Secondary hyperparathyroidism (SHPT) is frequently observed in dialysis patients leading to costly to treat health complications. OBJECTIVES: To assess whether direct medical costs significantly vary with SHPT conditions. METHODS:Direct medical cost and resource use data were retrospectively collected from patient medical files and invoices of 75 dialysis patients over maximum period of 2 years prior to death in 3 dialysis centres in Belgium. Total collection period was segmented into two SHPT condition episodes defined by PTH values £ or > 300 pg/mL, using linear interpolation lines between subsequent PTH measurements. Costs were retrieved from hospital invoices, including start and stop date of each bill and cost category (hospitalisations, medications, dialysis, laboratory tests, honoraria, and other). To overcome problems of incomplete invoices, an invoice coverage factor was developed comparing available invoices with healthcare resource utilisation from medical files. Periods with insufficient coverage were discarded. Resource use and costs were attributed to each individual PTH condition episode. Statistical modelling: Because of heavily skewed cost data, incomplete data sets, the need to adjust cost results for time to death and centre effect, mixed models on non-transformed and log-transformed costs and General Linear Model (GLM) with a gamma response probability distribution and a loglink function were used. RESULTS: Results expressed as mean cost per day show that periods with elevated PTH-values lead to higher costs compared to normal periods (average cost-difference per day: approximately €50; p <0.05). Results will further be discussed regarding distribution of residuals, and hypothesis testing. To assess robustness of results, three sensitivity analyses were performed: analysis on subjects having experienced both PTH conditions before death; analysis over different observation periods using 6 months incremental time periods; cost depreciation factor. CONCLUSIONS: The statistical method allowed obtaining robust and valuable cost estimates for different SHPT conditions.