Online Program

Hypothesis Testing in Multiple Health Care Databases for Scientific Insight Generation

*Shaum Kabadi, AstraZeneca 
Jigar Desai, Pfizer Inc. 
Matthew St. Louis, Pfizer Inc. 
Craig Hyde, Pfizer Inc. 
Vinicius Bonato, Pfizer Inc. 
Katrina Loomis, Pfizer Inc. 

Keywords: type 2 diabetes, meta-analysis, real world data, claims data, electronic health records, electronic medical records, genetic disorder

Patients with a diagnosis of hereditary fructose intolerance (HFI) or alpha-1 antitrypsin deficiency (A1AT) and type 2 diabetes (T2D) were identified in four large observational health care databases. The association between both inherited disorders and T2D was compared to the association between T2D and seven negative control chronic diseases with no established relationship with T2D. The unadjusted pooled odds ratio (OR), calculated using a random-effects model meta-analysis, was 3.48 (95% CI: 2.21-5.46) for HFI and 2.71 for A1AT (95% CI: 1.75-4.20). After pooling all patients and adjusting for the negative control chronic diseases using a random-effects model meta-analysis, it was found that HFI patients have a 73% increased odds of T2D (ratio of odds ratios [ROR]=1.73, 95% CI: 1.08-2.75) compared to patients with negative control diseases; the association was stronger when using a fixed-effects model meta-analysis (ROR=2.19, 95% CI: 2.07-2.31). The adjusted association between A1AT and T2D was statistically significant in the fixed-effects (ROR=1.33, 95% CI: 1.27-1.40) model meta-analysis, but not the random-effects model meta-analysis (ROR=1.35, 95% CI: 0.86-2.12). *The work was conducted when the presenter was employed at Pfizer.