Regulatory Forum Opinon Piece: Raw Data in Pathology—Always a Conundrum,Often a Controversy*

There are few topics that can garner as much angst, concern, and sometimes downright hostility among non-pathologists discussing issues in Good Laboratory Practice (GLP) studies as the topic of raw data in pathology. The very idea that a pathologist determines, with few restraints, the content of raw data for histopathology, and indeed, when the data “become raw,” can lead to considerable consternation. I have had the experience of speaking at a conference of the Society of Quality Assurance about the nature of pathology raw data, and some members of the audience indicted pathologists at large with the charge of “falsifying data.” Although this incident illustrates how emotionally charged this topic can be among non-pathologists, what is our understanding of this issue within our own ranks?

I would like to believe that most pathologists involved in reading, writing, or reviewing GLP pathology work know what raw data in pathology actually are, since this topic has been addressed many times in the literature. But is that a true assumption? Since there has been a lot of recent regulatory discussion about pathology issues such as peer review, sponsor influence, and interim data submission that impinge upon the question of raw data for pathology, I would like to review the basic concepts of raw data in pathology again and raise some questions that have shaken my previous conviction that this was a settled matter.

To begin with, the GLP regulations define raw data as follows:

Raw data means any laboratory worksheets, records, memoranda, notes, or exact copies thereof, that are the result of original observations and activities of a nonclinical laboratory study and are necessary for the reconstruction and evaluation of the report of that study.

The issue with histopathology data in a toxicology study is that our original observations are often refined, modified, or simply changed as we progressively evaluate the slides and gain more information, perspective, and insight into the lesions. So our initial observations may be extensively modified or altered completely by the end, and they would not, therefore, satisfy the second attribute of raw data—that they are “necessary for the reconstruction and evaluation of the report of that study”; that claim could be made only for the final version of the data. In order to address this issue, United States Food and Drug Administration officials decided in 1987, after almost 10 years of consideration, that raw data for histopathology would be “the final signed pathology report,” which meant, of course, that any “interim notes” about changes observed on the slides were not raw data and could therefore be discarded without any record. This decision was rendered in 1987 based on initial guidance established in 1978, when data were often collected with paper and pen, and tables and reports were usually issued and signed in hard copy, page by page, by the various investigators and study director. However, since most pathology data are now collected and even signed electronically, is this guidance anachronistic, and does it alter our perspectives on what constitutes raw data for histopathology?

Currently, most histopathology data are collected using a well-validated electronic data acquisition system that is typically configured to allow free modification of the data until a point at which the pathologist electronically designates the data as “locked.” Subsequently, modifications are allowed, but they are tracked via an audit trail. Previously in my own mind, I had considered the “locked” database as raw data, based simply on the fact that one cannot alter it without a reason and an audit trail. I assumed that the “signed pathology report” referred to in the regulations was now the “locked” electronic data tables. However, recent anecdotes suggest that regulators have reaffirmed the position that, based on a literal interpretation of the written regulations, raw data for histopathology are still both the data tables using “locked” data, as well as a written narrative signed by the study pathologist. I am in agreement that regulators cannot make decisions on the safety of patients without having raw data on which to base those decisions. But what seems to be the current conundrum is whether raw data should be just the locked database, or whether they also should require a signed narrative.

On one hand, the “locked” database of diagnoses could be considered like any other database of clinical pathology, organ weight data, or clinical signs observations, none of which requires an associated narrative to explain them or make them raw data. Any narrative associated with those data is seen as providing an additional insight, with which one is free to agree or disagree. It may also be argued in a parallel fashion that any explanatory narrative for histopathology data can be provided in comments within the database associated with diagnoses and be similarly “locked” within the database as part of the raw data.

On the other hand, the contrary position holds that histopathology data are different, and that the database is only part of the story that must be coupled with the report narrative that explains exactly what the pathologist was inferring when he or she made the diagnosis. Thus, someone would be prevented from misinterpreting a diagnosis in a table to mean something that the study pathologist did not intend. Tables of diagnoses do not necessarily contain all the information needed to determine a treatment effect in a given study. Even if the report recipient is an experienced pathologist (not usually the case), these decisions may depend on things that are not necessarily in the data tables, such as the appearance of control tissues, co-localization of focal changes within the tissue, knowledge of common spontaneous findings encountered in that facility, and so on. Such decisions, which may be crucial for the interpretation of a study, can be made only by the study pathologist, who is familiar with the tissues. His or her opinion is particularly critical for studies with a small number of animals, or in which the incidence of changes is low, as it often is at doses where the no observed effect level/lowest observed effect level decision is made. Lesions with different pathogenesis may have the same morphological appearance, or conversely, lesions with the same pathogenesis may have differing morphological appearances, especially with differing chronicity. Similarly, the cause of important in-life clinical signs (including death) cannot always be inferred from the pathology data tables and require the study pathologist’s opinion as to their cause and hence relationship to treatment. Thus, without the study pathologist’s written opinion, the data may be incomplete; the data tables and the narrative are inextricably linked, and alone, either would be incomplete. An important implication that stems from this thinking is that a reader cannot lightly disagree with the diagnoses or written interpretation (notwithstanding obvious errors or poor writing) without review of the glass slides.

I set forth my perspective on this conundrum since I have been thinking hard about this for some time, and I would appreciate hearing debate about this interesting and controversial subject. Although I have tried not to take a side here, I do have my own bias that, for the sake of objectivity (or at least the perception of it), I have tried not to reveal. However, since the GLP regulations are currently being reviewed by the Food and Drug Administration with the aim of creating an updated code of GLP, I firmly believe that it is incumbent upon the STP and sibling organizations to provide clear and robust guidance about this and other matters pertaining to the best practice of toxicologic pathology.