False Indications of Dose-Response Nonlinearity in Large Epidemiologic Cancer Radiation Cohort Studies; A Simulation...

Area of Science:

• Epidemiology
• Radiation Oncology
• Biostatistics

Background:

• Large cohort studies on radiation exposure and cancer risk are crucial for policy.
• Increasingly, analyses rely on detecting nonlinear dose-response relationships.
• Isolated findings of nonlinearity may represent statistical fluctuations rather than true effects.

Purpose of the Study:

• To investigate the prevalence of false positive indications of dose-response nonlinearity in radiation cancer epidemiology.
• To assess the impact of statistical fluctuations on nonlinearity findings.
• To evaluate methods for correcting potential biases.

Main Methods:

• Monte Carlo simulations were used to replicate analyses on six radiation exposure datasets.
• Five indicators of nonlinearity were tested against a linear null hypothesis.
• Akaike's Information Criterion (AIC) was employed for model selection and correction.

Main Results:

• False positive rates for nonlinearity indicators reached approximately 25% per study and 50% across six studies.
• False above-zero threshold doses were identified over 50% of the time.
• AIC-based correction reduced false occurrences to 8-19%.

Conclusions:

• The study highlights a significant risk of false positives when identifying dose-response nonlinearity in radiation cancer studies.
• Uncorrected biases can distort meta-analyses and influence regulatory decisions.
• Reporting threshold doses requires noting the potential for high false prevalence rates.