26 HER2 Low-Level Interpretation Analysis – AI for Breast Cancer Analysis

27.1 Background & Clinical Context

The distinction between HER2 Score 0 and Score 1+ has become one of the most clinically consequential calls in breast cancer pathology. With the approval of trastuzumab deruxtecan (T-DXd, Enhertu) for HER2-low breast cancer, this historically “negative-versus-negative” distinction now determines access to targeted therapy.

This chapter provides a focused deep-dive into HER2 interpretation at the low end of the scoring spectrum, complementing the broader HER2-low overview in Chapter 14.

27.1.1 The Evolving HER2 Classification

Category	IHC Pattern	Clinical Relevance
Score 0	No staining or incomplete faint membrane staining in <=10% of cells	HER2-negative; T-DXd not eligible
HER2-ultralow	Faint incomplete staining — between 0 and 1+ (emerging concept)	Under investigation for T-DXd benefit
Score 1+	Faint/barely perceptible complete membrane staining in >10% of cells	HER2-low; T-DXd eligible
Score 2+/FISH-	Weak-moderate complete membrane staining in >10%, FISH negative	HER2-low; T-DXd eligible
Score 2+/FISH+	Weak-moderate staining, FISH positive	HER2-positive
Score 3+	Strong complete membrane staining in >10% of cells	HER2-positive

27.1.2 Why the 0 vs 1+ Distinction Is So Difficult

Subjective threshold: The difference between “no staining” and “faint, barely perceptible” staining is inherently subjective
Pre-analytical variability: Fixation time, antibody clone, and staining protocol all affect faint staining detection
Low reproducibility: Published kappa values for 0 vs 1+ range from 0.40-0.60 (moderate agreement)
High clinical stakes: T-DXd costs approximately $15,000/month; misclassification affects treatment access

Note for Pathologist: This analysis specifically investigates how AI affects the 0 vs 1+ call — the most subjective distinction in HER2 scoring. We examine whether AI pushes pathologists toward or away from Score 1+, whether some pathologists are more affected than others, and whether specimen type matters.

27.2 Load and Prepare Data

HER2 observations: 2368

Cases: 296

Pathologists: 4

FISH data loaded: 33 cases with results
  FISH Negative: 29 
  FISH Positive: 4

27.3 HER2 Score Distribution at the Low End

27.3.1 Overall Score Distribution: Focus on 0 and 1

Phase	HER2 Score¹	N	Total	%
HER2 Score Distribution: Pre-AI vs Post-AI
All 4 scores shown; Score 0 and 1 are the focus
Post-AI	0	317	1102	28.8
Post-AI	1	493	1102	44.7
Post-AI	2	160	1102	14.5
Post-AI	3	132	1102	12.0
Pre-AI	0	345	1107	31.2
Pre-AI	1	438	1107	39.6
Pre-AI	2	192	1107	17.3
Pre-AI	3	132	1107	11.9
¹ Scores 0 and 1 (highlighted) are the focus of this analysis

27.3.2 Score 0 vs Score 1 Prevalence Shift

Score	N (Post)	N (Pre)	% (Post)	% (Pre)
Score 0 vs Score 1 Prevalence: Pre-AI vs Post-AI
Among cases scored 0 or 1 only
0	317	345	39.1	44.1
1	493	438	60.9	55.9

27.3.3 Visualization

27.4 Score 0 vs Score 1 Transition Analysis

27.4.1 Paired Transition Matrix (Pre-AI to Post-AI)

Pre-AI Score	Post-AI Score
Full HER2 Score Transition Matrix¹
Pre-AI (rows) to Post-AI (columns)¹
Pre-AI Score	0	1	2	3
0	302	24	0	0
1	10	402	16	0
2	0	44	141	3
3	0	0	2	129
¹ Yellow cells = transitions within the 0/1 zone (focus of this analysis)

27.4.2 Score 0 <-> Score 1 Transitions Only

Transition	N	%
HER2 Score Transitions Involving Score 0 or 1
T-DXd eligibility changes highlighted
1 -> 1 (Stable HER2-low)	402	50.4
0 -> 0 (Stable negative)	302	37.8
2+/3+ -> 1 (Downgrade to low)	44	5.5
0 -> 1 (Gained T-DXd eligibility)	24	3.0
1 -> 2+/3+ (Upgrade from low)	16	2.0
1 -> 0 (Lost T-DXd eligibility)	10	1.3

27.4.3 Clinical Impact Summary

T-DXd ELIGIBILITY IMPACT (Score 0 <-> Score 1):
================================================

Total paired assessments: 1073
Assessments involving Score 0 or 1: 798

STABLE:
- Score 0 -> 0 (stable negative): 302
- Score 1 -> 1 (stable HER2-low): 402

CHANGED:
- Score 0 -> 1 (GAINED T-DXd eligibility): 24 (2.2% of all paired)
- Score 1 -> 0 (LOST T-DXd eligibility): 10 (0.9% of all paired)
- Net change: 14 cases

CONCORDANCE at 0/1 boundary:
- Among cases scored 0 or 1 in either phase: 95.4% stable

27.5 HER2-Low Classification (0 vs Low vs Positive)

27.5.1 Define HER2-Low Category

HER2-low includes Score 1+ and Score 2+/FISH-. Since FISH data is not available for Score 2+ cases in this dataset, we classify conservatively:

Score 0: HER2-negative
Score 1: HER2-low (definite)
Score 2: HER2-low (presumed, pending FISH) or HER2-positive (if FISH+)
Score 3: HER2-positive

Pre-AI	Post-AI
HER2 Classification Transition Matrix
Score 0 (Negative) vs Score 1 (HER2-low) vs Score 2+/3+ (Higher)
Pre-AI	Score 0 (Negative)	Score 1 (HER2-low)	Score 2+/3+ (Higher)
Score 0 (Negative)	302	24	0
Score 1 (HER2-low)	10	402	16
Score 2+/3+ (Higher)	0	44	275

27.5.2 Row Percentages

Pre-AI	Post-AI
HER2 Classification Transitions (Row Percentages)
What happened to cases in each Pre-AI category?
Pre-AI	Score 0 (Negative)	Score 1 (HER2-low)	Score 2+/3+ (Higher)
Score 0 (Negative)	302 (92.6%)	24 (7.4%)	0 (0.0%)
Score 1 (HER2-low)	10 (2.3%)	402 (93.9%)	16 (3.7%)
Score 2+/3+ (Higher)	0 (0.0%)	44 (13.8%)	275 (86.2%)

27.6 Per-Pathologist Analysis at the 0/1 Boundary

27.6.1 Per-Pathologist Score 0 vs Score 1 Transitions

Pathologist	N Cases	Score 0 (Pre)	Score 1 (Pre)	Score 0 (Post)	Score 1 (Post)	N Changed	Change %	0->1	1->0	Net¹
Per-Pathologist HER2 Score 0 vs 1 Transitions
Among cases scored 0 or 1 in both phases
Pathologist 1	190	81	109	71	119	12	6.3	11	1	10
Pathologist 2	202	65	137	67	135	16	7.9	7	9	-2
Pathologist 3	177	124	53	118	59	6	3.4	6	0	6
Pathologist 4	169	56	113	56	113	0	0.0	0	0	0
¹ Positive net = pathologist tends to upgrade 0->1 with AI; negative = downgrade 1->0

27.6.2 Per-Pathologist Score Shift Visualization

27.6.3 Per-Pathologist Alluvial: Where Do Score 0 Cases Go?

Pathologist	Post-AI Score
Where Do Pre-AI Score 0 Cases End Up Post-AI?
Per-pathologist destination of Score 0 cases
Pathologist	0	1
Pathologist 1	70 (86.4%)	11 (13.6%)
Pathologist 2	58 (89.2%)	7 (10.8%)
Pathologist 3	118 (95.2%)	6 (4.8%)
Pathologist 4	56 (100%)	0 (0.0%)

27.6.4 Where Do Score 1 Cases Go?

Pathologist	Post-AI Score
Where Do Pre-AI Score 1 Cases End Up Post-AI?
Per-pathologist destination of Score 1 cases
Pathologist	0	1	2
Pathologist 1	1 (0.9%)	108 (96.4%)	3 (2.7%)
Pathologist 2	9 (6.2%)	128 (88.9%)	7 (4.9%)
Pathologist 3	0 (0.0%)	53 (96.4%)	2 (3.6%)
Pathologist 4	0 (0.0%)	113 (96.6%)	4 (3.4%)

27.7 Interobserver Agreement at the 0/1 Boundary

27.7.1 Fleiss’ Kappa: Score 0 vs Score 1 (Binary)

Phase	Fleiss' Kappa	z	p-value	N Cases	Delta
Binary Agreement: Score 0 vs Score 1
Fleiss' Kappa — cases scored 0 or 1 by all pathologists
Pre-AI	0.550	15.251	0.00	128	NA
Post-AI	0.620	17.977	0.00	140	0.070

27.7.2 Fleiss’ Kappa: Full 4-Score Agreement

Scope	Kappa (Pre)	Kappa (Post)	N (Pre)	N (Post)	Delta
Agreement Comparison: Binary (0 vs 1) vs Full (0-3)
Is agreement worse at the 0/1 boundary than overall?
Binary (0 vs 1)	0.550	0.620	128	140	0.070
Full (0/1/2/3)	0.671	0.726	229	226	0.055

27.7.3 Pairwise Cohen’s Kappa: Score 0 vs Score 1

Pathologist Pair	Kappa (Pre)	Kappa (Post)	Delta
Pairwise Cohen's Kappa: Score 0 vs Score 1
Per pathologist pair
Pathologist 2 vs Pathologist 1	NA	NA	NA
Pathologist 2 vs Pathologist 4	NA	NA	NA
Pathologist 2 vs Pathologist 3	NA	NA	NA
Pathologist 1 vs Pathologist 4	NA	NA	NA
Pathologist 1 vs Pathologist 3	NA	NA	NA
Pathologist 4 vs Pathologist 3	NA	NA	NA

27.7.4 Overall Percent Agreement for Score 0 vs Score 1

Phase	Agreement on Score 0¹	Agreement on Score 1	Overall Agreement
Specific Agreement for Score 0 and Score 1
How well do pathologists agree on each specific score?
Pre-AI	79.4%	74.5%	77.6%
Post-AI	80.2%	81.8%	81.3%
¹ Specific agreement = 2*a/(n1+n2) where a=both agree, n1/n2=each rater's count

27.8 Cross-Pathologist Concordance for Individual Cases

27.8.1 How Often Do All Pathologists Agree on 0 vs 1?

Phase	N Cases	All Agree	% All Agree	Cases with 0/1	0/1 All Agree	% 0/1 Agree
Complete Concordance Among All Pathologists
Cases where all 4 pathologists assign the same HER2 score
Post-AI	226	144	63.7	170	90	52.9
Pre-AI	229	130	56.8	170	75	44.1

27.8.2 Disagreement Patterns Among Score 0/1 Cases

Pattern	Post-AI	Pre-AI
Disagreement Patterns Among Low-Score Cases
Cases with at least one Score 0 or 1 where pathologists disagree
0 vs 1 only	50	53
1 vs 2 only	30	40
2-step range (e.g., 0-2)	0	2

27.9 Confusion Matrix: Score 0 vs Score 1 (Pre-AI vs Post-AI)

27.9.1 Aggregate Confusion Matrix

Reference (Pre-AI)	Post-AI
Confusion Matrix: Score 0 vs Score 1
Pre-AI as reference, Post-AI as prediction
Reference (Pre-AI)	0	1
0	302	24
1	10	402

27.9.2 Precision, Recall, and F1

Category	TP	FP	FN	Precision	Recall	F1	Accuracy
Precision, Recall, F1: Score 0 vs Score 1
Pre-AI as reference, Post-AI as prediction
0	302	10	24	0.968	0.926	0.947	0.954
1	402	24	10	0.944	0.976	0.959	0.954

27.9.3 Per-Pathologist P/R/F1

Pathologist	Category	Precision	Recall	F1	Accuracy
Per-Pathologist P/R/F1: Score 0 vs Score 1
Pre-AI as reference, Post-AI as prediction
Pathologist 1	0	0.986	0.864	0.921	0.937
Pathologist 1	1	0.908	0.991	0.947	0.937
Pathologist 2	0	0.866	0.892	0.879	0.921
Pathologist 2	1	0.948	0.934	0.941	0.921
Pathologist 3	0	1.000	0.952	0.975	0.966
Pathologist 3	1	0.898	1.000	0.946	0.966
Pathologist 4	0	1.000	1.000	1.000	1.000
Pathologist 4	1	1.000	1.000	1.000	1.000

27.9.4 Confusion Heatmap: Full 4-Score

27.10 Biopsy Type Stratification

27.10.1 Score 0 <-> Score 1 Transitions by Specimen Type

Biopsy Type	N Cases	0 -> 1	1 -> 0	N Changed	Change %	Net
Score 0 <-> Score 1 Transitions by Biopsy Type
Does specimen type affect the 0/1 distinction?
Excision	428	14	9	23	5.4	5
Tru-cut	310	10	1	11	3.5	9

27.10.2 Visualization by Biopsy Type

27.11 Impact on Molecular Subtype and Treatment

27.11.1 Subtype Changes Driven by HER2 0 <-> 1 Transitions

HER2 Transition	Molecular Subtype (Pre-AI)	N Cases	%
Molecular Subtype Context of HER2 0 <-> 1 Transitions¹
Which subtypes are affected by the 0/1 reclassification?¹
HER2 0 -> 1	Hormone Weak Positive	5	20.8
HER2 0 -> 1	Luminal A	14	58.3
HER2 0 -> 1	Luminal B	3	12.5
HER2 0 -> 1	Triple Negative	2	8.3
HER2 1 -> 0	Hormone Weak Positive	2	20.0
HER2 1 -> 0	Luminal A	4	40.0
HER2 1 -> 0	Luminal B	3	30.0
HER2 1 -> 0	Triple Negative	1	10.0
¹ HER2 0->1 and 1->0 do not change molecular subtype (both are HER2-negative for subtyping) but DO change T-DXd eligibility

27.11.2 T-DXd Eligibility Impact by Subtype

Molecular Subtype	Gained T-DXd eligibility	Lost T-DXd eligibility
T-DXd Eligibility Changes by Molecular Subtype
Clinical impact of Score 0 <-> 1 reclassification
Hormone Weak Positive	5	2
Luminal A	14	4
Luminal B	3	3
Triple Negative	2	1

27.12 Concordance Rate Over Score Range

27.12.1 Concordance by HER2 Score

Pre-AI Score	N Cases	N Concordant	Concordance %
Pre-AI to Post-AI Concordance Rate by HER2 Score
Is Score 0 or Score 1 less stable than higher scores?
0	326	302	92.6
1	428	402	93.9
2	188	141	75.0
3	131	129	98.5

27.12.2 Visualization

27.13 Case-Level Exploration: Discordant 0/1 Cases

27.13.1 List of Discordant Cases

Case ID	Pathologist	Pre-AI	Post-AI	Direction	Biopsy Type
All Discordant HER2 Score 0 <-> 1 Cases
Individual cases where AI changed the 0/1 call
12343-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
16396-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
16497-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
19362-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
20235-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
22026-25	Pathologist 2	1	0	Downgrade (1->0)	Tru-cut
33089-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
35091-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
7468-25	Pathologist 2	1	0	Downgrade (1->0)	Excision
7468-25	Pathologist 1	1	0	Downgrade (1->0)	Excision
10183-25	Pathologist 3	0	1	Upgrade (0->1)	Excision
11632-25	Pathologist 2	0	1	Upgrade (0->1)	Excision
12343-25	Pathologist 1	0	1	Upgrade (0->1)	Excision
12440-25	Pathologist 2	0	1	Upgrade (0->1)	Tru-cut
13471-25	Pathologist 1	0	1	Upgrade (0->1)	Tru-cut
14933-25	Pathologist 1	0	1	Upgrade (0->1)	Excision
23259-25	Pathologist 2	0	1	Upgrade (0->1)	Tru-cut
27588-25	Pathologist 1	0	1	Upgrade (0->1)	Tru-cut
28310-25	Pathologist 1	0	1	Upgrade (0->1)	Excision
29841-25	Pathologist 1	0	1	Upgrade (0->1)	Excision
30017-25	Pathologist 3	0	1	Upgrade (0->1)	Excision
30029-25	Pathologist 3	0	1	Upgrade (0->1)	Excision
30130-25	Pathologist 2	0	1	Upgrade (0->1)	Tru-cut
30736-25	Pathologist 1	0	1	Upgrade (0->1)	Excision
30767-25	Pathologist 1	0	1	Upgrade (0->1)	Excision
31448-25	Pathologist 3	0	1	Upgrade (0->1)	Excision
31485-25	Pathologist 3	0	1	Upgrade (0->1)	Excision
33813-25	Pathologist 2	0	1	Upgrade (0->1)	Tru-cut
34681-25	Pathologist 1	0	1	Upgrade (0->1)	Tru-cut
6444-25	Pathologist 2	0	1	Upgrade (0->1)	Tru-cut
6444-25	Pathologist 1	0	1	Upgrade (0->1)	Tru-cut
6444-25	Pathologist 3	0	1	Upgrade (0->1)	Tru-cut
7010-25	Pathologist 2	0	1	Upgrade (0->1)	Excision
7010-25	Pathologist 1	0	1	Upgrade (0->1)	Excision

27.13.2 Cases with Pathologist Disagreement on 0 vs 1

INTER-PATHOLOGIST DISAGREEMENT on Score 0 vs 1:
- Pre-AI: 55 cases where some pathologists score 0 and others score 1
- Post-AI: 50 cases
- Change: -5 cases

Case ID	Pathologist 2	Pathologist 1	Pathologist 4	Pathologist 3	All 0/1?
Pre-AI: Cases Where Pathologists Disagree on 0 vs 1
Each column shows a pathologist's HER2 score
34604-25	1	0	1	0	TRUE
34186-25	1	1	1	0	TRUE
34286-25	0	0	1	0	TRUE
33988-25	1	1	1	0	TRUE
33813-25	0	0	1	0	TRUE
33015-25	1	0	0	0	TRUE
33089-25	1	0	0	0	TRUE
31448-25	1	1	1	0	TRUE
31485-25	1	0	0	0	TRUE
31276-25	1	0	1	0	TRUE
31467-25	1	1	1	0	TRUE
30736-25	1	0	1	0	TRUE
30767-25	1	0	1	1	TRUE
30689-25	1	1	1	0	TRUE
30029-25	1	1	2	0	FALSE
30017-25	1	1	1	0	TRUE
27902-25	0	0	1	0	TRUE
28310-25	1	0	1	0	TRUE
25784-25	1	0	1	0	TRUE
25166-25	1	1	1	0	TRUE
24569-25	1	0	1	0	TRUE
24009-25	1	1	1	0	TRUE
24185-25	1	1	1	0	TRUE
23100-25	1	1	1	0	TRUE
21963-25	1	0	0	0	TRUE
21951-25	1	0	1	0	TRUE
22026-25	1	0	0	0	TRUE
20823-25	1	1	2	0	FALSE
18131-25	1	1	1	0	TRUE
18004-25	1	1	1	0	TRUE
18341-25	1	1	1	0	TRUE
17716-25	1	1	1	0	TRUE
17071-25	1	1	0	0	TRUE
17077-25	1	1	1	0	TRUE
16396-25	1	0	0	0	TRUE
16497-25	1	1	1	0	TRUE
15319-25	0	1	0	0	TRUE
14698-25	1	1	0	0	TRUE
14902-25	1	1	0	0	TRUE
14815-25	1	1	1	0	TRUE
14933-25	1	0	1	1	TRUE
14262-25	1	1	0	0	TRUE
13582-25	1	1	1	0	TRUE
13472-25	1	1	1	0	TRUE
12343-25	1	0	0	0	TRUE
12545-25	1	0	0	0	TRUE
12030-25	1	1	1	0	TRUE
12440-25	0	0	1	1	TRUE
11632-25	0	1	1	1	TRUE
11286-25	1	1	1	0	TRUE
10183-25	1	1	1	0	TRUE
8132-25	1	1	1	0	TRUE
7468-25	1	1	0	0	TRUE
7012-25	0	0	0	1	TRUE
6592-25	1	1	1	0	TRUE

Case ID	Pathologist 2	Pathologist 1	Pathologist 4	Pathologist 3	All 0/1?
Post-AI: Cases Where Pathologists Disagree on 0 vs 1
Each column shows a pathologist's HER2 score
35131-25	1	0	1	0	TRUE
34681-25	1	1	1	0	TRUE
34604-25	1	0	1	0	TRUE
34186-25	1	1	1	0	TRUE
34286-25	0	0	1	0	TRUE
33988-25	1	1	1	0	TRUE
33813-25	1	0	1	0	TRUE
33015-25	1	0	0	0	TRUE
32396-25	1	0	1	0	TRUE
32195-25	1	1	1	0	TRUE
31276-25	1	0	1	0	TRUE
30736-25	1	1	1	0	TRUE
30689-25	1	1	1	0	TRUE
30130-25	1	0	0	0	TRUE
28718-25	1	1	1	0	TRUE
28310-25	1	1	1	0	TRUE
27588-25	1	1	1	0	TRUE
27247-25	1	1	1	0	TRUE
25533-25	1	1	1	0	TRUE
25166-25	1	1	1	0	TRUE
24009-25	1	1	1	0	TRUE
24433-25	1	1	1	0	TRUE
23100-25	1	1	1	0	TRUE
21963-25	1	0	0	0	TRUE
21951-25	1	0	1	0	TRUE
21114-25	1	1	1	0	TRUE
21524-25	1	1	1	0	TRUE
20823-25	1	1	1	0	TRUE
18131-25	1	1	1	0	TRUE
18300-25	1	1	1	0	TRUE
18413-25	1	1	1	0	TRUE
18004-25	1	1	1	0	TRUE
18341-25	1	1	1	0	TRUE
17716-25	1	1	1	0	TRUE
17077-25	1	1	1	0	TRUE
16497-25	0	1	1	0	TRUE
15319-25	0	1	0	0	TRUE
14815-25	1	1	1	0	TRUE
13582-25	1	1	1	0	TRUE
13472-25	1	1	1	0	TRUE
13471-25	1	1	1	0	TRUE
12343-25	0	1	0	0	TRUE
12545-25	1	0	0	0	TRUE
12030-25	1	1	1	0	TRUE
11352-25	1	1	1	0	TRUE
11286-25	1	1	1	0	TRUE
7010-25	1	1	1	0	TRUE
7012-25	0	0	0	1	TRUE
6592-25	1	1	1	0	TRUE
6444-25	1	1	0	1	TRUE

27.14 FISH-Integrated HER2 Classification

Data Source: FISH Results

The tables and figures in Sections 12-14 incorporate FISH (Fluorescence In Situ Hybridization) results from the laboratory database. FISH testing was performed on a subset of cases. These results allow definitive classification of Score 2+ cases as HER2-low (FISH-) or HER2-positive (FISH+).

27.14.1 FISH Data Overview

Classified As	Original FISH Value	N Cases
FISH Results Available
Cases with FISH testing from laboratory records
Negative	NEGATİF	26
Negative	NEGATİF (KONS MATERYALİ)	1
Negative	YAPILMAMIŞ (KONSBLOĞUNDA YAPILMIŞ NRGATİF)	1
Negative	YAPILMAMIŞ (MASTEKTOMİDE YAPILMIŞ NEGATİF)	1
Positive	POZİTİF	4
Source: aiforia breast - her2 - fish.xlsx

27.14.2 FISH Results Joined with HER2 IHC Scores

Cases with both HER2 IHC and FISH: 31 
Pathologist-case pairs: 119

FISH Result	HER2 Pre-AI	HER2 Post-AI	N
HER2 IHC Scores Among FISH-Tested Cases
Pre-AI and Post-AI scores by FISH result
Negative	0	0	8
Negative	0	1	2
Negative	1	1	34
Negative	2	1	14
Negative	2	2	42
Negative	2	3	1
Negative	3	2	1
Negative	3	3	1
Positive	2	2	8
Positive	2	3	2
Positive	3	3	6
Source: FISH results from laboratory database joined with HER2 IHC scores

27.14.3 Definitive HER2 Classification Using FISH

Phase	HER2 Category	N	%
Definitive HER2 Classification (IHC + FISH)
Score 2+ cases resolved using FISH results
Post-AI	HER2-Negative (Score 0)	8	6.7
Post-AI	HER2-Low (Score 1+)	50	42.0
Post-AI	HER2-Low (Score 2+/FISH-)	43	36.1
Post-AI	HER2-Positive (Score 2+/FISH+)	8	6.7
Post-AI	HER2-Positive (Score 3+)	10	8.4
Pre-AI	HER2-Negative (Score 0)	10	8.4
Pre-AI	HER2-Low (Score 1+)	34	28.6
Pre-AI	HER2-Low (Score 2+/FISH-)	57	47.9
Pre-AI	HER2-Positive (Score 2+/FISH+)	10	8.4
Pre-AI	HER2-Positive (Score 3+)	8	6.7
Source: HER2 IHC scores + FISH results from laboratory database

27.15 FISH-Based Transition Analysis

Data Source: FISH Results

This section uses FISH results to provide definitive HER2 classification for Score 2+ cases.

27.15.1 Definitive HER2 Transition Matrix (with FISH)

Pre-AI	Post-AI
Definitive HER2 Transition Matrix (IHC + FISH)
Pre-AI (rows) to Post-AI (columns) — Score 2+ resolved by FISH
Pre-AI	HER2-Negative (Score 0)	HER2-Low (Score 1+)	HER2-Low (Score 2+/FISH-)	HER2-Positive (Score 3+)	HER2-Positive (Score 2+/FISH+)
HER2-Negative (Score 0)	8	2	0	0	0
HER2-Low (Score 1+)	0	34	0	0	0
HER2-Low (Score 2+/FISH-)	0	14	42	1	0
HER2-Positive (Score 2+/FISH+)	0	0	0	2	8
HER2-Positive (Score 3+)	0	0	1	7	0
Source: HER2 IHC scores + FISH results from laboratory database

27.15.2 Clinically Relevant Transitions (with FISH)

Pre-AI	Post-AI
3-Tier HER2 Transitions: Negative / Low / Positive (with FISH)
HER2-Low = Score 1+ or Score 2+/FISH-; HER2-Positive = Score 2+/FISH+ or 3+
Pre-AI	HER2-Negative	HER2-Low	HER2-Positive
HER2-Negative	8 (80%)	2 (20%)	0 (0.0%)
HER2-Low	0 (0.0%)	90 (98.9%)	1 (1.1%)
HER2-Positive	0 (0.0%)	1 (5.6%)	17 (94.4%)
Source: HER2 IHC scores + FISH results from laboratory database

27.15.3 T-DXd Eligibility with Definitive Classification

Change	N	%
T-DXd Eligibility Changes (Definitive Classification with FISH)
HER2-Low (Score 1+ or Score 2+/FISH-) = T-DXd eligible
No change	115	96.6
Gained T-DXd eligibility	3	2.5
Lost T-DXd eligibility	1	0.8
Source: HER2 IHC scores + FISH results from laboratory database

27.15.4 Score 2+ Cases: FISH-Resolved Classification

Case ID	Pathologist	IHC Pre	IHC Post	FISH	Definitive Pre	Definitive Post	Biopsy
Score 2+ Cases: Definitive Classification by FISH
Cases with HER2 IHC Score 2 in Pre-AI or Post-AI
10074-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10074-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10074-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10074-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10157-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Tru-cut
10157-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10157-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10157-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
10676-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
10676-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
10676-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
12675-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
12675-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
12675-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
12675-25	Pathologist 3	2	3	Negative	HER2-Low (Score 2+/FISH-)	HER2-Positive (Score 3+)	Excision
12818-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
12818-25	Pathologist 1	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Tru-cut
12818-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
12818-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
14058-25	Pathologist 1	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Tru-cut
14058-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
17896-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
17896-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
17896-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
17896-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
18012-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
18012-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
24311-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
24311-25	Pathologist 4	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
25564-25	Pathologist 1	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Tru-cut
28926-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
28926-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
33681-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
33681-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6225-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
6476-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Tru-cut
6922-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6922-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6922-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6937-25	Pathologist 2	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6937-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6937-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
6937-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
7291-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
7291-25	Pathologist 1	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
7291-25	Pathologist 4	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
7291-25	Pathologist 3	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
8754-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
9347-25	Pathologist 2	2	1	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 1+)	Excision
9347-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
9347-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
9347-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
9885-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
9885-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Excision
9930-25	Pathologist 2	3	2	Negative	HER2-Positive (Score 3+)	HER2-Low (Score 2+/FISH-)	Tru-cut
9930-25	Pathologist 1	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
9930-25	Pathologist 4	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
9930-25	Pathologist 3	2	2	Negative	HER2-Low (Score 2+/FISH-)	HER2-Low (Score 2+/FISH-)	Tru-cut
15428-25	Pathologist 2	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Excision
15428-25	Pathologist 1	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Excision
15428-25	Pathologist 4	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Excision
15428-25	Pathologist 3	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Excision
15441-25	Pathologist 2	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Tru-cut
15441-25	Pathologist 1	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Tru-cut
15441-25	Pathologist 4	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Tru-cut
15441-25	Pathologist 3	2	2	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 2+/FISH+)	Tru-cut
20256-25	Pathologist 2	2	3	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 3+)	Tru-cut
20256-25	Pathologist 4	2	3	Positive	HER2-Positive (Score 2+/FISH+)	HER2-Positive (Score 3+)	Tru-cut
Source: HER2 IHC scores + FISH results from laboratory database

27.16 FISH-Based Agreement Analysis

Data Source: FISH Results

This section calculates agreement metrics using the definitive HER2 classification that incorporates FISH results for Score 2+ cases.

27.16.1 Agreement: 3-Tier Classification (Negative / Low / Positive)

Fleiss' Kappa	z	p-value	N Cases	Phase	Delta
Agreement: 3-Tier HER2 Classification (with FISH)
Fleiss' Kappa — Negative / HER2-Low / HER2-Positive
0.861	13.737	0.00	26	Pre-AI	NA
0.904	14.245	0.00	26	Post-AI	0.043
Source: HER2 IHC scores + FISH results from laboratory database

27.16.2 Confusion Matrix: Definitive Classification (with FISH)

Reference (Pre-AI)	Post-AI
Confusion Matrix: Definitive HER2 Classification (with FISH)
3-Tier: HER2-Negative / HER2-Low / HER2-Positive
Reference (Pre-AI)	HER2-Negative	HER2-Low	HER2-Positive
HER2-Negative	8	2	0
HER2-Low	0	90	1
HER2-Positive	0	1	17
Source: HER2 IHC scores + FISH results from laboratory database

27.16.3 Precision, Recall, F1: Definitive Classification (with FISH)

Category	TP	FP	FN	Precision	Recall	F1	Accuracy
P/R/F1: Definitive HER2 Classification (with FISH)
Pre-AI as reference, Post-AI as prediction
HER2-Negative	8	0	2	1.000	0.800	0.889	0.966
HER2-Low	90	3	1	0.968	0.989	0.978	0.966
HER2-Positive	17	1	1	0.944	0.944	0.944	0.966
Source: HER2 IHC scores + FISH results from laboratory database

27.16.4 Comparison: IHC-Only vs IHC+FISH Classification

Classification Method¹	N Assessments	Concordance (%)
Concordance Comparison: IHC-Only vs IHC+FISH
Among cases with FISH results available
IHC Score Only (0/1/2/3)	119	83.2
Definitive (IHC + FISH)	119	96.6
¹ IHC-only uses raw 4-score (0/1/2/3); Definitive resolves Score 2 by FISH
Source: HER2 IHC scores + FISH results from laboratory database

27.16.5 FISH Summary

FISH DATA INTEGRATION SUMMARY
=============================

FISH-tested cases in dataset: 33
  - FISH Negative: 29
  - FISH Positive: 4

Cases matched to HER2 IHC data: 31
Score 2+ cases resolved by FISH: 68

CLINICAL SIGNIFICANCE:
- FISH resolves Score 2+ ambiguity: 2+/FISH- = HER2-Low (T-DXd eligible)
- FISH Positive cases (4): Definitive HER2-positive (trastuzumab eligible)
- FISH Negative Score 2+ cases are reclassified as HER2-Low
- Score 0 and Score 1 do NOT require FISH (classification unchanged)

27.17 Summary and Clinical Implications

SUMMARY - HER2 Low-Level Interpretation Analysis
=================================================

SCOPE:
- Total paired HER2 assessments: 1073
- Assessments involving Score 0 or 1: 798

SCORE 0 <-> 1 TRANSITIONS:
- Score 0 -> 1 (gained T-DXd eligibility): 24
- Score 1 -> 0 (lost T-DXd eligibility): 10
- Net: 14 cases

AGREEMENT (Fleiss' Kappa for 0 vs 1):
- Pre-AI: 0.55
- Post-AI: 0.62
- Delta: 0.07

INTER-PATHOLOGIST DISAGREEMENT:
- Pre-AI: 55 cases with 0/1 disagreement
- Post-AI: 50 cases with 0/1 disagreement

27.17.1 Clinical Recommendations

1. Score 0 vs Score 1: The Most Critical Call

The 0/1 distinction is the single most consequential call in HER2 scoring for T-DXd eligibility
AI assistance shows improvement in interobserver agreement
Any AI-suggested change from 0 to 1 or 1 to 0 should trigger manual pathologist review

2. Pathologist-Specific Patterns

Some pathologists may systematically favor Score 0 or Score 1 with AI assistance
Monitoring individual pathologist reclassification rates is recommended
Consensus scoring should be considered for borderline cases

3. Specimen Type Considerations

Pre-analytical factors (fixation, processing) disproportionately affect faint staining detection
Biopsy type may influence the reliability of the 0/1 distinction
Special attention needed for small specimens (tru-cut, vacuum biopsies)

4. HER2-Ultralow: Emerging Concept

Recent studies suggest cases between Score 0 and Score 1 (“ultralow”) may also benefit from T-DXd
Current IHC scoring does not capture this nuance (binary 0 vs 1)
Future AI tools may need to distinguish ultralow from true zero staining

5. Quality Assurance

Track 0 <-> 1 reclassification rates as a quality metric
Consider reflex re-staining or second opinion for discordant cases
T-DXd annual cost (~$180,000/patient) justifies careful assessment

Analysis completed: 2026-02-10
HER2 Score 0 vs 1 is the most subjective and clinically consequential distinction in current breast cancer pathology

27 HER2 Low-Level Interpretation Analysis

27.1 Background & Clinical Context

27.1.1 The Evolving HER2 Classification

27.1.2 Why the 0 vs 1+ Distinction Is So Difficult

27.2 Load and Prepare Data

27.3 HER2 Score Distribution at the Low End

27.3.1 Overall Score Distribution: Focus on 0 and 1

27.3.2 Score 0 vs Score 1 Prevalence Shift

27.3.3 Visualization

27.4 Score 0 vs Score 1 Transition Analysis

27.4.1 Paired Transition Matrix (Pre-AI to Post-AI)

27.4.2 Score 0 <-> Score 1 Transitions Only

27.4.3 Clinical Impact Summary

27.5 HER2-Low Classification (0 vs Low vs Positive)

27.5.1 Define HER2-Low Category

27.5.2 Row Percentages

27.6 Per-Pathologist Analysis at the 0/1 Boundary

27.6.1 Per-Pathologist Score 0 vs Score 1 Transitions

27.6.2 Per-Pathologist Score Shift Visualization

27.6.3 Per-Pathologist Alluvial: Where Do Score 0 Cases Go?

27.6.4 Where Do Score 1 Cases Go?

27.7 Interobserver Agreement at the 0/1 Boundary

27.7.1 Fleiss’ Kappa: Score 0 vs Score 1 (Binary)

27.7.2 Fleiss’ Kappa: Full 4-Score Agreement

27.7.3 Pairwise Cohen’s Kappa: Score 0 vs Score 1

27.7.4 Overall Percent Agreement for Score 0 vs Score 1

27.8 Cross-Pathologist Concordance for Individual Cases

27.8.1 How Often Do All Pathologists Agree on 0 vs 1?

27.8.2 Disagreement Patterns Among Score 0/1 Cases

27.9 Confusion Matrix: Score 0 vs Score 1 (Pre-AI vs Post-AI)

27.9.1 Aggregate Confusion Matrix

27.9.2 Precision, Recall, and F1

27.9.3 Per-Pathologist P/R/F1

27.9.4 Confusion Heatmap: Full 4-Score

27.10 Biopsy Type Stratification

27.10.1 Score 0 <-> Score 1 Transitions by Specimen Type

27.10.2 Visualization by Biopsy Type

27.11 Impact on Molecular Subtype and Treatment

27.11.1 Subtype Changes Driven by HER2 0 <-> 1 Transitions

27.11.2 T-DXd Eligibility Impact by Subtype

27.12 Concordance Rate Over Score Range

27.12.1 Concordance by HER2 Score

27.12.2 Visualization

27.13 Case-Level Exploration: Discordant 0/1 Cases

27.13.1 List of Discordant Cases

27.13.2 Cases with Pathologist Disagreement on 0 vs 1

27.14 FISH-Integrated HER2 Classification

27.14.1 FISH Data Overview

27.14.2 FISH Results Joined with HER2 IHC Scores

27.14.3 Definitive HER2 Classification Using FISH

27.15 FISH-Based Transition Analysis

27.15.1 Definitive HER2 Transition Matrix (with FISH)

27.15.2 Clinically Relevant Transitions (with FISH)

27.15.3 T-DXd Eligibility with Definitive Classification

27.15.4 Score 2+ Cases: FISH-Resolved Classification

27.16 FISH-Based Agreement Analysis

27.16.1 Agreement: 3-Tier Classification (Negative / Low / Positive)

27.16.2 Confusion Matrix: Definitive Classification (with FISH)

27.16.3 Precision, Recall, F1: Definitive Classification (with FISH)

27.16.4 Comparison: IHC-Only vs IHC+FISH Classification

27.16.5 FISH Summary

27.17 Summary and Clinical Implications

27.17.1 Clinical Recommendations