Prognostic factors in soft tissue sarcoma.

Despite major advances in the knowledge of soft tissue sarcoma (STS) during the last decades, no significant improvement in survival has been observed. Detailed data on the prognosis of STS are crucial in order to identify patients who might benefit from more aggressive treatment. Such data can be obtained from properly designed databases; however, the validation of data is crucial in order to obtain valid, reliable results. Furthermore, the majority of prognostic studies in STS have been limited by potential selection bias, low power, and biased estimates due to the statistical methods used, e.g., dichotomizing continuous variables, censoring competing events, as well as not adjusting for important confounders. The overall aim of this thesis was to investigate the prognosis of STS patients using data from the Aarhus Sarcoma Registry (ASR), covering western Denmark in the period from 1979 to 2008. In study I, we systematically validated data in the ASR and evaluated the validity, including completeness of patient registration and accuracy of data. In study II, we investigated the prognostic impact of patient-, tumor-, and treatment-related factors on local recurrence and disease-specific mortality. These were analyzed in a competing risk model in which continuous variables were included as cubic splines and possible confounders were selected based on directed acyclic graphs. In study III, we examined the impact of comorbidity on overall and disease-specific mortality. In study IV, we compared mortality in patients with abnormal biomarkers to those with normal values, assessed the significance of adjusting for comorbidity, as well as constructed a prognostic biomarker score. In study V, we described the relative mortality, i.e., the mortality in STS patients compared with the mortality in a general population, and compared relative and disease-specific estimates. The mortality in the general population was determined using an individually age- and sex-matched comparison cohort. All five studies were conducted in western Denmark within a population of approximately 2.5 million. Individual linkage between the ASR and national registries was made possible by the unique Danish civil registration number. The National Patient Registry and the LABKA research database were used to obtain data on comorbidity and biomarkers. In studies II to V we used a time-to-event-analysis approach that included cumulative incidence functions as well as crude and confounder adjusted Cox proportional hazard regression. In study I, we established that the overall validity of data in the ASR, after validation, was satisfactory and that the ASR included 85.3% of sarcoma patients from western Denmark between 1979 and 2008. In study II, we found a five-year local recurrence and disease-specific mortality of 16% and 24%, respectively. We excluded depth as a prognostic factor, and established that age, duration of symptoms, tumor size, anatomical and compartmental location, as well as radiotherapy were important prognostic factors for disease-specific mortality. In study III, we found that the level of comorbidity before or at diagnosis was an independent prognostic factor for both overall and disease-specific mortality, even after adjustment for age. In study IV, we showed that pretreatment levels of albumin, hemoglobin, and neutrophil to lymphocyte ratios were independently correlated with disease-specific mortality, and that adjusting for comorbidity was significant. In study V, we found five- and ten-year relative mortalities of 32.8% and 36.0%, respectively. The mortality in patients with low-grade STS was not significantly increased compared with the general population. The five- and ten-year disease-specific mortalities were underestimated by 3.1 and 1.9 percentage points compared to the relative mortality, respectively. We showed that relative mortality provided an accurate method to differentiate between cancer-specific and non-cancer-specific deaths. In conclusion, we showed that the ASR is a valid source of population-based data on STS. Improving the statistical methods used in prognostic studies of STS is important in order to obtain unbiased and reliable results. The level of comorbidity and bio-markers were important prognostic factors and should be used to identify high-risk STS patients who might benefit from more aggressive treatment.