Page 149 - Haematologica3

Experiences from the Swedish Myeloma Registry
A
B
or plasmacytoma. We estimated observed survival using the Kaplan-Meier method. When estimating relative survival (RS), rel- ative to the general Swedish population, we used the Ederer II method for expected survival. For observed survival (OS), we esti- mated Hazard Ratios (HR) using Cox proportional hazards regres- sion modeling. Also for RS, we estimated HR using proportional hazards regression, but in transformed time.20 Survival time was calculated from date of diagnosis to death or censoring. Patients were censored at the end of follow up in the study or loss to fol- low up. Age-standardized RS was calculated in each age group separately and then weighted together using weights from a stan- dard population, in this case, the International Cancer Survival Standard (ICSS) 1. We used a proportional hazard model of RS by year of diagnosis in all patients to estimate changes in survival over time. The survival analysis by year of diagnosis included both SMM and MM, and the date of diagnosis refers to the date of the primary diagnosis, whether this was SMM or MM. To evaluate the impact of the treating hospital, we estimated a proportional hazard model of RS by hospital type, in the categories “university hospital” or “not”, and hospitals reporting treatment on more or less than 10 patients per year. The survival analysis by treatment
Figure 2. Survival in active myeloma (MM) in the Swedish Myeloma Registry: observed (A) and relative (B) survival, by 10-year age cohorts. n: number.
response and by hospital type was carried out on symptomatic MM patients only (including patients who had developed symp- tomatic disease after SMM or plasmacytoma) with reported 1- year follow up, to enable comparison with statistics on treatment. When adjusting for ISS stage in regression analysis, we treated patients with missing values in the stage variable as a category within the ISS stage variable in order to not exclude the cohort of patients with missing data on ISS stage. P<0.05 was considered statistically significant. All data preparation and analysis were car- ried out using R statistical software.21
Results
A total of 5222 patients with plasma cell diseases diag- nosed in the period 2008-2015 had been reported to the Swedish Myeloma Registry as of December 31st 2016, with 97% coverage when compared with the Swedish Cancer Registry.
Clinical data at diagnosis were available for 4904 MM and SMM patients diagnosed in the period 2008-2015
haematologica | 2018; 103(3)
509