BACKGROUND Secondary hyperparathyroidism (SHPT) is a complex complication in chronic kidney disease (CKD) and is associated with an increased risk of cardiovascular disease (CVD), fracture and mortality in both dialysis and non-dialysis (ND) patients. Randomized clinical trials (RCTs) are often shorter than one year and focused on the effects on biomarkers of mineral metabolism. Only one published health economic model on SHPT treatment in ND CKD patients could be identified.

OBJECTIVES This study developed a cost-effectiveness model framework for assessing SHPT treatment in ND CKD patients by linking clinical surrogates to hard outcomes.

METHODS To capture the complexity of CKD disease progression, a Markov multi-state microsimulation model was developed. Patients’ starting biomarker levels, sampled individually for each patient from a representative distribution dependent on the patients’ CKD stage, are altered by treatment. In turn, these affect the CKD disease pathway; patients start in CKD Stage-3 or 4 and progress to dialysis, transplant, or death. Simulating the individual heterogeneity in biomarker levels makes it possible to model disease progression as a function of the proportion of patients reaching clinically significant thresholds. In each one-year cycle, patients can progress in CKD, experience CVD and fracture events and are at risk of hypercalcemia as an adverse event from treatment. The best available data on the linkage between biomarkers and hard endpoints were collected from literature reviews conducted during model development.

CONCLUSION Modeling SHPT treatment in ND CKD patients poses methodological issues that previously published models did not fully tackle. The absence of hard endpoints in RCTs calls for a health economic model to link the clinical evidence available to tangible health outcomes. Our model framework proposes a method for estimating cost-effectiveness as a function of treatment effects, patient-level heterogeneity, and the intermediate link between surrogate biomarkers and hard endpoints.