24 hours after IVIG infusion. Total leucocyte count >13.1×10⁹/L demonstrated the highest combined sensitivity (91%) and specificity (89%).

### Autoimmune Neurological Disorders

Three studies examined biomarkers in distinct neurological conditions, revealing markedly different predictive approaches.

In Guillain-Barré syndrome, Maddur et al. investigated regulatory T cell (Treg) frequency as a post-treatment biomarker. Treg cells were measured by flow cytometry for surface CD4 and intracellular Foxp3 at baseline and weeks 1, 2, and 4-5 post-IVIG. Clinical recovery correlated with Treg cell response, suggesting Treg frequency represents a potential immunological biomarker.

In chronic inflammatory demyelinating polyneuropathy (CIDP), Mausberg et al. identified natural killer (NK) cell markers measured 24 hours post-IVIG initiation as predictive. CD56bright NK cells and gene transcripts KLRB1 and KLRK1 demonstrated significant association with response.

In myasthenia gravis, Irgens et al. used the QMG Score for Disease Severity as both outcome measure and potential severity stratification tool. Patients with baseline QMG >10.5 (more severe disease) demonstrated the greatest improvement, suggesting baseline severity may predict magnitude of response.

### Other Autoimmune Conditions

In systemic lupus erythematosus, Levy et al. examined autoantibodies and complement levels as potential predictors. The overall response rate was 85% (17/20 patients), with SLAM scores decreasing significantly.

For dermatomyositis, Charles-Schoeman et al. examined disease activity scores and autoantibodies. Multivariate analysis identified higher MDAAT cutaneous scores as predicting better TIS improvement.

## Synthesis

The heterogeneity in predictive biomarkers across conditions reflects fundamental differences in disease pathophysiology, IVIG mechanisms of action, and optimal measurement timing rather than inconsistent or contradictory findings.