Forskning
Udskriv Udskriv
Switch language
Region Hovedstaden - en del af Københavns Universitetshospital
Udgivet

Discovery and validation of a colorectal cancer classifier in a new blood test with improved performance for high-risk subjects

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

DOI

  1. Distinct proteome pathology of circulating microparticles in systemic lupus erythematosus

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  2. Serum proteome pool changes in type 2 diabetic patients treated with anakinra

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  1. Evaluation of algorithm development approaches: Development of biomarker panels for early detection of colorectal lesions

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  2. Prognostic factors for relapse in patients with clinical stage I testicular cancer: protocol for a Danish nationwide cohort study

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  3. Genome-wide cell-free DNA fragmentation in patients with cancer

    Publikation: Bidrag til tidsskriftLetterForskningpeer review

  4. Mandibular advancement device therapy for obstructive sleep apnea: a prospective study on predictors of treatment success

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

Vis graf over relationer

BACKGROUND: The aim was to improve upon an existing blood-based colorectal cancer (CRC) test directed to high-risk symptomatic patients, by developing a new CRC classifier to be used with a new test embodiment. The new test uses a robust assay format-electrochemiluminescence immunoassays-to quantify protein concentrations. The aim was achieved by building and validating a CRC classifier using concentration measures from a large sample set representing a true intent-to-test (ITT) symptomatic population.

METHODS: 4435 patient samples were drawn from the Endoscopy II sample set. Samples were collected at seven hospitals across Denmark between 2010 and 2012 from subjects with symptoms of colorectal neoplasia. Colonoscopies revealed the presence or absence of CRC. 27 blood plasma proteins were selected as candidate biomarkers based on previous studies. Multiplexed electrochemiluminescence assays were used to measure the concentrations of these 27 proteins in all 4435 samples. 3066 patients were randomly assigned to the Discovery set, in which machine learning was used to build candidate classifiers. Some classifiers were refined by allowing up to a 25% indeterminate score range. The classifier with the best Discovery set performance was successfully validated in the separate Validation set, consisting of 1336 samples.

RESULTS: The final classifier was a logistic regression using ten predictors: eight proteins (A1AG, CEA, CO9, DPPIV, MIF, PKM2, SAA, TFRC), age, and gender. In validation, the indeterminate rate of the new panel was 23.2%, sensitivity/specificity was 0.80/0.83, PPV was 36.5%, and NPV was 97.1%.

CONCLUSIONS: The validated classifier serves as the basis of a new blood-based CRC test for symptomatic patients. The improved performance, resulting from robust concentration measures across a large sample set mirroring the ITT population, renders the new test the best available for this population. Results from a test using this classifier can help assess symptomatic patients' CRC risk, increase their colonoscopy compliance, and manage next steps in their care.

OriginalsprogEngelsk
TidsskriftClinical Proteomics
Vol/bind14
Udgave nummer28
Sider (fra-til)1-12
ISSN1542-6416
DOI
StatusUdgivet - 1 jul. 2017

ID: 51545050