Abstract
METHODS
In a cohort of 5,370 ≥ 5-year HL survivors aged 15 to 50 years at HL diagnosis and treated in the Netherlands between 1965 to 2012, we used RT fields and prescribed dose to estimate mean lung dose (MLD). The twinning method was used to divide the cohort into homogenous parts: 80% for model development and 20% for validation using Inverse Probability of Censoring Weighting (IPCW) Area Under the Curve (AUC). Cox proportional hazards models allowing for time-dependent coefficient(s) were used to model time from HL diagnosis to LC and LC-free death as competing event for predicting absolute LC risk up to 30 years after HL diagnosis.
CONCLUSION
We developed a well-calibrated prediction tool that estimates long-term risk of LC with good discrimination based on patient characteristics and MLD, allowing application in HL survivors and contemporary HL patients.
RESULTS
Treatment information was composed of supra-diaphragmatic RT in 75.2% of patients with median MLD of 15.8 Gray. During follow-up, 218 survivors developed LC. Older age, male gender, smoking at HL diagnosis and higher MLD were associated with higher LC risk. The median estimated 30-year absolute LC risk in our cohort was 1.2% (Interquartile range (IQR): 0.8%-1.9%) in non-smokers and 6.9% (IQR: 4.6%-10.9%) in smokers at HL diagnosis. The 20- and 30-year IPCW AUCs were 0.79 (95% Confidence interval (CI): 0.73-0.85) and 0.75 (95% CI: 0.69-0.81), respectively.
BACKGROUND
Hodgkin lymphoma (HL) patients are at increased risk of developing lung cancer (LC), especially after chest radiotherapy (RT). However, there are no tools to predict LC risk for different HL treatments.