In clinically aggressive diseases, patients experience pathophysiological changes that often alter concentrations of lipids and environmental lipophilic factors; such changes are related to disease signs and symptoms. The aim of the study was to compare the effects of correcting for total serum lipids (TSL) and other clinical factors on the odds of mutations in the K-ras oncogene by organochlorine compounds (OCs), in logistic models, in 103 patients with exocrine pancreatic cancer (EPC) using a causal directed acyclic graph (DAG) framework. Results and likelihood of bias were discussed in the light of possible causal scenarios. The odds of K-ras mutated EPC was associated with some TSL-corrected OCs, including p,p′-DDT (p-value: 0.008) and polychlorinated biphenyl 138 (p-trend: 0.024). When OCs were not corrected by TSL, the OR of a K-ras mutation was significant for p,p′-DDT (p-trend: 0.035). Additionally adjusting for cholestatic syndrome increased the ORs of TSL-corrected OCs. When models were adjusted by the interval from first symptom to blood extraction (ISE), the ORs increased for both TSL-corrected and uncorrected OCs. Models with TSL-corrected OCs and adjusted for cholestatic syndrome or ISE yielded the highest ORs. We show that DAGs clarify the covariates necessary to minimize bias, and demonstrate scenarios under which adjustment for TSL-corrected OCs and failure to adjust for symptoms or ISE may induce bias. Models with TSL-uncorrected OCs may be biased too, and adjusting by symptoms or ISE may not control such biases. Our findings may have implications as well for studying environmental causes of other clinically aggressive diseases.