PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • Study design and data sources: This population-based cancer-registry study was conducted in the BCR. The BCR routinely collects information regarding new cancer diagnoses in Belgium including patient characteristics (such as age at diagnosis, region of residence) and tumour characteristics (such as date of diagnosis, cancer topography and morphology) on a national level since the incidence year 2004. In Belgium, the specific Health Law of December 13th 2006 [18] has made cancer registration compulsory for the pathological anatomy laboratories and for the oncological care programs. Moreover, the reimbursement of multidisciplinary oncological consultation for each reported case of cancer acts as an incentive for the cancer registration. This provides high quality data and high completeness of BCR database, which is estimated to be more than 95% complete using cross-validation techniques. [19] The remaining incompleteness is most likely due to elderly patients with a very poor prognosis at diagnosis and outpatients with a clinical diagnosis only. Based on the aforementioned law, the BCR has also the legal authorization to use the National Social Security Number (NSSN) as a unique patient identifier. This NSSN allows the BCR to deterministically link its data with data from the Crossroads Bank for Social Security (CBSS) to retrieve information on vital status. The NSSN can also be used to deterministically link the BCR data with reimbursement data from the health insurance companies, provided by the Intermutualistic Agency (IMA). As health insurance is mandatory in Belgium, administrative, medical claims data from IMA are available for almost the entire Belgian population. The IMA database contains information regarding all reimbursed medications and medical acts. Since statins are reimbursed drugs that are only available on prescription in Belgium, all dispensed statin prescriptions can be identified within IMA data. In addition, for more than 90% of the BCR population, [19] death certificates provided by the regional administrations can be used to retrieve causes of death using a linkage based on several variables (such as date of birth, date of death, sex, residence at time of death). The linkage of the BCR-, the IMA-, the CBSS- and the regional administrations’ databases thus provides information on all patients diagnosed with cancers in Belgium, on their vital status, their date and cause of death, the cancer treatments they received, and their medication use. Specific ethical approvals were not required because the analyses were conducted within the legal framework of the BCR, using only anonymized data and without any access to private information on individual patients. All patients with an OC (ICD-10 code C56) diagnosed between 2004 and 2012 were identified from the BCR. Patients with borderline malignant tumours or with non-epithelial cancers were excluded. Patients with another invasive cancer diagnosis (apart from non-melanoma skin cancer) prior to their OC diagnosis were also excluded. Additionally, patients who died within six months after OC diagnosis were excluded as these patients were mainly diagnosed in advance stage disease. It seems unlikely that short term postdiagnostic medication usage could influence their survival. Therefore, the beginning of the follow-up time was set to 6 months after diagnosis. The other exclusion criteria were related to missing data or absence of follow-up data, thus, patients were also excluded if they did not have an available NSSN or available IMA data, if they died on the day of diagnosis or if they were lost to follow-up at diagnosis. Exclusions based on these criteria are expected to be rare in our study population because in the general BCR database, only 1.5% and 3.4% of patients don’t have a NSSN or IMA data available, only 0.16% of patients died on the day of diagnosis and 1.2% of patients are lost to follow-up at diagnosis. [20] Patients were followed from six months after diagnosis to a maximum of 3 years after their OC diagnosis (or to the end of follow-up on July 1st, 2015). Use of statins was identified from IMA data using the Anatomical Therapeutic Chemical (ATC) codes ‘C10AA’. To avoid immortal time bias, [21] statin use was modelled as a time varying covariate, i.e. patients were initially considered as nonusers and then became users after their first statin prescription. As recommended in studies of medication use and cancer survival, [22] a lag period of 6 months was applied to exclude prescriptions that fall within the 6 months period prior to death because these may be affected by palliative treatments. Statins were also examined by each statin type and by subgroups of statins according to solubility type, i.e. lipophilic agents (simvastatin and fluvastatin) and hydrophilic agents (atorvastatin, pravastatin and rosuvastatin). Statin exposure was further investigated in a dose-response analysis by counting the total daily defined doses (DDD) of statins that the patient received during the follow-up. After their first statin prescription, patients were considered as light users and they became heavy users as soon as they cumulated at least 365/2 DDDs. The available covariates were age at diagnosis, year of cancer diagnosis, histological grade, combined stage (according to the TNM Classification of Malignant Tumours, 7th edition), cancer histologic subtype, and cancer treatments (including surgery, chemotherapy, and radiotherapy in the 9 months after diagnosis). Cardiovascular comorbidities and diabetes were estimated using a previously described methodology based on specific drugs prescriptions in the year before the diagnosis. [23] Descriptive statistics were used to characterize patients at the time of diagnosis. These analyses were conducted within subgroups of statin users and nonusers in order to compare characteristics in these two groups. Patient- and tumor characteristics in each group are reported as frequencies and percentages. Age was the only variable considered both in categories and as a continuous variable. Time-dependent Cox regression models were used to calculate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for postdiagnostic statin users compared to nonusers and risk of overall mortality within 3 years after diagnosis. Analyses were adjusted for the following potential confounders: age in categories (≤ 49 years, 50–74 years, ≥ 75 years), year of diagnosis (in 3-year bands: 2004 to 2006, 2007 to 2009 and 2010 to 2012), cancer stage (I, II, III, IV), cancer treatments (neoadjuvant chemotherapy, adjuvant chemotherapy, chemotherapy only, surgery only and no treatment), cardiovascular comorbidities (yes or no) and diabetes (yes or no). Subgroup analyses were conducted by age categories, year of diagnosis, cancer stage, histologic subtype, cancer treatment and prediagnostic statin use. Tests for interactions were performed using interaction terms within Cox regression models. A sensitivity analysis that controls for immortal time bias without requiring time varying covariate was conducted. This simplified analysis used Cox regression to compare statin users to statin nonusers in the first 6 months after diagnosis. After the exclusion of deceased patients having missing death certificates (n = 521, 10%), analyses were also performed for ovarian cancer specific-mortality (S1 and S2 Tables and S1 Fig). The statistical significance level was set to P<0.05. All analyses were performed using the statistical software SAS Enterprise Guide 5.1.
rdf:type