Influence of drugs and comorbidity on serum potassium in 15 000 consecutive hospital admissions
Henz, Samuel ; Maeder, Micha T. ; Huber, Stephanie ; Schmid, Michael ; Loher, Marcel ; Fehr, Thomas
In: Nephrology Dialysis Transplantation, 2008, vol. 23, no. 12, p. 3939-3945
Ajouter à la liste personnelle- Summary
- Background. Drug trials often exclude subjects with relevant comorbidity or comedication. Nevertheless, after approval, these drugs will be prescribed to a much broader collective. Our goal was to quantify the impact of drugs and comorbidity on serum potassium in unselected patients admitted to the hospital. Methods. This was a retrospective pharmacoepidemiologic study in 15 000 consecutive patients admitted to the medical department of the Kantonsspital St. Gallen, a 700-bed tertiary hospital in eastern Switzerland. Patients with ‘haemolytic' plasma and patients on dialysis or with an estimated glomerular filtration rate (GFR) <10 mL/min/1.73 m2 were excluded. For the remaining 14 146 patients, drug history on admission, age, sex, body weight, physical findings, comorbidity (ICD-10 diagnoses) and laboratory information (potassium and creatinine) were extracted from electronic sources. Results. Estimated GFR was the strongest predictor of serum potassium (P < 0.0001). Angiotensin-converting enzyme inhibitors, cyclosporine, loop diuretics and potassium-sparing diuretics all showed a significant effect modification with decreasing GFR (P < 0.001). Similarly, in patients with liver cirrhosis a significantly stronger effect on potassium was found for angiotensin receptor blockers, betablockers and loop diuretics (P < 0.01). Several significant drug-drug interactions were identified. Diabetes, male sex, older age, lower blood pressure and higher body weight were all independently associated with higher serum potassium levels (P < 0.001). The model explained 14% of the variation of serum potassium. Conclusions. The effects of various drugs on serum potassium are highly influenced by comorbidity and comedication. Although the presented model cannot be used to predict potassium in individual patients, we demonstrate that clinical databases could evolve as a powerful tool for industry-independent analysis of postmarketing drug safety