PATHWORK TISSUE OF ORIGIN TEST KIT-FFPE (ORIGIN TEST KIT-FFPE

The Pathwork Tissue of Origin Test Kit-FFPE is an in vitro diagnostic tool that employs microarray-based RNA expression profiling of formalin-fixed, paraffin-embedded tumor samples to determine their likely tissue of origin. It utilizes a machine learning-derived algorithm to compare a tumor's gene expression signature against a database of 15 known tumor types, generating Similarity Scores to assist clinicians in complex cancer diagnoses.

The Pathwork® Tissue of Origin Test Kit-FFPE is an in vitro diagnostic system designed to aid in identifying the tissue of origin for poorly differentiated, undifferentiated, or metastatic tumors. It analyzes RNA expression patterns from formalin-fixed, paraffin-embedded (FFPE) tumor specimens, which must contain at least 60% viable tumor cells. The methodology involves extracting total RNA (minimum 150 ng), performing a two-cycle RNA amplification, and then hybridizing the biotin-labeled cDNA target to the custom Pathchip microarray. This microarray, similar to the Affymetrix GeneChip HG-U133A, utilizes 2000 gene-specific probesets, selected via machine learning from a database of 2196 specimens, to measure transcript levels. An algorithm then processes these expression values to generate 15 Similarity Scores, one for each of the tissue types on the test panel (e.g., bladder, breast, colorectal), which sum to 100, indicating the likelihood of origin from each respective tissue. The test's performance was established through non-clinical and clinical validation. Reproducibility studies across three laboratories showed high correlation for Similarity Scores (r = 0.92-0.93) and an overall lab-to-lab concordance of 89.3%. A clinical validation study with 462 specimens demonstrated that the test results matched available diagnoses 88.5% of the time (95% CI [85.3, 91.3]). A Similarity Score of ≤5 was 99.8% effective in ruling out a tissue type as the origin. The test has specific limitations: it is not intended to diagnose cancers of unknown primary that are unclassifiable by current methods, subclassify known tumors, predict prognosis, or distinguish primary from metastatic sites. Furthermore, analysis of tumor types not included in the 15-tissue database (off-panel specimens) showed they can yield high Similarity Scores for panel tissues, indicating potential false positive associations; for example, small cell lung cancer specimens often showed similarity to non-small cell lung cancer.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the two-cycle RNA amplification protocol for FFPE samples, utilizing both polydT and random primers, compare in fidelity and representation of the transcriptome to a one-cycle protocol, especially concerning degraded RNA common in FFPE tissues?; What were the specific machine learning algorithms and cross-validation strategies employed in the development of the predictive model using the 2196 specimens, and how was model overfitting addressed?; Given that tumor types not in the 15-tissue database can show high similarity scores (e.g., small cell lung cancer with non-small cell lung, CNS/endometrial tumors with ovarian), what further molecular or immunohistochemical markers are recommended to differentiate these in clinical practice when results are ambiguous or unexpected?; The test requires at least 60% viable tumor; how does the algorithm account for or mitigate the influence of RNA from non-tumor cells (stroma, immune cells) or necrotic tissue (up to 40% allowed) on the gene expression signature and resulting Similarity Scores?; Considering the performance differences across various tumor types (e.g., Gastric PPA 72.0% vs. Hepatocellular PPA 96.0%), what are the underlying biological or technical reasons for this variability, and are there plans to refine the marker set for less performant tissue types?