Biomarkers for Atherosclerotic Cardiovascular Events in Rheumatoid Arthritis: Towards Validation of a Biomarker-Enhanced Risk Model

In a nested case-control study of rheumatoid arthritis patients, adding the biomarkers high-sensitivity cardiac troponin T and soluble TNF receptor 1 to clinical risk factors significantly improved the predictive accuracy for future cardiovascular events, though prospective validation is still required before clinical implementation.

The Gist

Imagine you are trying to predict a storm. For decades, meteorologists have used standard tools (like barometers and wind speed) to tell everyone when a storm is coming. These tools work pretty well for the general population. But for people with Rheumatoid Arthritis (RA), the "weather" is different. They have a chronic, internal inflammation that acts like a hidden, smoldering fire, making them much more likely to have a heart attack or stroke than the standard tools predict.

This paper is like a team of scientists trying to build a better, more specialized weather forecast specifically for people with RA. They wanted to know: Can we add a few extra "sensors" to our forecast to make it more accurate?

The Problem: The Old Map Doesn't Fit

Think of the standard heart disease risk scores (like the Framingham Risk Score) as a generic map. If you try to use a map of a flat city to navigate a mountainous region, you're going to get lost. Similarly, using general population tools to predict heart risk in RA patients often misses the mark, underestimating the danger.

Scientists have tried to fix this by adding RA-specific details to the map (like how long you've had the disease or if you take steroids). It helped a little, but they wanted to see if they could do even better by looking inside the body for chemical clues.

The Experiment: The "Chemical Smoke Detectors"

The researchers gathered a group of 1,345 RA patients who had no history of heart trouble yet. They took blood samples from them and looked for eight different "chemical smoke detectors" (biomarkers) that might signal trouble brewing in the arteries.

Think of these biomarkers as different types of smoke:

• Some smoke comes from general inflammation (like hsCRP or SAA).
• Some smoke comes from the heart muscle itself being stressed (like hsTnT).
• Some smoke comes from the immune system's "alarm bells" ringing (like sTNFR1).

They followed these patients for several years to see who actually had a heart event (a "storm").

The Discovery: Two Key Sensors

When they analyzed the data, most of the smoke detectors didn't seem to add much new information. However, two specific sensors stood out as being incredibly useful:

1. hsTnT (High-sensitivity Troponin T): Imagine this as a tiny, sensitive microphone on the heart muscle. Even if the heart isn't having a full-blown heart attack, this sensor can hear the faintest "whispers" of stress or micro-damage. In RA patients, this whisper was a loud warning sign of future trouble.
2. sTNFR1 (Soluble TNF Receptor 1): Think of this as a flag raised by the immune system. It shows that the body's inflammatory fire is burning hot and is directly linked to clogging the pipes (arteries).

The Result: A Sharper Forecast

The researchers built two models:

• Model A: Used only the standard map (age, blood pressure, diabetes, etc.) plus RA details.
• Model B: Used the standard map PLUS the two new sensors (hsTnT and sTNFR1).

The verdict? Model B was significantly better at predicting who would have a heart event.

• The old model was about 76% accurate at sorting people into "high risk" vs. "low risk."
• The new model with the biomarkers jumped to 80% accurate.

While that 4% difference might sound small, in the world of medicine, it's like upgrading from a blurry photo to a high-definition image. It means doctors can better identify the patients who are truly in danger and need aggressive prevention (like stronger medication or lifestyle changes) versus those who might be safe.

The Catch: Still a Draft

The authors are careful to say this is a "preprint," meaning it's a draft that hasn't been fully peer-reviewed yet. Also, the study was done at a single hospital, so we need to see if these "sensors" work in other hospitals and with different groups of people.

The Bottom Line

This study suggests that for people with Rheumatoid Arthritis, looking at just the standard risk factors isn't enough. By adding two specific blood tests that measure heart stress and immune inflammation, doctors might soon have a much sharper tool to predict and prevent heart attacks. It's like adding a night-vision camera to a standard security system; you can finally see the threats that were previously hiding in the dark.

Technical Summary

1. Problem Statement

Patients with Rheumatoid Arthritis (RA) face a significantly elevated risk of cardiovascular disease (CVD), which remains the leading cause of mortality in this population. However, standard CVD risk stratification tools (e.g., Framingham Risk Score, QRISK2, PREVENT) derived from the general population consistently underestimate CVD risk in RA patients. This is likely due to the chronic systemic inflammation inherent to RA, which accelerates atherosclerosis in ways not captured by traditional risk factors alone.

While previous attempts to improve RA-specific risk models have incorporated RA-specific clinical variables (e.g., disease duration, glucocorticoid use) or proprietary biomarker panels, there is a lack of validated, evidence-based models that integrate specific, clinically accessible biomarkers to enhance risk prediction accuracy.

2. Methodology

The study employed a nested case-control design within the Brigham and Women's Hospital Rheumatoid Arthritis Sequential Study (BRASS) cohort.

• Study Population:
  • Source: 1,581 RA patients from the BRASS registry.
  • Exclusions: Patients with pre-existing CVD at baseline (n=213) and unconfirmed self-reported events (n=80).
  • Final Cohort: 1,288 patients (Parent Cohort).
  • Cases: 123 patients with confirmed incident atherosclerotic CVD events (MI, stroke, TIA, unstable angina, revascularization, or CV death) during follow-up.
  • Controls: 123 patients matched 1:1 to cases by age (±5 years) and sex, selected via incidence density sampling.
• Outcome Definition: A composite of non-fatal MI, non-fatal stroke/TIA, unstable angina requiring hospitalization, revascularization, or CV death. Events were adjudicated via medical record review (by an experienced adjudicator) and administrative data (Medicare/Medicaid).
• Biomarker Selection:
  • Eight candidate biomarkers were selected based on prior associations with vascular inflammation (FDG PET/CT) or general CVD risk: Adiponectin, hsCRP, Lp(a), Osteoprotegerin (OPG), hsTnT, SAA, YKL-40, and sTNFR1.
  • Samples were collected at cohort entry (or 1-year visit), stored at -80°C, and assayed in a blinded, CLIA-certified laboratory using commercial ELISA kits.
• Statistical Analysis:
  • Modeling: Conditional logistic regression (accounting for the matched design).
  • Selection Process: Biomarkers were first tested individually (age-adjusted). Those with $p < 0.05$ were retained for multivariable modeling.
  • Model Comparison: Two primary models were constructed:
    1. Clinical Model: Traditional CVD risk factors + RA clinical variables.
    2. Biomarker-Enhanced Model: Clinical Model + selected biomarkers.
    3. ERS-RA Model: A previously validated "Extended Risk Score for RA" was also tested with and without biomarkers.
  • Performance Metrics: Area Under the Curve (AUC), Akaike Information Criterion (AIC), Likelihood Ratio Tests, and Net Reclassification Index (NRI).

3. Key Contributions

• Validation of Specific Biomarkers: The study identifies high-sensitivity cardiac troponin T (hsTnT) and soluble TNF receptor 1 (sTNFR1) as the only two biomarkers from a panel of eight that significantly predict future CVD events in an independent RA cohort after adjusting for age.
• Quantification of Incremental Value: The paper provides statistical evidence that adding these two biomarkers to clinical risk factors significantly improves model discrimination and reclassification, specifically addressing the gap in current RA-specific risk tools.
• Rigorous Adjudication: Unlike many prior studies relying solely on self-report or administrative codes, this study utilized rigorous medical record adjudication for CVD outcomes.

4. Key Results

• Cohort Characteristics: The median age was 63 years; 77% were female. Cases had higher rates of diabetes, current tobacco use, and oral glucocorticoid use compared to controls.
• Biomarker Association:
  • In age-adjusted models, hsTnT (OR 1.80 per SD increase) and sTNFR1 (OR 1.45 per SD increase) were significantly associated with CVD events.
  • Other biomarkers (hsCRP, Lp(a), OPG, etc.) did not show significant associations in the multivariable context.
• Model Performance:
  • Clinical Model (Traditional + RA factors): AUC = 0.758 (95% CI 0.689–0.829).
  • Biomarker-Enhanced Model (Clinical + hsTnT + sTNFR1): AUC increased to 0.802 (95% CI 0.718–0.998).
  • Statistical Significance: The improvement in AUC was significant (Likelihood Ratio Test $p = 0.004$). The AIC improved from 320.5 to 313.3, indicating a better model fit.
  • ERS-RA Model: Adding biomarkers to the ERS-RA score also improved the AUC from 0.743 to 0.756 ($p=0.002$).
• Reclassification (NRI):
  • The Net Reclassification Index (NRI) for the clinical model with added biomarkers was 16.3% (95% CI 1.0–31.8).
  • Crucial Nuance: The improvement in reclassification was driven entirely by the controls (patients who did not have events). The model better identified low-risk individuals, but showed no improvement in reclassifying actual cases.

5. Significance and Limitations

Significance:

• The study demonstrates that incorporating hsTnT (a marker of subclinical myocardial injury) and sTNFR1 (a marker of TNF-mediated inflammation) into risk models significantly enhances the ability to stratify CVD risk in RA patients.
• This supports a shift toward "biomarker-enhanced" risk stratification, potentially allowing clinicians to better identify RA patients who require aggressive preventative care (e.g., statins, CAC scanning) versus those who do not.
• The findings suggest that general population scores are insufficient and that RA-specific models must account for both inflammatory burden and subclinical cardiac injury.

Limitations:

• Sample Size: The number of events (n=123) limits statistical power, particularly for detecting interactions or smaller effect sizes.
• Generalizability: The cohort was drawn from a single academic center (Brigham and Women's Hospital), which may limit applicability to broader, diverse RA populations.
• Lipid Data: Actual lipid parameters were missing for many patients, preventing the use of standard lipid-based risk scores (e.g., Pooled Cohort Equations) and necessitating the use of proxy variables.
• Reclassification Bias: The NRI improvement was driven by controls, suggesting the model is better at ruling out risk than confirming it in high-risk patients.
• Preprint Status: As a preprint, the study has not yet undergone peer review.

Conclusion:
The authors conclude that while the addition of hsTnT and sTNFR1 improves statistical model fit, prospective validation in external cohorts is required before these biomarkers can be integrated into clinical decision-making for CVD prevention in RA. Future studies should assess whether these biomarkers improve the prediction of Coronary Artery Calcium (CAC) scores, a gold standard for guiding preventative therapy.