Can Pathologists Reliably Categorize Visually Interpreted Immunohistochemical Stains at a Level More Precise Than Present, Equivocal, Absent?
Funda Vakar-Lopez, Jing Xia, Ruth Etziioni, Lawrence D True. University of Washington, Seattle, WA; Fred Hutchinson Cancer Research Center, Seattle, WA
Background: Visually interpreted immunoperoxidase stains (IPOX) are the most widely used prognostic and predictive tissue-based biomarker assays (TBBA). Clinical decisions rely upon accurate interpretation of TBBAs, i.e. expression of ER or her2 predicts response of breast cancer to tamoxifen and trastuzumab, respectively. Though high rates of inter-lab variance have decreased when sources of pre-analytical variability, i.e. length of fixation, have been standardized, variance by pathologists in assessment of IPOXs remains a problem. As new androgen deprivation agents are developed to treat prostate cancer, the demand for TBBAs to qualify a patient for targeted therapy will increase.
Design: Our goal was to determine the largest number of categories of staining intensity that pathologists could reproducibly use in assessing IPOXs. Two pathologists independently assessed anti-androgen receptor immunostains of tissue microarrays containing 398 cores of prostate cancer and paired benign prostate glands from prostatectomy specimens. Cancer and benign glands were visually assessed as % of cells (at 5% increments) at each of 3 levels of staining - absent, faint/equivocal, intense. Concordance of the pathologists in assessing intense staining was evaluated using 6 different scales. Percent of cancer and benign glands that stained intensely in each of the 398 cores was grouped into categories of percent. The categories of percent of each element in each core that stained intensely ranged from 4 to 12 (on a scale of 0 to 100%).
Results: Of the 398 cores, we differed on whether cancer or benign glands, respectively, were present in 109 cores. We excluded these cores from the study. Concordance of assessment of intense staining ranged from 3% (36 to 43% of cancer cells in the 12 category scale) to 54% (51 to 74% of cancer cells in the 4 category scale).
Conclusions: Since concordance was, at best, 54%, we question whether IPOX for biomarkers can be used at any level of precision greater than 3 levels of staining – Present, Equivocal, or Absent. However, our study has caveats: (1) It is retrospective. The pathologists did not prospectively categorize immunostained elements into a pre-designated range of staining intensity. (2) Since we didn't have a threshold (at present there is no prostate cancer TBBA on which treatment decisions are based), we couldn't base our evaluations on a threshold value. (3) Concordance may be higher for other TBBAs, which may have narrower ranges of staining intensity.
Category: Genitourinary (including renal tumors)
Wednesday, March 6, 2013 1:00 PM
Poster Session VI # 179, Wednesday Afternoon