Abstract P5-10-06: Risk of recurrence estimates with IHC4 are tolerant of variations in staining and scoring
Bibliographic record
Abstract
Abstract Aims 1. To determine the degree to which alterations in immunohistochemical (IHC) staining for ER, PgR, and Ki67 affect IHC4 scores and distant recurrence estimates. 2. To examine the level of scoring precision needed for accurate use of the IHC4 score. Background The IHC4+clinical (C) score combines assessment of protein expression levels of ER, PgR, HER2 and Ki67 with clinicopathological parameters to identify breast cancer patients at low, intermediate or high risk of distant disease recurrence so aiding treatment decision-making (Cuzick J, et al. JCO 2011;29:4273). The score is used in clinical decision making in our hospital (RMH). Our published studies have shown it provides information that improves decision-making on adjuvant chemotherapy in >65% of patients (Barton S, et al. BrJCancer 2012;106:1760; Yeo BJ, et al. AnnalsOncol 2014;25suppl_1:i2). Despite its low cost and wide availability reported use of IHC4+C in other institutions remains limited (Lakhanpal R, et al. JCO 2014;32:abstr2549); one explanation is the perception that IHC-based methods and the assessment of them lack precision, reproducibility and portability. We have examined the effect of decentralized testing and easily reproduced estimate-based scoring methods on the IHC4 score to determine its suitability for wider adoption. Methods A TMA was constructed from 30 breast cancer cases representative of those for which IHC4+C is requested at RMH. Sections were distributed to three centers that undertake diagnostic breast cancer IHC and that use reagents and platforms from Dako, Leica or Ventana who provide most IHC for oncology globally. Centers carried-out staining using their standard procedures and returned slides for central assessment. Results were compared against those obtained at RMH using standardized methods previously described (Barton S, et al. BrJCancer 2012;106:1760), and were used to calculate IHC4+C and 10-year distant recurrence probability (%DRprob) scores. In parallel the TMA slide stained at RMH for ER was scored by a variety of simplified non-counting based methods. Results were compared with those produced by counted H-Score when used to calculate IHC4+C and %DRprob. Results There was a high-degree of correlation between individual IHC results produced by all three external centers and those of RMH, and of the IHC4 and %DRprob scores derived from them (Tables 1 & 2). Scoring method for ER could be adapted to require less precision without significantly affecting IHC4+C and %DRprob results (correlation coefficients range: 0.90 - 0.98). IHC4%DRprob RMHDakoLeicaRMHDakoLeicaDako0.92 0.97 Leica0.890.95 0.980.98 Ventana0.930.960.950.980.980.98 Table 1. Correlation coefficients for all pair-wise comparisons MeanMedian difference RMHDakoLeicaRMH9.9% Dako8.4%1.0% Leica8.9%0.9%0.5% Ventana9.1%0.6%0.5%0.6% Table 2. Showing for each center, the mean %DRprob score and median value obtained when the absolute difference was calculated between matched scores Conclusion The IHC4+C algorithm is tolerant of variation in staining and ER-scoring method used. Although additional comparative studies are required to confirm them, these data support the use of IHC4+C in routine clinical practice outside its institute of origin. Citation Format: Andrew Dodson, Lila Zabaglo, Belinda Yeo, Keith Miller, Ian Smith, Mitch Dowsett. Risk of recurrence estimates with IHC4 are tolerant of variations in staining and scoring [abstract]. In: Proceedings of the Thirty-Seventh Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2014 Dec 9-13; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2015;75(9 Suppl):Abstract nr P5-10-06.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.022 | 0.102 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.014 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".