NEJM: A Quartet of Papers about Non Invasive CRC Screening

 NEJM offers a quartet of papers on new tests for non invasive Colorectal Cancer screening.

• Chung et al. report a 10,000 patient study with specificity of about 90% and cancer sensitivity of about 83%-87% depending on definition.  
• Imperiale et al. report a next-generation of a stool test, 20,000 patients, with sensitivity and specificity both in the 90% range.
• Carethers of UCSD provides an op ed. 
• Lo of Chinese Univ Hong Kong provides a brief overview of cell-free DNA detection methods.

NYT here.  AP here.  Genomeweb here.   Medpage here.

##

AI Corner (GPT4): Summarizing the OpEd

[Sources like USPSTF also remark, like Carethers, that "the best CRC screening test is the one that gets completed."]

In his NEJM Op Ed, Dr Carethers discusses the evolution and effectiveness of noninvasive colorectal cancer screening methods. Historically, guaiac-based tests, despite their limitations such as poor sensitivity and the need for dietary adjustments, demonstrated benefits like early cancer detection and reduced mortality. The introduction of Fecal Immunochemical Testing (FIT) improved sensitivity and specificity without requiring diet changes, although annual testing is still recommended. The addition of multitarget stool DNA markers to FIT, as seen in the DeeP-C Study, further increased sensitivity for detecting colorectal cancer and advanced adenomas, albeit with a reduction in specificity and an increase in false positives, recommending a three-year testing interval.

The article highlights two promising studies: the BLUE-C Study, which evaluated a next-generation multitarget stool DNA test showing preserved sensitivity for colorectal cancer and improved specificity, and the ECLIPSE study, which explored a blood-based "liquid biopsy" showing notable sensitivity and specificity for colorectal cancer detection. These advancements aim to enhance screening adherence and effectiveness, especially given the below-target screening rates in the pre-COVID-19 era.

The op-ed emphasizes the critical role of screening in saving lives and the importance of selecting the most effective and accessible tests to increase screening participation and adherence. The best screening test is described as the one that gets completed, underlining the importance of ease of use, cost-effectiveness, and appropriate testing intervals to improve screening rates and reduce colorectal cancer mortality.

(I asked GPT4 for, "an interesting illustration, be sure it is interesting.")

##

In a sidebar, I asked GPT4 to adopt the voice of a journalist and summarize both studies.  I needed to ask for a re-write, which it also did.  Here.

##

NEJM text on Cologuard Tech:

The next-generation [stool] test incorporated a new molecular panel (including the methylated DNA markers ceramide synthase 4 gene [LASS4], leucine-rich repeat-containing protein 4 gene [LRRC4], serine–threonine protein phosphatase 2A 56-kDa regulatory subunit gamma isoform gene [PPP2R5C], and the reference marker zinc finger DHHC-type containing 1 gene [ZDHHC1], while retaining fecal hemoglobin). [from Imperiale]  

Per text in its NCD, Cologuard today is run as: Cologuard tests for two DNA methylation markers [NDRG4, BMP3], seven point mutations on K-ras [codons 12 and 13], quantitative DNA [β-actin], and fecal hemoglobin.  

##

AI-shortened ECLIPSE method, based on the long technical appendix.   

Plasma from each subject was pooled and sent for cfDNA extraction, which was then divided into methylated and unmethylated fractions based on methylation states. Unique dual barcodes were applied to DNA molecules for identification, with subsequent PCR amplification and enrichment targeting about 1Mb of the human genome.

The analysis distinguished between methylated and unmethylated cfDNA, focusing on specific genomic regions. A neoplasia detection algorithm, the Shield Test, evaluated thousands of features to identify cancer-specific signals, incorporating epigenetic changes and somatic mutations. Results were based on methylation-based tumor fraction regression (TFR) and an integrated cfDNA score, determining the sample's status as either positive (abnormal) or negative (normal).

The algorithms were trained on a diverse sample set, including healthy individuals, CRC-negative donors, and CRC patients, refining the detection of tumor-derived cfDNA. This involved quantifying aberrant methylation, assessing molecule counts, and employing logistic regression models to predict the presence of tumor-derived molecules based on methylation status, fragmentation patterns, and detected somatic mutations.