[1] "Filtered to common cohort. Merged N: 1184"[1] "Before NA removal: 1184"[1] "After NA removal: 1037"9 Clinical Impact Assessment
9.1 Objective
Evaluate how AI-assisted diagnosis affects treatment decisions and clinical outcomes for breast cancer patients.
Note for Pathologist: This is the “So What?” section. We translate the changes in scores (e.g., Ki67 going from 19% to 21%) into actual treatment decisions (e.g., “Chemotherapy Recommended”). we calculate how many patients would have their treatment plan changed based on the AI’s input—whether that means adding chemotherapy, avoiding it, or changing eligibility for anti-HER2 drugs.
9.2 Setup
9.3 Load Data
9.4 Treatment Decision Framework
Based on standard breast cancer treatment guidelines:
9.4.1 Endocrine Therapy Eligibility
- ER/PR Positive: ER > 1% OR PR > 1% → Endocrine therapy recommended
- ER/PR Negative: ER = 0% AND PR = 0% → No endocrine therapy
9.4.2 HER2-Targeted Therapy Eligibility
- HER2 Positive (Score 3+): Anti-HER2 therapy (Trastuzumab, Pertuzumab)
- HER2 Equivocal (Score 2+): Requires FISH testing
- HER2 Negative (Score 0/1+): No anti-HER2 therapy
9.4.3 Chemotherapy Decision
- Luminal A: Often chemotherapy can be avoided (low Ki67, strong hormonal)
- Luminal B: Chemotherapy generally recommended (high Ki67)
- HER2 Positive: Chemotherapy + anti-HER2 therapy
- Triple Negative: Chemotherapy recommended
9.5 Treatment Category Changes
| Treatment Decision Changes by Pathologist | |||||||||
| Number and percentage of cases where AI changed treatment recommendation | |||||||||
| Pathologist | Total Cases | Endocrine Therapy Changes | HER2 Therapy Changes | Chemotherapy Changes | Any Treatment Change | % | % | % | % |
|---|---|---|---|---|---|---|---|---|---|
| Pathologist 1 | 259 | 3 | 10 | 52 | 55 | 1.2 | 3.9 | 20.1 | 21.2 |
| Pathologist 2 | 275 | 3 | 29 | 42 | 49 | 1.1 | 10.5 | 15.3 | 17.8 |
| Pathologist 3 | 252 | 2 | 15 | 42 | 46 | 0.8 | 6.0 | 16.7 | 18.3 |
| Pathologist 4 | 251 | 0 | 11 | 33 | 34 | 0.0 | 4.4 | 13.1 | 13.5 |
Note for Pathologist: This table shows, for each pathologist, how many cases had their treatment recommendation change after AI. The “Any Treatment Change” column is the most important: it tells you the overall rate at which AI influences clinical decisions. Variations between pathologists reflect differing AI adoption patterns.
9.6 Endocrine Therapy Changes
| Endocrine Therapy Eligibility Transitions | |||||
| Changes in treatment recommendation after AI | |||||
| pathologist | endocrine_eligible_pre | endocrine_eligible_post | Not Eligible → Not Eligible | Not Eligible → Eligible | Eligible → Eligible |
|---|---|---|---|---|---|
| Pathologist 1 | FALSE | FALSE | 40 | 0 | 0 |
| Pathologist 1 | FALSE | TRUE | 0 | 3 | 0 |
| Pathologist 1 | TRUE | TRUE | 0 | 0 | 216 |
| Pathologist 2 | FALSE | FALSE | 43 | 0 | 0 |
| Pathologist 2 | FALSE | TRUE | 0 | 3 | 0 |
| Pathologist 2 | TRUE | TRUE | 0 | 0 | 229 |
| Pathologist 3 | FALSE | FALSE | 41 | 0 | 0 |
| Pathologist 3 | FALSE | TRUE | 0 | 2 | 0 |
| Pathologist 3 | TRUE | TRUE | 0 | 0 | 209 |
| Pathologist 4 | FALSE | FALSE | 45 | 0 | 0 |
| Pathologist 4 | TRUE | TRUE | 0 | 0 | 206 |
Cases with changed endocrine therapy eligibility:Total: 8 Gained eligibility: 8 Lost eligibility: 0 | Examples of Endocrine Therapy Eligibility Changes | |||||||||
| case_id | pathologist | er_pre | er_post | pr_pre | pr_post | endocrine_eligible_pre | endocrine_eligible_post | comment | change_type |
|---|---|---|---|---|---|---|---|---|---|
| 23175-25 | Pathologist 2 | 0.0 | 2.0 | 0.0 | 0.0 | FALSE | TRUE | kesit kalitesi dusuk | Gained Eligibility |
| 21524-25 | Pathologist 2 | 0.0 | 1.0 | 0.0 | 0.0 | FALSE | TRUE | stromal hucrelerin ERsini invaziv hesapliyor | Gained Eligibility |
| 18689-25 | Pathologist 2 | 0.0 | 2.0 | 0.0 | 0.0 | FALSE | TRUE | — | Gained Eligibility |
| 19713-25 | Pathologist 1 | 0.0 | 1.0 | 0.0 | 0.0 | FALSE | TRUE | — | Gained Eligibility |
| 18689-25 | Pathologist 1 | 0.0 | 1.0 | 0.0 | 0.0 | FALSE | TRUE | — | Gained Eligibility |
| 11985-25 | Pathologist 1 | 0.0 | 2.0 | 0.0 | 0.0 | FALSE | TRUE | — | Gained Eligibility |
| 18961-25 | Pathologist 3 | 0.0 | 1.0 | 0.0 | 0.0 | FALSE | TRUE | — | Gained Eligibility |
| 14324-25 | Pathologist 3 | 0.0 | 2.0 | 0.0 | 3.0 | FALSE | TRUE | — | Gained Eligibility |
9.7 HER2 Therapy Changes
| HER2 Therapy Decision Transitions | ||
| How AI changed HER2-targeted therapy recommendations | ||
| Pre-AI Decision | Post-AI Decision | Number of Cases |
|---|---|---|
| Anti-HER2 Therapy | FISH Required | 2 |
| FISH Required | Anti-HER2 Therapy | 3 |
| FISH Required | No Anti-HER2 | 44 |
| No Anti-HER2 | FISH Required | 16 |
Note for Pathologist: The Sankey (flow) diagram shows how HER2 therapy decisions moved from Pre-AI to Post-AI. Follow the colored bands: thick bands flowing between different categories represent cases where the treatment recommendation changed. Movements from “No Anti-HER2” to “FISH Required” or “Anti-HER2 Therapy” are especially significant as they imply additional testing or treatment.
9.8 Chemotherapy Recommendation Changes
| Chemotherapy Recommendation Transitions | ||||
| Rows = Pre-AI, Columns = Post-AI | ||||
| Pre-AI Recommendation | Case-by-Case | Consider Avoiding | Recommended | Strongly Recommended |
|---|---|---|---|---|
| Case-by-Case | 120 | 9 | 20 | 4 |
| Consider Avoiding | 5 | 243 | 74 | 7 |
| Recommended | 0 | 3 | 116 | 5 |
| Strongly Recommended | 9 | 11 | 22 | 389 |
Significant chemotherapy recommendation changes: 128 | Significant Chemotherapy Recommendation Changes | |||||||||
| Cases where treatment intensity changed substantially | |||||||||
| case_id | pathologist | molecular_subtype_pre | molecular_subtype_post | chemo_recommendation_pre | chemo_recommendation_post | er_pre | er_post | ki67_pre | ki67_post |
|---|---|---|---|---|---|---|---|---|---|
| 34682-25 | Pathologist 2 | Hormone Weak Positive | Luminal B | Case-by-Case | Recommended | 100.0 | 95.0 | 13.0 | 30.0 |
| 33681-25 | Pathologist 2 | Luminal B | Luminal A | Recommended | Consider Avoiding | 50.0 | 50.0 | 35.0 | 25.0 |
| 33533-25 | Pathologist 2 | Hormone Weak Positive | HER2 Positive | Case-by-Case | Strongly Recommended | 95.0 | 97.0 | 15.0 | 22.0 |
| 32252-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 100.0 | 100.0 | 25.0 | 33.0 |
| 31276-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 100.0 | 100.0 | 29.0 | 30.0 |
| 31467-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 95.0 | 95.0 | 20.0 | 30.0 |
| 30689-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 95.0 | 100.0 | 23.0 | 35.0 |
| 24311-25 | Pathologist 2 | HER2 Positive | Luminal A | Strongly Recommended | Consider Avoiding | 98.0 | 97.0 | 3.0 | 6.0 |
| 24570-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 90.0 | 89.0 | 23.0 | 32.0 |
| 23916-25 | Pathologist 2 | HER2 Positive | Luminal A | Strongly Recommended | Consider Avoiding | 100.0 | 99.0 | 8.0 | 10.0 |
| 21963-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 95.0 | 95.0 | 27.0 | 37.0 |
| 21524-25 | Pathologist 2 | Triple Negative | Hormone Weak Positive | Strongly Recommended | Case-by-Case | 0.0 | 1.0 | 35.0 | 48.0 |
| 20823-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 80.0 | 77.0 | 21.0 | 30.0 |
| 19258-25 | Pathologist 2 | Luminal A | HER2 Positive | Consider Avoiding | Strongly Recommended | 100.0 | 95.0 | 22.0 | 27.0 |
| 18676-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 90.0 | 93.0 | 27.0 | 40.0 |
| 18012-25 | Pathologist 2 | HER2 Positive | Hormone Weak Positive | Strongly Recommended | Case-by-Case | 2.0 | 1.0 | 96.0 | 97.0 |
| 15965-25 | Pathologist 2 | HER2 Positive | Hormone Weak Positive | Strongly Recommended | Case-by-Case | 100.0 | 95.0 | 7.0 | 10.0 |
| 15305-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 90.0 | 86.0 | 28.0 | 44.0 |
| 14815-25 | Pathologist 2 | Luminal A | Luminal B | Consider Avoiding | Recommended | 100.0 | 95.0 | 28.0 | 30.0 |
| 14987-25 | Pathologist 2 | HER2 Positive | Luminal A | Strongly Recommended | Consider Avoiding | 80.0 | 90.0 | 26.0 | 28.0 |
9.9 Molecular Subtype Reclassification
Total molecular subtype reclassifications: 173 Percentage of all evaluations: 16.7 %
| Most Common Molecular Subtype Transitions | ||
| Top 15 reclassification patterns | ||
| Pre-AI Subtype | Post-AI Subtype | Number of Cases |
|---|---|---|
| Luminal A | Luminal B | 74 |
| HER2 Positive | Luminal B | 22 |
| Hormone Weak Positive | Luminal B | 20 |
| HER2 Positive | Luminal A | 11 |
| Hormone Weak Positive | Luminal A | 9 |
| HER2 Positive | Hormone Weak Positive | 7 |
| Luminal A | HER2 Positive | 7 |
| Luminal A | Hormone Weak Positive | 5 |
| Luminal B | HER2 Positive | 5 |
| HER2 Positive | Triple Negative | 4 |
| Hormone Weak Positive | HER2 Positive | 4 |
| Luminal B | Luminal A | 3 |
| Triple Negative | Hormone Weak Positive | 2 |
9.10 Impact on FISH Testing
| Impact on FISH Testing Requirements | ||||||
| Cases requiring HER2 FISH (Score 2+) | ||||||
| Pathologist | FISH Needed (Pre) | FISH Needed (Post) | Total Cases | Change | % Pre | % Post |
|---|---|---|---|---|---|---|
| Pathologist 1 | 42 | 38 | 259 | -4 | 16.2 | 14.7 |
| Pathologist 2 | 47 | 34 | 275 | -13 | 17.1 | 12.4 |
| Pathologist 3 | 46 | 37 | 252 | -9 | 18.3 | 14.7 |
| Pathologist 4 | 49 | 46 | 251 | -3 | 19.5 | 18.3 |
Overall change in FISH testing needs: -29 Interpretation: AI reduced FISH testing by 29 cases 9.11 Cost-Effectiveness Implications
| Treatment Decision Changes | |||||
| Potential clinical impact per pathologist | |||||
| Pathologist | FISH Cost Change | Increased Chemo | Decreased Chemo | Gained HER2 Tx | Lost HER2 Tx |
|---|---|---|---|---|---|
| Pathologist 1 | −$2,000.00 | 37 | 4 | 0 | 0 |
| Pathologist 2 | −$6,500.00 | 21 | 11 | 1 | 1 |
| Pathologist 3 | −$4,500.00 | 24 | 6 | 1 | 1 |
| Pathologist 4 | −$1,500.00 | 23 | 2 | 1 | 0 |
Overall Clinical Impact Summary:Total cases with changed chemo recommendation: 128 - More intensive: 105 - Less intensive: 23 Total cases with changed HER2 therapy: 5 - Gained therapy: 3 - Lost therapy: 2 9.12 Concordance with Consensus
For cases where we can define a “majority consensus” (3+ pathologists agree).
| Impact on Consensus Achievement | ||
| Molecular subtype consensus (3+ pathologists agree) | ||
| Outcome | Cases | % |
|---|---|---|
| Achieved Consensus | 22 | 8.2 |
| Lost Consensus | 19 | 7.1 |
| Maintained Consensus | 214 | 79.6 |
| No Consensus Both | 14 | 5.2 |
9.13 Conclusion
9.13.1 Clinical Implications
Treatment Decision Changes: AI substantially affects treatment recommendations in a meaningful proportion of cases.
Endocrine Therapy: Changes in ER/PR interpretation can determine whether patients receive hormonal therapy.
HER2 Therapy: Critical for anti-HER2 therapy eligibility and FISH testing needs.
Chemotherapy Decisions: Molecular subtype changes affect chemotherapy recommendations, particularly for borderline cases.
FISH Testing: AI may reduce or increase FISH testing needs depending on HER2 scoring patterns.
Consensus Building: AI’s impact on achieving diagnostic consensus varies by case.
9.13.2 Recommendations
- Cases with treatment-altering changes should be reviewed in multidisciplinary tumor boards
- Quality assurance processes should focus on cases where AI caused clinically significant changes
- Consider AI as a “tie-breaker” in cases lacking consensus
- Monitor real-world outcomes for AI-influenced treatment decisions