New LIH study identifies major gaps in survival and mortality prediction models for older adults with cancer

Review of models reveals widespread methodological bias, precluding their use in clinics

• Ageing, Cancer, and Disparities Research Unit – ACADI
• Department of Precision Health

A comprehensive review of 250 prediction models for overall survival and mortality in adults aged 65 and older with cancer has found that all models suffer from significant methodological flaws, precluding their use in clinics. Conducted by the Ageing, Cancer, and Disparities Research Unit (ACADI) group and members from the International Society of Geriatric Oncology (SIOG), the study highlights a critical gap between current modelling practices and clinical applicability. The findings emphasise the need for more robust approaches to improve risk prediction in ageing cancer populations.

A new study published in The Lancet by the ACADI group from the Department of Precision Health at the Luxembourg Institute of Health (LIH) has uncovered substantial limitations in survival and mortality prediction models for older adults with cancer. Postdoctoral researcher Dr Pauline Duquenne and ACADI group leader Dr Sophie Pilleron led a collaborative project involving 27 clinicians and researchers who systematically reviewed 250 published models designed to predict the survival and mortality outcomes of patients aged 65 and above suffering from cancer.

Analysis found that more than half of the models did not include critical variables such as comorbidity, nutritional status, or cognitive function. These factors are widely recognised as essential in assessing disease prognosis in older populations.

More importantly, all models included in the review were found to have a high risk of bias. The primary methodological concerns included inappropriate handling of continuous predictors, inadequate predictor selection, and insufficient validation procedures. These shortcomings preclude the applicability of the models in clinical practice.

Our findings show that none of the prediction models reviewed are fit for purpose in older adults with cancer. There is a clear need to adhere more rigorously to established methodological standards when developing these tools,

said Sophie Pilleron, leader of the ACADI group.

The authors stress that future research must follow established scientific guidelines for prediction model development and validation. Incorporating geriatric variables and improving statistical methodology are essential steps towards building clinically useful tools.

This work represents a significant contribution to the modelling domain, particularly in highlighting the disconnect between current research practices and the real-world needs of ageing populations.

Funding & Collaborations

This research was done in the framework of ATTRACT project ReDiCo (16731054) funded by the Luxembourg National Research Fund (FNR).