BACR10: The use of Latent Class Analysis to predict cancer risk based on a panel of serum biomarkers in the Swedish AMORIS study

Aida Santaolalla1,Anita Grigoriadis1,Ton Coolen1,Lars Holmberg1,Hans Garmo1,Niklas Hammar2,3,Håkan Malmström2,Göran Walldius2,Ingmar Jungner4,Mieke Van Hemelrijck1

1Kings College London, London, UK,2Karolinska Institutet, Stockholm, Sweden,3AstraZeneca Sverige, Gothenborg, Sweden,4Karolinska Institutet and CALAB Research, Stockholm, Sweden

Presenting date: Monday 2 November
Presenting time: 13.10-14.00

Background

To better understand the complexity of cancer, it is critical to identify biomarkers linked to carcinogenesis and their intrinsic associations. Several methods have been used to explore the relationship among biomarkers. Latent Class Analysis (LCA), a model-based cluster analysis, is used here to capture information from a series of serum biomarkers.

Method

We applied LCA on data from 13,615 participants in the Swedish Apolipoprotein MOrtality RISk (AMORIS) study to evaluate whether a panel of serum biomarkers predicts risk of cancer. The biomarkers represent the metabolisms of lipids (i.e. ratio of triglycerides/HDL cholesterol, apolipoprotein B/A-I), sugars (i.e. glucose and fructosamine), phosphate, calcium, and irons (i.e. iron and total iron binding capacity), as well as liver (i.e. gamma-glutamyl transferase, alanine aminotransferase, and aspartate aminotransferase) and kidney function (i.e. creatinine) and the immune system (i.e. C-reactive protein, albumin, and leukocytes). All biomarkers were dichotomised based on their clinical cut-offs.

Results

LCA indicated that the population consists of four classes based on whether they had normal values for all biomarkers (reference), high values for lipid markers (class 1), high values for liver functioning markers (class 2), or high values iron metabolism markers (class 3). Cox proportional hazards model, adjusted for age, sex, Charlson comorbidity index and education status, revealed that all classes were associated with an increased risk of cancer, compared to the reference: HR: 1.15 (95%CI: 1.02-1.28), 1.37 (1.17-1.62), and 1.11 (0.90-1.36) for class 1, 2, and 3, respectively.

Conclusion

Using LCA we showed that lipid metabolism plays an important role in classifying individuals based on their serum biomarkers, along with serum markers of liver function and iron metabolism. Abnormal values for biomarkers of these three metabolisms could be indicators of cancer susceptibility or early cancer development and calls for further investigation into their biological interactions.