Signature patterns predict which patients with pre-existing diseases (breast cancer, diabetes, RA) will develop MI, enabling precision prevention within disease categories.
BC→MI Group (Left): - 🚨 MASSIVE Signature 5 rise (pink/red) in final 2-4 years before MI - Sharp acceleration from year -4 onward - Signature 8 (green) shows strong depletion - Signature 4 (yellow-green) decreases
BC No MI Group (Right): - Relatively FLAT signature patterns - No dramatic Signature 5 acceleration - Stable Signature 8 - These patients have breast cancer at the same age but don’t activate cardiovascular signatures
✅ Having breast cancer alone doesn’t cause MI signature changes ✅ Only BC patients destined for MI show cardiovascular signature activation ✅ Signatures identify high-risk cancer survivors years before MI
Clinical Translation: > “Monitor Signature 5 in breast cancer survivors. Rising patterns indicate need for aggressive cardiovascular prevention and cardio-oncology referral.”
BC→MI Group (Left): - Progressive signature divergence over 8 years post-BC diagnosis - Signature 5 (light blue/cyan) rises steadily - Signature 4 (green) shows dramatic depletion - Signature 8 (orange) depletes - Multiple signatures activate simultaneously
BC No MI Group (Right): - Relatively stable signatures post-diagnosis - Signature 0 (light blue/cyan) dominates but stays flat - Minimal Signature 4 changes - No cardiovascular signature cascade
✅ Cancer treatment/disease doesn’t universally cause MI signatures ✅ Only those who will develop MI show progressive signature changes ✅ Post-cancer signature monitoring can identify at-risk survivors
Possible Mechanisms: 1. Cardiotoxic chemotherapy (anthracyclines, trastuzumab) 2. Radiation-induced cardiovascular damage 3. Shared biology between cancer and CVD 4. Lifestyle/behavioral factors post-cancer diagnosis
Clinical Translation: > “Implement longitudinal signature monitoring in cancer survivor clinics. Rising Signature 5 post-treatment triggers cardioprotective interventions (ACE inhibitors, beta-blockers, statins).”
For patients who eventually develop MI, showing signature trajectories based on pre-existing conditions: - Rheumatoid arthritis (n=509) - Diabetes (n=533) - Hypertension (n=8,650) - Hypercholesterolemia (n=5,067) - Obesity (n=1,465) - Major depressive disorder (n=1,183) - Anxiety disorder (n=615) - No transition / Control (n=14,227)
Pattern: - Dramatic Signature 5 rise starting around age 60 - Sharp acceleration in 70s-80s - Signature 8 shows depletion
Pathway Match: Pathway 2 (Multimorbid/Inflammatory)
Mechanism: - Chronic systemic inflammation drives cardiovascular disease - Cytokines (IL-6, TNF-α) promote atherosclerosis - RA is independent CV risk factor (RR ~1.5-2.0)
Clinical Implication: - RA patients with sharp Sig 5 rise need aggressive inflammatory control - Consider biologics (anti-TNF, IL-6 inhibitors) for dual benefit - Add CV prevention to RA treatment plans early
Pattern: - Signature 5 rises progressively throughout life - Signature 8 (green) shows dip/depletion - Gradual rather than sharp changes
Pathway Match: Pathway 3 (Metabolic)
Mechanism: - Metabolic dysfunction develops over decades - Insulin resistance, dyslipidemia, endothelial dysfunction - Diabetes increases MI risk 2-4 fold
Clinical Implication: - Diabetics with Sig 5 rise + Sig 8 depletion pattern = highest risk - Prioritize cardioprotective diabetes drugs (GLP-1 agonists, SGLT2 inhibitors) - More intensive BP/lipid control thresholds
Pattern: - Steady Signature 5 rise starting in 50s-60s - Continuous acceleration through 70s-80s - Large sample size = robust pattern
Pathway Match: Pathway 0 (Progressive Ischemia)
Mechanism: - Chronic pressure overload damages endothelium - Left ventricular hypertrophy - Accelerated atherosclerosis
Clinical Implication: - HTN patients with steep Sig 5 slopes need intensive BP control (<120/80) - Consider advanced therapies if signatures continue rising despite treatment - May identify treatment-resistant hypertension earlier
Pattern: - Similar to hypertension - Progressive Signature 5 elevation from age 50 onward - Steeper acceleration in 70s
Pathway Match: Pathway 0 (Progressive Ischemia)
Mechanism: - LDL-driven atherosclerosis - Plaque accumulation over decades - Classic cardiovascular risk factor
Clinical Implication: - High-cholesterol patients with rising Sig 5 need aggressive lipid lowering - Target LDL <70 mg/dL (or lower) - Consider PCSK9 inhibitors if signatures don’t stabilize on statins
Pattern: - Similar to diabetes - Signature 5 rises with age - Some Signature 8 depletion
Pathway Match: Pathway 3 (Metabolic)
Mechanism: - Adipose tissue inflammation - Insulin resistance - Dyslipidemia, hypertension (metabolic syndrome)
Clinical Implication: - Obese patients with metabolic signature pattern need intensive lifestyle intervention - Consider weight-loss medications (GLP-1 agonists) or bariatric surgery - Focus on metabolic health, not just weight
Pattern: - Signature 5 rises in 70s-80s - Later onset than other conditions - Interesting psychiatric → cardiovascular connection
Possible Mechanisms: 1. Medication effects: Antipsychotics, tricyclic antidepressants (weight gain, metabolic syndrome) 2. Stress biology: Chronic cortisol elevation, inflammation 3. Behavioral factors: Smoking, physical inactivity, poor diet 4. Shared biology: Depression and CVD have inflammatory overlap
Clinical Implication: - Screen depression patients for cardiovascular risk in 60s-70s - Consider cardioprotective antidepressants (SSRIs > tricyclics) - Integrate mental health and CV prevention
Pattern: - Similar to depression - Signature 5 rise in 70s-80s - Smaller sample size
Mechanism: - Chronic stress/sympathetic activation - Inflammation - Behavioral factors
Clinical Implication: - Monitor CV risk in anxiety patients - Stress reduction interventions may have CV benefit
Pattern: - These patients have diseases but DON’T develop MI - Signature 5 still rises (normal aging) - BUT: The rate and magnitude differ from MI progressors
Key Observation: This is the most important comparison. It shows: - Having diabetes/HTN/etc doesn’t guarantee MI - Signature dynamics add predictive value beyond disease presence - Within disease categories, signatures identify highest-risk subgroups
Clinical Implication: - Don’t treat all diabetics the same - Signatures stratify risk within disease cohorts - Personalize prevention intensity based on signature trajectories
| Pre-MI Disease | n | Signature Pattern | Pathway | Key Mechanism | Clinical Action |
|---|---|---|---|---|---|
| RA | 509 | Sharp Sig 5 rise (60s-70s) | 2 (Inflammatory) | Systemic inflammation | Aggressive anti-inflammatory Rx |
| Diabetes | 533 | Gradual Sig 5 ↑, Sig 8 ↓ | 3 (Metabolic) | Metabolic dysfunction | GLP-1s, intensive control |
| Hypertension | 8,650 | Steady Sig 5 rise (50s+) | 0 (Ischemic) | Endothelial damage | Intensive BP control |
| Hypercholesterol | 5,067 | Steady Sig 5 rise (50s+) | 0 (Ischemic) | Atherosclerosis | Aggressive lipid lowering |
| Obesity | 1,465 | Sig 5 ↑, Sig 8 ↓ | 3 (Metabolic) | Metabolic syndrome | Weight loss, GLP-1s |
| Depression | 1,183 | Late Sig 5 rise (70s-80s) | Mixed | Stress/medications | CV screening, drug choice |
| Anxiety | 615 | Late Sig 5 rise (70s-80s) | Mixed | Chronic stress | Stress management + CV Rx |
| Breast Cancer | 64-67 | Late Sig 5 acceleration | Variable | Cardiotoxicity/shared biology | Cardio-oncology monitoring |
Traditional Approach: - “Patient has diabetes → universal diabetes management” - “Patient has RA → focus on joint disease” - “Cancer survivor → monitor for recurrence”
Signature-Based Approach: - Diabetic with rising Sig 5 + depleting Sig 8 → highest CV risk, aggressive prevention - Diabetic with stable signatures → standard care - RA patient with sharp Sig 5 spike → intensify anti-inflammatory treatment - Cancer survivor with Sig 5 acceleration → cardio-oncology referral
This is precision medicine WITHIN disease cohorts.
This validates that your 4 MI pathways capture real biological mechanisms: - Inflammatory diseases (RA) → Pathway 2 signature pattern - Metabolic diseases (diabetes, obesity) → Pathway 3 signature pattern - Cardiovascular risk factors (HTN, cholesterol) → Pathway 0 signature pattern - Psychiatric diseases → Mixed/variable patterns
The pathways aren’t arbitrary - they reflect underlying biology.
Cardio-Oncology Application: - Breast cancer patients who show Signature 5 rise post-treatment are at high MI risk - Could be due to: - Cardiotoxic chemotherapy (anthracyclines, trastuzumab) - Radiation effects on heart/vessels - Shared risk factors (obesity, smoking) - Biological connection between cancer and CVD
Actionable: 1. Monitor signatures in all cancer survivors 2. Flag those with rising Sig 5 for cardio-oncology referral 3. Prophylactic cardioprotection (beta-blockers, ACE inhibitors) in high-risk patients 4. Adjust chemo regimens if signatures rise during treatment
14,227 patients with various diseases who don’t develop MI show: - Signature 5 rises with normal aging - But rate and magnitude differ from MI progressors - Having a disease doesn’t guarantee MI if signatures remain stable
Implication: Within any disease cohort, signatures identify: - High-risk subgroup: Rising Sig 5 → aggressive prevention - Low-risk subgroup: Stable signatures → standard care
This is the definition of precision medicine.
Problem: Which cancer survivors need aggressive CV monitoring? Solution: Monitor Signature 5 post-treatment - Rising Sig 5 → cardio-oncology referral, cardioprotective drugs - Stable Sig 5 → standard survivorship care - Predict chemotherapy cardiotoxicity risk before starting treatment
Problem: Which RA patients need intensified treatment? Solution: Monitor signature trajectories - Sharp Sig 5 spike → escalate to biologics, add CV prevention - Stable signatures → continue current DMARDs - Dual benefit: Control inflammation + prevent MI
Problem: Which diabetics are at highest CV risk? Solution: Stratify by signature pattern - Sig 5 ↑ + Sig 8 ↓ pattern → GLP-1 agonists, SGLT2 inhibitors, intensive BP/lipid control - Stable signatures → standard diabetes care - Earlier intervention in high-risk subgroup
Problem: How to personalize CV prevention? Solution: Risk stratify by signatures, not just diseases - HTN + rising Sig 5 → intensive BP control (<120/80) - HTN + stable Sig 5 → standard BP targets - Depression + Sig 5 rise → screen for CV disease, adjust medications
Problem: When to intensify prevention? Solution: Signature velocity as biomarker - Calculate ΔSig5/Δtime - Threshold for “dangerous” rate of rise - Trigger for advanced therapies (PCSK9 inhibitors, aggressive BP lowering)
For each disease (diabetes, RA, BC, etc.):
- AUC for MI prediction using signatures
- Compare to: Framingham score, PRS, standard risk factors
- Hypothesis: Signatures add incremental value within disease groupsCalculate rate of Signature 5 change:
- ΔSig5/Δtime in disease→MI vs disease no MI
- Optimal threshold for clinical decision-making
- Time-dependent Cox regression models
- Net reclassification improvementTest whether having multiple conditions has:
- Additive signature effects (BC + diabetes)
- Synergistic effects
- Competing effects
- Does signature pattern differ from either alone?Can signatures predict:
- Response to cardioprotective therapy?
- Chemotherapy cardiotoxicity before starting?
- Optimal timing for prevention intervention?
- Who benefits most from intensive treatment?Formal comparison across disease groups:
- Cluster analysis of signature trajectories
- Do patterns match the 4 MI pathways?
- Are there disease-specific signatures?
- Shared vs unique signature featuresPanel A: Breast Cancer Analysis - Left: BC→MI signature trajectories - Right: BC no MI trajectories - Highlight: Diverging Sig 5 in years -4 to 0
Panel B: Disease-Specific Signature Heatmap - Rows: Different pre-MI diseases (RA, diabetes, HTN, etc.) - Columns: Signatures 0-20 - Color: Deviation magnitude - Shows characteristic patterns by disease type
Panel C: Predictive Performance - ROC curves for MI prediction within each disease cohort - Signatures vs standard risk scores - Demonstrate added value
Panel D: Clinical Decision Algorithm - Flowchart: “Patient with diabetes + rising Sig 5 → Action” - Precision prevention based on signatures - Match prevention intensity to signature risk
Panel E: Signature Velocity Analysis - ΔSig5/Δtime as continuous predictor - Threshold for high-risk classification - Survival curves stratified by velocity
Reviewer Question: “Why do we need signatures when we already know disease diagnoses?”
Your Answer: > “Because diseases are heterogeneous. Two diabetics can have completely different cardiovascular trajectories. Signatures capture this heterogeneity and enable precision prevention within disease categories.”
1. External Validation: - [ ] Replicate in US cohorts (All of Us, eMERGE) - [ ] Test in diverse populations - [ ] Validate in prospective studies
2. Intervention Trials: - [ ] Signature-guided cardioprotection in BC survivors - [ ] Signature-guided anti-inflammatory therapy in RA - [ ] Signature-guided diabetes management
3. Mechanistic Studies: - [ ] What biological processes do signatures reflect? - [ ] Can we measure signatures with blood tests? - [ ] Molecular validation (RNA-seq, proteomics)
4. Clinical Implementation: - [ ] Develop signature-based risk calculator - [ ] Test in real-world clinics - [ ] Health economics analysis - [ ] Integration with EHR systems
5. Expand Disease Coverage: - [ ] Test in other high-risk conditions - [ ] COPD, CKD, autoimmune diseases - [ ] Build comprehensive disease-signature map
Old Paradigm: - “Patient has diabetes → standard diabetes care” - “Patient has RA → focus on joints” - “Cancer survivor → monitor for recurrence”
New Paradigm: - “Patient has diabetes + rising Sig 5 → highest MI risk → aggressive CV prevention” - “Patient has diabetes + stable signatures → standard care” - “Patient has RA + sharp Sig 5 spike → intensify anti-inflammatory + add CV Rx” - “Cancer survivor + Sig 5 acceleration → cardio-oncology referral”
“We demonstrate that longitudinal disease signatures stratify myocardial infarction risk within established disease cohorts, enabling precision cardiovascular prevention tailored to individual biological trajectories. This approach transforms static disease labels into dynamic risk prediction, identifying high-risk subgroups who require intensified prevention strategies.”
Right now, clinicians ask: “Does this patient have diabetes/RA/cancer?”
With signatures, clinicians will ask: “Does this patient’s biological trajectory indicate high MI risk?”
This is the future of precision preventive medicine.
Principal Investigator: Sarah Urbut, MD PhD
Collaboration Opportunities: - External validation cohorts - Cardio-oncology implementation studies - Intervention trial design - Mechanistic investigations - Clinical implementation research
Manuscript Status: In preparation for Nature
Files Referenced: -
bc_progression_Breast_cancer_matched_on_bc_age.png - BC
pre-MI signatures -
transition_deviations_age_matched_Breast_cancer_to_myocardial_infarction.png
- BC post-diagnosis signatures - image.png (×3) -
Disease-specific signature trajectories
“Signatures don’t just predict who will get MI in the general population - they predict who will get MI within high-risk disease cohorts, enabling truly personalized prevention strategies.”