An efficient sampling strategy for selection of biobank samples using risk scores

Area of Science:

• Epidemiology
• Biobanking
• Biostatistics

Background:

• Case-control studies are crucial for investigating disease etiology.
• Biobanks offer rich data but require efficient sample selection strategies.
• Existing methods may not fully leverage available covariate data for optimal sample selection.

Purpose of the Study:

• To propose a new sample-selection strategy for case-control studies utilizing biobank data.
• To develop a risk-score based approach for stratifying samples.
• To improve the efficiency and precision of biobank research.

Main Methods:

• Utilized an ongoing Swedish case-control study on fetal exposure to endocrine disruptors and childhood overweight.
• Defined cases (BMI ≥18 kg/m²) and controls (BMI ≤17 kg/m²).
• Developed a logistic regression model to estimate overweight risk scores based on covariates, categorizing children into low, medium, and high-risk groups for sample selection.

Main Results:

• The risk-score model incorporated smoking, birth weight, parental BMI, residence type, and economic situation (AUC=67%).
• Case group risk profiles were: low (12%), medium (46%), and high (43%).
• The proposed strategy, stratifying by risk score and matching on sex, demonstrated consistent improvements in statistical precision via simulations.

Conclusions:

• Risk scores derived from survey or register data can enhance sample selection in biobank studies.
• This approach improves the ability to study heterogeneity of exposure effects.
• The strategy offers a more precise and efficient method for biobank-based epidemiological research.