Finally, we conducted separate regression analyses predicting cigarettes per day (CPD) and mFTQ score (each square-root transformed to reduce skewness) from nicotine metabolism as measured by the NMR. The models also included the potential confounding variables of age, gender, race/ethnicity, and duration of smoking. We then performed backward selection using the Akaike information criterion (AIC) to drop non-contributing variables from the models; the resulting models are reported as the adjusted models. Because the mFTQ also contains information about smoking rate (CPD), we also re-ran the equation predicting mFTQ while removing CPD from the mFTQ scoring.
