Abstract: TH-PO451
Estimating the Association Between Urinary Cadmium and Hypertension: Bias Introduced by Missing Data on Renal Function
Session Information
- Hypertension and CVD: Epidemiology, Risk Factors, Prevention
October 25, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Hypertension and CVD
- 1401 Hypertension and CVD: Epidemiology, Risk Factors, and Prevention
Authors
- Gaggl, Martina M., Medical Universitiy of Vienna, Vienna, Austria
- Wang, Tiansheng, Gillings School of Global Public Health, UNC, Chapel Hill, North Carolina, United States
- Franceschini, Nora, Gillings School of Global Public Health, UNC, Chapel Hill, North Carolina, United States
Background
Cadmium exposure is associated with hypertension and ultimately all-cause mortality. Reduced renal function leads to decreased urinary cadmium excretion, and is aside from smoking an important confounder of the causal relationship of urinary cadmium and hypertension.
The current project aimed to demonstrate an easily applicable method to overcome missing data to establish an unbiased association between urinary cadmium and hypertension in a large U.S. cohort of postmenopausal women sampled between 1993-1998.
Methods
Data were derived from the Women’s health Initiative. Filtered urinary cadmium was measured by Inductively Coupled Mass Spectrometry and normalized to urine creatinine (µg/g). Hypertension was defined as a systolic blood pressure (SBP) >140 mmHg or a diastolic BP (DBP) > 90 mmHg or use of antihypertensive medication. Estimated glomerular filtration rate (eGFR) was calculated using the CKD-EPI equation.
We estimated the odds of having hypertension in relation to log-transformed urinary cadmium adjusted for age, ethnicity/race, education, U.S. region, smoking, body mass index, and anemia (hemoglobin<12g/dL). Because of 76% missing data on eGFR, we performed analysis (1) unadjusted for eGFR, (2) adjusted for eGFR but restricted to a complete sample, and (3) adjusted for eGFR with imputed eGFR values. We performed multiple imputation (SAS 9.4. Proc MI) using a Monte-Carlo-Chain method (100 iterations) and estimated variance with Rubin’s formula.
Results
In 1460 women aged 63(±7) years, the mean urinary cadmium was 0.61(±46)µg/g and 44% had hypertension. Eight percent were current, 37% were past, and 54% were never smokers. The unadjusted odds ratio (OR) for prevalent hypertension was 0.84 (95% CI 0.73 – 0.96). The OR for model 1 was 0.92 (95% CI 0.78-1.07), whereas in model 2 restricted to complete sample on eGFR OR was 0.63 (95% CI 0.42-0.94). After imputation for eGFR (model 3), the OR was 0.91 (95% CI 0.77-1.07), which is similar to estimates obtained in model 1. Assumptions of missing out of randomness and normal distribution were met for the imputation.
Conclusion
Our example demonstrates that using a study sample without missing data may provide biased estimates of association, and imputation of important confounders and covariates can be used if missing is random.
Funding
- Other NIH Support