Age at Vaccination a Potential Confounder of Studies of Aluminum Adjuvants

Two large observational studies examining the association between aluminum adjuvant exposure from childhood vaccines and subsequent asthma diagnosis yielded conflicting results: Daley et al. (2023) reported a positive hazard ratio (HR) of 1.19 per mg (95% CI: 1.14-1.25), while Andersson et al. (2025) found a negative HR of 0.96 per mg (95% CI: 0.94-0.98). This discrepancy potentially arises from a failure to test linearity assumptions in the dose-response relationship. This paper reanalyzes the Daley et al. dataset using the same population and Cox proportional hazards methodology, but incorporates tests for nonlinearity by dividing the 0-24 month exposure window into six age-based sub-windows (0-1m, 1-3m, 3-5m, 5-7m, 7-15m, 15-23m) and assessing hazard ratios per sub-window. Results revealed significant nonlinearity as a function of age at exposure, with sub-window HRs ranging from 0.83 (0-1m) to 1.94 (1-3m), all statistically differing from the overall HR of 1.27 via two-sample t-tests (adjusted t-statistics: -9.27 to 5.98). No nonlinearity by dose was observed. These findings indicate that differing vaccine schedules in the U.S. and Denmark may have sampled distinct domains of a nonlinear function, confounding the results. Uncorrected dose-response nonlinearities can cause arbitrarily large errors and future research protocols should include explicit tests for these possible nonlinearities.