Abstract 1565P
Background
There is international consensus that risk models are more effective than fixed criteria as pack-years and age to select high-risk subjects for lung cancer screening. Here we tested the HUNT LCM against the PLCOm2012, in a Norwegian prospective lung cancer screening study. Both models were thoroughly validated in external large dataset. These risk models have not been compared to each other directly by all participants answering questions from both models. This is of interest for implementation of screening in several countries.
Methods
In the Norwegian prospective lung cancer screening pilot TIDL, 1004 subjects with age 60-79 and > 35 packyears or PLCOm2012 risk > 2.6% were screened in 2023 and re-screened in 2024. All were asked to provide the variables needed for the HUNT LCM and PLCOm2012 calculators. Risk scores for both models were calculated for all subjects at inclusion. The performance of the models was tested by area under the ROC curve (AUC) and by ranking of risk by sample-level risk scores versus the cumulative number of lung cancers diagnosed provided by the two models with the Kolgomorov-Smirnov test.
Results
Out of the 1004 included in the screening trial, 884 had had signed consent to be included in this risk study. In total at 1st and 2nd screening round, 25 individuals were diagnosed with lung cancer. Median pack-years was 41.5 and 43.8 among cases and controls respectively (p=0.44). Median risk score of all screened and the lung cancer cases was 3.7% and 4.4% versus 2.15% and 2.19% by the PLCOm2012 and the HUNT LCM respectively (p>0.05). The AUC for lung cancer diagnosis in one year was 0.522 and 0.546 in the PLCO and HUNT LCM respectively. When ranked according to risk score, the HUNT LCM had a higher detection rate (p<0.001).
Conclusions
In this Norwegian pilot screening study of very high risk individuals, none of the two methods were able to predict lung cancer at the first two screening rounds correctly according to AUC. However, HUNT LCM was statistically superior over PLCOm2012 in ranking and predicting lung cancer diagnosis two years after inclusion. We will follow this up with yearly updates to assess the usefulness of the two models.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
Akershus University Hospital, Lørenskog, Norway.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.
Resources from the same session
1199P - Developing and systematically validating homologous recombination repair gene detection method based on next-generation sequencing
Presenter: Yi Sun
Session: Poster session 10
1200P - Investigation of multiphoton microscopy as an innovative tool for intraoperative section-free histologic investigations in just a few minutes
Presenter: Martí Homs Soler
Session: Poster session 10
1201P - Novel deep learning model and validation of whole slide images in lung cancer diagnosis
Presenter: Alhassan Ahmed
Session: Poster session 10
Resources:
Abstract
1202P - A deep learning approach using routine pathology images to guide precision medicine in metastatic CRC
Presenter: Chaitanya Parmar
Session: Poster session 10
1203P - Analytical evaluation of whole genome sequencing for acute myeloid leukemia
Presenter: Guidantonio Malagoli Tagliazucchi
Session: Poster session 10
1204P - Real-world utility of whole genome sequencing for patients with cancer: Evaluation of a regional implementation of the 100,000 genomes project
Presenter: Helen Robbins
Session: Poster session 10
1205P - A retrospective machine learning-based analysis of nationwide cancer CGP data across cancer types to identify features associated with recommendation of mutation-based therapy
Presenter: Hiroaki Ikushima
Session: Poster session 10
1478P - Dual single-nucleotide polymorphism biomarker combination to select opioid for cancer pain management
Presenter: Yoshihiko Fujita
Session: Poster session 10
1479P - Use of rescue opioids and pain control after ketamine initiation in refractory cancer pain: A multicentric observational study
Presenter: Pablo Gallardo Melo
Session: Poster session 10
1480P - Long term therapy with denosumab and zoledronic acid: A comparative real-world retrospective observational study on skeletal-related events and pain in patients with metastatic breast cancer
Presenter: Giacomo Massa
Session: Poster session 10