Methods
Disclaimer
Despite I use systematic review methodology, this is not intended to serve as a formal systematic review for publication. In addition, the article summary is generated by AI (large language models) and may contain inconsistencies or errors. I do not supervise the contents of article selection or summary, is all made by AI. In addition, we do not provide access to the full articles. To read the full articles, you need to have a subcription from the journal or be affiliated to a institution that has access to it.
Citation
If you feel like using part of this information, I kindly ask you to cite it:
Lazar Neto, F. NewCancerPapers Project. Cancer Clinical Trials. https://newcancerpapers.com/cancerclinicaltrials/methods
Search Strategy
The search strategy was optimized in an iterative process to include most clinical trials of cancer topics. However, as expected, its much more sensitive than specific. Here is the search I use:
("neoplasms"[MeSH Terms] OR "cancer"[Title/Abstract] OR "carcin*" [Title/Abstract] OR "adenoc*" [tiab] OR "melanoma" [tiab] OR "glioblastoma" [tiab] OR "astrocytoma" [tiab] OR "tumor*" [tiab] OR "oligodendroglyoma" [tiab] OR "sarcom*" [tiab]) AND ((clinical[Title/Abstract] AND trial[Title/Abstract] ) OR (phase [tiab] AND trial [tiab]) OR clinical trials as topic[MeSH Terms] OR clinical trial[Publication Type] OR NCT0*[tiab] OR random*[Title/Abstract] OR random allocation[MeSH Terms] OR therapeutic use[MeSH Subheading]) NOT (review [pt] OR "Cell Line, Tumor" [MeSH Terms] OR "Xenograft Model Antitumor Assays" [MeSH Terms] OR "Tumor Cells, Cultured" [MeSH Terms] OR "Cell Line, Tumor" [MeSH Terms] OR "Propensity Score" [MeSH Terms] OR "Retrospective Studies" [MeSH Terms]) NOT (animals[MeSH Terms] NOT humans[MeSH Terms])The results of this search are filtered further to include only Q1 journals using Clarivate Journal Citation Reports. This was intended to increase the specificity of the articles selection, even though it may loose some articles that match the criteria.
For this review, I classify any article with an impact factor ≥ 10 as a Top-Tier Journal
AI Review and Scoring
The prompt I use is not currently available but I intend to make it available in the future (still need to proper organize a github repository for this project). Using chain-of-though and llm reasoning, I ask the LLM to classify the articles with a score from 0% to 100% in how likely it matches all the criteria. In addition, I ask it to provide a reasoning for each criteria it assessed. The criteria for this specific topic:
- INCLUDE: Prospective human clinical trials of cancer treatments that report results regardless of treatment assignment method (single-arm, comparative, blinded, unblinded, among others).
- EXCLUDE: Systematic reviews or meta-analyses.
- EXCLUDE: Studies involving animals.
- EXCLUDE: Retrospective studies.
- EXCLUDE: Studies primarily focused on developing predictive modelling algorithms or nomograms.
- EXCLUDE: Prospective studies which do not involve any kind of active treatment assignment (not a clinical trial).
- EXCLUDE: In-vitro studies.
- EXCLUDE: Behavioral, lifestyle changes, or dietary interventions.
- EXCLUDE: Protocol type of studies (studies that only report a protocol without results)
- EXCLUDE: Studies where the intervention is surgery or any technical procedure where the outcomes are short-term post-procedure clinical outcomes or non-cancer related outcomes.Only articles with 90% or more of higher are displayed in this website.
AI Cancer Classification
Journal articles with a score of 90% or higher are further classified into different cancer topics (listed in the sidebar). Although some articles have MESH tags, most don’t have it when first publised and thats the reason why I use AI to classify them.
Time-Frame
Here are displayed only published articles from the past 60 days. Specifically for the latest articles section, we display only the last 15 days because of the number of articles displayed.