Exploratory evidence maps for the WHO Classification of Tumours 5th edition for lung and thymus tumors

One of the key requirements for accurate cancer diagnosis is the availability of a comprehensive classification of different tumor types. The World Health Organization (WHO) therefore regularly updates WHO Classification of Tumours (WCT), produced as a website and as the WHO Blue Books. These are essential resources for pathologists and researchers among many other specialists worldwide. The WCT is based on multidimensional criteria, including clinicopathological features; these criteria are described for each tumor type, and essential and desirable criteria are listed to make reliable tumor diagnoses [1, 2].

Tumor classification has seen and is expected to see more major changes in the recent decades due to increased understanding resulting from the application of new technologies, e.g., large-scale molecular profiling of tumor DNA or artificial intelligence and digital pathology [3‐7]. Likewise, new technologies in the field of cancer research are emerging, thus highlighting the importance to approach the classification of tumors from multiple dimensions [3‐5].

Given the emergence of new techniques, as well as the heterogeneity of biomedical and cancer research, the quality and level of evidence (LOE) for the study designs of citations in the WCT often vary significantly [8]. Additional reasons are the rarity of some tumor types that are covered in the WCT and the lack of LOE adapted to tumor pathology. The WCT sets the standards for cancer diagnosis and research. Therefore, the WCT is required to follow the principles of evidence-based medicine (EBM) and aim to synthesize the best available evidence to date. However, identification of the most relevant evidence and assessment of LOE and quality for published data is often difficult. The reasons for this include the high number of existing references and their variable quality. To address and assist in the evaluation of existing evidence, a new technology called the Evidence Gap Map (EGM) has emerged. EGMs synthesize and display available evidence for a given research question and are predominantly used in social sciences and public health [9‐11]. No EGMs or evidence maps had been produced to evaluate the cited evidence for tumor classification before, and the WCT EVI MAP project was developed to fill in this gap.

The present study is encompassed within the WCT EVI MAP project, which is funded by the European Commission horizon grant (101057127) and commenced in 2022. This project’s objective is to create living maps of current evidence to identify gaps and pockets of low-level evidence to assist in the production of the WCT next edition [12].

In this study, we have developed evidence maps for the cited references in the WCT 5th edition (WCT-5) for thoracic tumors [1], particularly for lung and thymus tumors, as well as exploring the LOE of the cited studies for different tumor types and tumor descriptors. For this purpose, we used the traditional LOE for assessing the cited evidence, i.e., in accordance with criteria from the Centre for Evidence Based Medicine (CEBM) of the University of Oxford [13], available for this purpose during the period of study.

We developed exploratory evidence maps according to our previously published protocol [14] for the cited evidence in the WCT-5 for thoracic tumors, which include tumors of the lung and thymus.

EPPI-Reviewer [15] and EPPI-Mapper [16] had been previously selected to develop the evidence maps. Both are open access software designed for reviews which permit the import of items in different format for coding and graphical interactive EGM production [15, 16].

A list of cited evidence in the lung and thymus tumor chapters for thoracic tumors [1] was exported from the WCT-5 and imported to EPPI-Reviewer. Therefore, no structured literature search was performed since our objective was to map citations in the WCT-5 for these tumors, thus creating evidence maps instead of EGMs. We extracted the references from the database that holds the WCT content at IARC, as PubMed IDs, and imported the text as a list to EPPI-Reviewer. Eligibility was restricted to citation in one of the target chapters. All study designs were included. All articles had an abstract written in English.

We applied predefined coding tool for the study design and tumor descriptors. Coding in EPPI-Reviewer for each chapter was done by a single reviewer and another reviewer reassessed the citation if the study design was unclear. If the study design was not assigned by information displayed in the title and abstract, the full text was accessed and assessed. LOE was assigned for the study designs based on criteria from the CEBM of the University of Oxford [13]. We then imported the data to EPPI-Mapper to generate an interactive evidence map to graphically display the results.

The resulting maps include the following key components: tumor types constituting the columns of the map, categories of tumor descriptors (Supplementary Table 1) constituting the rows, and LOEs (Supplementary Fig. 1) inside the resulting squares as bubbles (segmenting attribute), i.e., encompassing both tumor type and tumor descriptor. If a particular study had been cited more than once for different tumor types and/or tumor descriptors, it was plotted multiple times. The number of references per tumor type and tumor descriptor was reflected in the size of the bubbles, whereas the LOE was represented through a four-color code: low level evidence in red, moderate level in orange, and high level in blue. Uncodable citations as per study design were displayed in blue. Systematic reviews and randomized controlled trials were considered high; cohort and case control studies moderate; and cross sectional studies, as well as case series and reports and expert opinions and non-systematic reviews were considered low LOE. Only LOE was considered and not the quality of the study or the journal impact factor. In addition, the aim of this study was to map the cited evidence in WCT-5 but not to assess any author bias.

We observed that references cited in the WCT-5 for thoracic tumors, chapters on tumors of the lung and thymus, mostly correspond to a low LOE based on the CEBM LOE and study design. Additionally, the number of cited references was heterogeneously distributed, with some tumor types including hundreds of citations and others less than 10. It is likely that these differences are related to tumor incidence, i.e., the higher the incidence of a specific tumor, the higher the number of published studies addressing it. In line with this, invasive non-mucinous adenocarcinoma of the lung, which is a high-incidence lung tumor, had the highest number of citations [17]. This is further supported by the fact that the tumors with the highest proportion of high LOE were high incidence tumors and the tumor types presenting only low LOE evidence were rare and very rare tumors. This could be because achieving higher sample numbers and, thus, performing high LOE studies is difficult for rare tumor entities. However, with regard to tumors presenting a higher incidence, we observed that only 3.5% of all cited evidence addressed squamous cell carcinoma, i.e., variations exist across tumor entities. Another possible explanation is the lack of existing evidence for some neoplasms, mainly for rare tumor types, as well as the inaccessibility of some studies. For example, certain types of lung cancer, such as adenocarcinoma, lung squamous cell carcinoma (SQCC) and small cell lung cancer (SCLC) present a higher proportion in non-smoking people of East Asian origin in comparison to non-smokers from elsewhere [18‐20]. Consequently, research written in a language different than English may be less accessible to the research community and editors of the WCT. Yet another possible explanation is that existing evidence is written by non-pathologists and, therefore, does not use tumor descriptors that appear in the WCT headings for certain tumor types.

In addition, the tumor descriptor sections with the highest number of references overall were Prognosis and Prediction and Pathogenesis. This may be due to research focusing more on these areas, or of greater clinical relevance. Therefore, a focus on improving under-researched or under-cited categories of tumor descriptors and pockets of low LOE may be desirable to inform future WCT editions for thoracic tumors. This is especially important for rare tumors, for which solid evidence is often lacking. This could be achieved through international consortiums/groups focusing on particular rare tumors to achieve higher sample numbers and thus perform higher LOE research. In addition to mapping citations within the WCT, we plan to map all available evidence since 2018, when WCT-5 started publication, using an extensive search strategy for tumor types within the WCT EVI MAP project [12], to identify gaps in published research, both for tumor type and tumor descriptor, as well as pockets of low LOE. Knowing the lack of literature in certain tumor types and descriptors through this project will enable researchers to identify research future research areas to focus on.

For lung tumors, the type with the highest number of references, i.e., invasive non-mucinous adenocarcinoma of the lung, also presented the highest number of high LOE. However, we did not observe the same for thymus tumors. We believe it would be interesting to explore the distribution of high LOE references across for other tumors in the WCT [12].

We analyzed the cited evidence in accordance with tumor types and descriptors and LOE as per study designs. However, visualizations of the geographical area in which the studies were performed, and publication date distribution of the selected evidence would help identifying additional gaps and pockets of LOE. Most of the study settings in tumor research are in high-income countries and therefore addressing lung and thymic tumors in people of European origin [21, 22].

Our study has other limitations. Most of the coding of each chapter was done by a single reviewer. However, to overcome this limitation, a second reviewer was employed to solve doubts and disagreements. In addition, an assessment of the risk of bias or overall methodological quality of the selected studies is usually performed in evidence synthesis. The value of a study not only depends on the study design although it is an important aspect. Given the exploratory character of our study, we did not evaluate critically the included studies for the quality.

Finally, the CEBM LOE were primarily designed for clinical and therapeutic medical procedures. Given the specific characteristics of pathology, the possibility of performing studies that correspond to high CEBM LOE is limited [23]. This may explain the high number of low LOE studies in our exploratory map. Having identified this limitation, we recently developed a hierarchy for evidence in tumor pathology (HETP), which adapts CEBM LOE to tumor pathology [24]. This study performed prior to the development of HETP lends support to the concept that CEBM LOE may not accurately reflect the LOE pertaining to tumor pathology. The preparation of evidence maps applying HETP are currently in progress within the WCT EVI MAP project [12], which will enable comparison facilitating adoption of HETP for evidence maps for tumor pathology in future.

The present study represents a first step in the WCT EVI MAP project [12], which aims to assess existing evidence for different tumor types and their descriptors. We observed that, when applying CEBM LOE, most studies were coded as low LOE. We also observed great heterogeneity throughout all assessed areas. The next step will involve mapping the existing evidence for tumor types in WCT-5 using HETP to create living maps to support the editorial boards for the new edition of the WCT for each organ and system. Additionally, we hope to promote high LOE research in the future by mapping existing evidence, both generally and for certain tumor types.