This article is 8 months old
Reaction: computer model predicts who needs to be screened for lung cancer

A machine learning model can use data on people's age, duration of smoking addiction and number of cigarettes smoked per day to predict lung cancer risk and identify who needs to be screened, according to a new study published in PLOS Medicine.

03/10/2023 - 20:00 CEST
 
Expert reactions

Isabel Portillo - cribado cáncer pulmón EN

Isabel Portillo

Coordinator of Colorectal and prenatal cancer screening at the Basque Health Service-Osakidetza, researcher in the Cancer Biomarkers group at the Biobizkaia Health Research Institute and secretary of the Board of Directors of the Spanish Society of Epidemiology

Science Media Centre Spain

The study has been carried out using data from different sources by applying mathematical models. It does not add value to what has already been published on known risk factors. 

The quality of the analysis could be improved, as it finally only uses three predictor variables (age, duration of smoking and packs per year). These factors seem insufficient to determine risk, as both gender and socio-economic status should be taken into account to determine and advise individualised screening. 

The work is in line with the evidence (clinical trials), although it does not provide new evidence to those published (NELSON, NLST, PLCO). On the other hand, the authors compare studies of different populations, which may represent a study design bias. The mathematical models are difficult to interpret with my knowledge of the field, as biostatisticians and bioinformaticians are needed to assess the adjustments made and the validation of the models. 

Although the authors present it as a predictive tool, its applicability at a practical level (health professionals) is debatable in the case of indicating a person for screening (low-dose CT). It should be adjusted to be a tool to help both professionals and patients (Framingham study of cardiovascular risk). Biomarkers, environmental, occupational factors should be considered.

"I am part of the LUCIA (Lung Cancer Risk Factors and their Impact Assessment) Horizon EU Study which aims to identify risk factors also by applying predictive mathematical models".

EN
Publications
Assessing eligibility for lung cancer screening using parsimonious ensemble machine learning models: A development and validation study
  • Research article
  • Peer reviewed
  • Modelling
Journal
PLoS Medicine
Authors

Callender et al.

Study types:
  • Research article
  • Peer reviewed
  • Modelling
The 5Ws +1
Publish it
FAQ
Contact