Histopathological Scoring

Histopathological scoring (or grading) is key in most comparative and toxicologic pathologists’ daily practice. Scoring of histologic lesions enables quantitative assessment and statistical analysis of phenotype or treatment effects. There are nearly as many scoring paradigms as disease models, yet novel or modified scoring systems are often needed when approaching a new model or toxicologic event. Given the ubiquity and importance of scoring systems in comparative and toxicological pathology, the Comparative Pathology Special Interest Group held a full-day premeeting workshop in conjunction with the Annual Meeting of the American College of Veterinary Pathologists in Vancouver, BC, on approaches to quantitative and semiquantitative scoring in translational pathology studies. Here we present general insights and key takeaway points and references on some of these topics. Included are 4 commentaries expected to be of broad interest on topics that have not been previously published in this forum. Other published resources are available for veterinary pathologists. ^{4,12,16,23,25,28}

There is no “one size fits all” scoring system, either for hematoxylin and eosin–stained slides or immunohistochemistry. Individual study needs, previously established scoring systems for the disease process or model at hand, statistical analysis, reproducibility, and data presentation will define and inform scoring systems. Ultimately, to be useful, any semiquantitative scoring/grading system needs to be (1) well defined (precise grading criteria and consistent terminology for each lesion type), (2) relevant (focused on those lesions shown to be important in disease progression), and (3) reproducible (provide similar grades/scores when applied by different pathologists (Peter Vogel, personal communication, 2017). ^12,13

Blinding

Many nonpathologists requesting the support of a pathologist are generally under the impression that blinding the pathologist to the details of the project and the study groups is necessary to eliminate bias. However, it is critical that the pathologist has full disclosure to the details of the study to provide the most robust and valid data. ²³ The study materials can be reviewed in an unblinded fashion on the initial review to calibrate the pathologist to the ranges of responses and establish the appropriate criteria for a semiquantitative scoring system. The histologic sections can be blinded and rescored by the same person or in a peer-review process. In addition, knowledge of at least the different groups of animals helps prevent diagnostic drift in long-term studies. ^2,13,29

Statistics

It is important to choose statistical methods that are appropriate for analysis and robust reporting of data typically generated by scoring systems, and ideally this should include consultation with a statistician. ^{8 –11} The goal is to use the statistical tests that most accurately and appropriately answer the scientific questions, as detailed in this issue. ²² Morphological data are often ordinal and thus require nonparametric tests like the Mann-Whitney U. Reporting must include the numbers of animals and strain, sex, age, and weight of animals, yet only 59% of published studies report such parameters. ^15,24 Ultimately, the goal of reporting is to be transparent, thorough, and clear.

Severity Grading in Toxicologic Pathology

The STP Best Practices Guidelines ²³ have built upon Crissman et al, ³ which recommend that severity grades are definable, reproducible, and meaningful and recommend that “a description of the grading criteria/scheme should be included in the narrative for target lesions where severity is critical to interpretation of the study.” A similar approach for academic comparative pathology is detailed in this issue. ¹⁹

Digital Imaging, Automated Quantitation, and Stereology

Digital imaging is widely adopted and analysis of digital images is evolving with high-throughput image acquisition. Numerous analysis options range from simple to complex. As the technologies become more advanced, pathologists will increasingly be called upon to design studies, refine methodologies, define the regions and lesions of interest, assist in interpretation, and validate the analysis. Pathologists need to have a working understanding of the technologies, methodologies, data management, and validation to provide the pathobiological context of the big data generated by some of these systems. ^1,20

Cell Death

The classifications of types of cell death are rapidly evolving, are increasingly complex, and include molecular characterizations, ^7,18 in addition to morphology, to assist pathologists with distinguishing types of cells death beyond apoptosis and necrosis. Detailed in this issue are the current classifications of cell death, systems for scoring, and quantification in tissue sections. ¹⁴

Disease Models and Organ Systems

Individual disease models or organ systems may have unique considerations in approaches to scoring. Commonly used models may have a number published approaches, ¹⁶ and most pathologists will review the literature for published scoring systems prior to working on a novel model and then modify a previous system if needed to suit the particular needs of the study or pathologist’s preference. Certain organ systems have morphologically complex physiologic responses that can affect scoring (eg, hematopoietic system ⁶ and skeletal system). ²¹ Complex systemic pathophysiologic changes that occur with aging such as degenerative, inflammatory, or spontaneous neoplastic lesions can be captured and semiquantified using systematic approaches. ^26,27 Induced cancer models are diagnostically challenging, which greatly affects data collection and appropriate interpretation. For example, proliferative lesions in the lungs of mice with a conditional targeted mutation of Kras in type II pneumocytes are abundant and coalesce, making quantitation and characterization of the proliferative lesions difficult. ⁵ Numerous animal models of cancer are in widespread use, and many have well-established scoring systems. ^5,17