Page 226 - Haematologica April 2020

S. Thorsteinsdottir et al.
prostate cancer.14-16 To our knowledge, only three studies have assessed the effect of fractures on survival in MM. In the largest study so far, a total of 513 MM patients diag- nosed from 1998 to 2000 and enrolled in a clinical trial were included retrospectively; all patients had Durie- Salmon stage III, bone lesions, and a median age of 62 years. No significant difference in survival was observed between MM patients who developed a fracture during follow up as compared to patients that did not develop a fracture.16 The second study was a small case-control study in which overall survival was found to be inferior in MM patients with pathologic fractures (n=24) compared to patients with no pathologic fractures (n=25).17 However, in a large study from the UK including patients admitted with plasma cell dyscrasias (MM, plasmacytoma, and plasma cell leukemia), both previous and subsequent frac- tures were found to be a risk factor for worse survival after first admission for plasma cell dyscrasia.18
The aim of our study was to evaluate the impact of frac- tures and certain subtypes of fractures on survival after MM diagnosis in a whole population using real-world data on all MM patients diagnosed during a 23-year period in Sweden. We also aimed to compare the effect of frac- tures on survival in MM before and after the introduction of novel treatment agents that have greatly improved sur- vival in the MM patient population.19-21
Methods
The Swedish Cancer Registry is a centralized, nationwide reg- istry containing information on patients who have been diagnosed with a malignant disorder in Sweden since January 1st, 1958. Every physician and pathologist is obliged by law to report each case of cancer to the registry. The Swedish Cancer Registry contains infor- mation on sex, date of birth, date of diagnosis, and histopathologic diagnosis. In a validation study from 2007, the completeness and diagnostic accuracy of the Swedish Cancer Registry was found to be over 93% for MM patients.22 Patients diagnosed with MM in the period from January 1990 to December 2013 were identified from the Swedish Cancer Registry. Information on date of birth, diagnosis, and death were collected. Information on clinical stage and laboratory results were not available for the patients. Each person in Sweden receives a unique personal identification num- ber that is used to index all major health registers, making it pos- sible to link information in the registries. Information on fractures, using ICD-10 and ICD-9 diagnostic codes (see Online Supplementary Appendix) was obtained from the Swedish Patient Registry, that contains inpatient data from 1987 as well as infor- mation on outpatient visits from 2000. Fractures from ten years before MM diagnosis and then afterwards were included (to adjust for previous fractures before MM diagnosis). Information on date of death was gathered from the Swedish Cause of Death Registry. End of follow up was December 31st, 2013. The study period was divided into two calendar periods: 1990-1999 and 2000-2013, respectively before and after the introduction of novel treatment options for MM in Sweden that have been shown to improve survival.19,20 A total of 333 patients were excluded from analysis because of unknown age; 68 of these were patients with a fracture.
A Cox regression model was used to estimate the effect of a fracture at diagnosis (defined as 30 days before or after MM diag- nosis) on survival after the time of MM diagnosis. All results were adjusted for age, sex, previous fracture and year of MM diagnosis. Results were presented as hazard ratios (HR) with 95% confi-
dence intervals (CI). P<0.05 was considered significant and all sta- tistical analysis was performed in R version 3.5.2.23
A landmark analysis was performed by selecting a subset of MM patients alive at six months after MM diagnosis and stratify- ing patients according to whether they had developed a fracture from the day of MM diagnosis (day 0) until six months after diag- nosis or not. Fractures before the day of MM diagnosis were not included in the landmark analysis to avoid any immortal time bias.24 A Cox regression model was used to assess the association of fracture status at six months and survival, and a Kaplan-Meier graph was generated to visualize the difference in survival between the two groups.
A Cox regression model was used with fracture as a time- dependent variable to assess the association of fracture and sur- vival after MM diagnosis (from the day of MM diagnosis). The effect of fracture was assessed for any fracture or a specific sub- type of fracture. Specifically, we assessed pathologic (all fractures registered as pathologic fractures), vertebral (both pathologic and others), hip, femoral, humerus, forearm, rib, pelvis, and ankle frac- tures. Either first fracture or the first subtype of fracture was used in the analysis. The effect of fracture was analyzed for males and females separately for two age groups (<70 years and ≥70 years old) and for the two calendar periods. Results were adjusted for age, sex, time of diagnosis, and previous fractures. To compare the difference in the association of fracture and survival between the two calendar periods, the two age groups, and sexes, the interac- tion effect of the variable and fracture was assessed in a Cox regression model. Because outpatient visits were included from year 2000, we performed an additional analysis with only inpa- tient diagnoses of fractures when comparing the calendar periods. Furthermore, we assessed the association of fractures and death after MM diagnosis for two calendar periods after year 2000: 2000-2006 and 2007-2013.
As an additional analysis, four controls, matched by gender, year of birth, and county of residence, were chosen randomly from the Swedish Register of Total Population. All controls were alive and free of MM at the time of MM diagnosis for the corre- sponding patient. Cox regression model was used with fracture as a time-dependent variable from MM diagnosis in the correspon- ding case to assess the association of fracture and survival in the matched controls. The effect of fracture was assessed for any frac- ture or a specific subtype of fracture (vertebral, femoral, humerus, rib, or ankle fracture).
Results
A total of 14,013 patients were diagnosed with MM in the period from 1st of January 1990 to 31st of December 2013. The median age was 72 years (range 20-99 years) and 54.9% were males (Table 1). A total of 4,146 (29.6%) patients developed a fracture including fractures that occurred a year before MM diagnosis and thereafter, with a sharp rise in fracture diagnoses around the time of MM diagnosis (Figure 1). Overall, 3,235 (23.1%) patients were diagnosed with a fracture at the same day or after MM diagnosis (Figure 1). A similar proportion of MM patients developed a fracture in the two calendar periods 1990- 1999 and 2000-2013 (22.2 and 23.7%, respectively) (Table 1). During the period of follow up, 10,731 (76.6%) MM patients died, of whom 2,520 (77.9%) were patients with fractures. Median follow-up time from diagnosis to death or end of follow up was 2.4 years (range 1 day-23.8 years). Median overall survival for all MM patients was 3.0 years for the whole study period and improved from 2.6 years
1068
haematologica | 2020; 105(4)