Regulatory Forum Opinion Piece

Abstract

The definition of diagnostic thresholds is an important aspect of identification and recording of histopathologic lesions in toxicology studies. Although the primary goal of the pathology examination is to identify and interpret lesions associated with the administration of the test article, the toxicologic pathologist will encounter many changes in the tissues that are variations in tissue morphology, tissue artifacts, and spontaneous background findings. The pathologist must establish appropriate thresholds to produce a comprehensive record of the findings so that potentially treatment-related lesions may be identified. However, the findings should not be so detailed as to create overly complex data with the appearance of differences when none exist. Care must be taken to be consistent in the identification and recording of background lesions, since they are important for historical control data, which is often used as a reference when interpreting findings in current studies. Insufficient or inconsistent recording of findings may result in a deficiency in the historical control data for the identification and interpretation of a finding in the future.

The primary goals of a pathology examination in a toxicology study are to identify and interpret lesions related to administration of a test article. To achieve these goals, the pathology examination must be thorough, accurate, and consistent. A number of opinion pieces have recently been written addressing appropriate methods of recording and evaluating histopathologic lesions in toxicology studies. Measuring and ordering of lesions have been suggested as being more powerful and objective than more common methods, such as scoring/grading for detecting subtle differences among groups (Holland 2005; Holland and Holland, 2011a; Holland and Holland 2011b). Counterpoints to these recommendations point out that a toxicologic pathology examination is usually conducted in two stages, an “identification stage” and a “confirmation stage,” and that the more objective evaluation methods, if at all applicable, are not applicable to the “identification stage” (Levin 2011; Wolf 2011). Best practices guidelines for toxicologic pathology indicate that the identification stage should be performed in an unblinded fashion (i.e., with knowledge of treatment, more importantly, with knowledge of the baseline “control”) to minimize the chance that subtle lesions will be missed (Crissman et al. 2004). The confirmation stage may be performed in a blinded fashion, since it can be more focused and analytical once the diagnostic features of a potentially treatment-related finding are identified.

In the identification stage of tissue evaluation, all differences from controls in each organ/tissue must be recorded so that lesions potentially related to the test article may be identified. Variations in tissue morphology in controls that differ substantively from those expected for the species/strain should also be recognized to identify potential issues with the animals or testing procedures. Identification of changes indistinguishable from spontaneous findings in tissues should also be identified in the analysis of biomarker studies (Burkhardt et al. 2011).

In the identification stage, pathologists engage in thresholding, which may be defined as the practice of determining which variations in normal morphology will be recorded and which variations below a threshold will not be recorded. Threshold may also be related to the sensitivity of the test (histopathology examination) in terms of the least detectable lesion that can be accurately identified and/or differentiated from normal variations in morphology. This can be viewed as the concept of a “minimal” grade, the amount of tissue change that barely exceeds that which is considered to be within normal limits.

There is a tradeoff in the degree to which a threshold can be applied to variations in tissue morphology. If too many variations are recorded, the resultant data may be overly complex and apparent differences may appear to be related to the treatment when no true differences exist. On the other hand, insufficient recording of lesions may lead to missing test article–related differences in incidence, severity, or distribution of lesions that may originate from or be similar to background lesions. Insufficient recording of findings may also result in a deficiency of historic control data for identification/interpretation of lesions in the future. Setting appropriate thresholds can aid in streamlining the number of diagnoses produced in a study so that treatment-related changes are easier to detect.

Variations in tissue morphology that are considered lesions by some pathologists may be considered by others to be within the range of normal for a particular species or strain of animal. An experienced toxicologic pathologist is familiar with the variations in tissue morphology and background lesions typically encountered in the animal species and strains normally used in toxicology studies. A lesion is usually thought of as a change in tissue morphology outside of the range of normal variation for a particular species or strain. In a population of untreated animals, there are normal, subtle variations in the morphology of tissues, so some degree of thresholding is needed to provide a meaningful compilation of microscopic pathology data. Factors the pathologist must consider in determining appropriate diagnostic thresholds include the morphology noted in the concurrent control animals and the pathologist’s understanding of normal morphology for the age, strain, and source of the test species. An examination of the concurrent controls is essential to determine which variations in tissue morphology are normal for the particular group of animals and experimental procedures. For example, variations in the histomorphology of hepatocytes might include variation in nuclear size, mitotic activity, number of binucleated cells, amount of staining intensity, and vacuolation of cytoplasm (Zbinden 1976). Mild infiltrates of inflammatory cells in the liver may be a normal finding in control animals, but there can be substantive differences among control groups in different studies or among control and treated groups in a single study. Therefore, the concurrent controls cannot be used as the sole reference standard for a study. Observations should also be made relative to a reference standard of other controls from the same species or strain, which may consist of the experience of the pathologist, a collection of histologic preparations of tissues considered to be “normal” for the species and strain being tested, or a complete description of these tissues.

An individual pathologist’s professional judgment might dictate that since specific lesions are common in the particular animal used in the study, the lesions should not be recorded. This approach may be acceptable from an individual animal diagnostic standpoint, an approach in which only important lesions are identified. However, it can lead to problems with toxicologic (experimental) studies. Changes in the incidence, severity, and/or distribution of a particular background lesion may not be detected. Changes may be recognized but recorded only for the dose group(s) in which they occurred, recorded only when they are perceived as increased in severity, or recorded only after they are first recognized. In any of these cases, there may be differences among animals or among dose groups as to how lesions are recorded. Reviewing pathologists may have doubts as to the ability of the pathologist to correctly and consistently identify lesions or may question the degree of care that went into the histologic evaluation of the study. Even if no major lesions have been identified in a particular study, there may be questions of credibility associated with the study under review or with the laboratory that conducted the study. If all lesions have been identified, there should be no questions relative to the thoroughness of the evaluation of the tissues. It is also important that microscopic observations be recorded using descriptive rather than diagnostic terminology. In many cases, a disease diagnosis implies a particular pathogenesis or impact on organ function based on what is known about the specific spontaneous disease, which may be misleading in the experimental setting of a toxicity study.

The consistent identification and recording of background lesions is also important to establish historic control data to interpret the occurrence of lesions in current studies. Historic control data is very important in the interpretation of studies with large numbers of lesions (e.g., rodent carcinogenicity studies) and studies with low numbers of animals (e.g., non-rodent toxicology studies). In the former, the overall incidence and variability of particular lesions may be important in interpretation. In the latter, simply the rare occurrence of a lesion may aid the interpretation. To develop accurate historic data, lesions must be consistently identified (which includes consistent application of thresholds) and consistent terminology must be applied. Irizarry Rovira et al. (2011) discuss the importance of recording of spontaneous or background microscopic findings by contract research organizations (CROs). The sponsor should discuss with the CRO how these findings are recorded, the consistency of the recording practices, and how these findings are graded. This step is important for establishing historical data to aid in the interpretation of lesion occurrence/severity in toxicology tests. Irizarry Rovira et al. are concerned from the sponsor’s perspective, for consistency of reporting across a series of tests with the same test agent, and having background data to aid in interpretation. The sponsor should make sure that this procedure is being appropriately followed by the CRO. However, establishing the thresholds, terminology, and grading for these background lesions is the responsibility of the CRO. The model belongs to the CRO. The thresholds, terminology, and grading for background lesions should not be significantly altered from study to study within a CRO, depending on individual sponsor’s preferences. However, it must be recognized that even within the same CRO, there are often individual differences in thresholds and grading between pathologists and even from study to study with the same toxicologic pathologist. Variations in threshold and/or grading may, at times, facilitate interpretation of lesions based on the specific aspects of the study being recorded. However, these variations should be kept to a minimum or defined within the context of the study so that the control historic database may be used to aid in interpretation of lesions.

The second major goal of a pathology examination is to interpret the significance or importance of treatment-related lesions. It stands to reason that only lesions that are identified can be interpreted. The identification of a definitive relationship to treatment can be made only after all observations are made, so as not to bias the observations. This would not be a problem with indisputable lesions (i.e., moderate to marked hepatocellular necrosis), but it may present a problem with background lesions that may be increased or decreased in severity or incidence because of treatment. Excessive detail in the observation and identification of lesions may result in the identification of treatment-related lesions that are of little or no consequence to the animal. Because of relative dosages of the test article, differences in metabolism, or enzymatic functions of particular tissues, these lesions may be of no importance to humans. The study pathologist has the responsibility to define these points and to put these lesions into proper context.

In summary, the identification of all lesions is important in the initial microscopic evaluation (the identification stage) of tissues in a toxicology study. The application of appropriate thresholds should be applied to adequately identify potentially treatment related lesions, but it should not be so detailed as to create overly complex data with the appearance of differences when none exist. The development of consistent historic control data is important to aid in the identification and interpretation of treatment-related lesions.