Primary central nervous system (CNS) tumours are heterogeneous, with different treatment pathways and prognoses depending on their histological and molecular classification. Due to their anatomical location, all CNS tumours, regardless of malignancy, can be debilitating. We used vital statistics linked to Canadian Cancer Registry data to estimate the age-standardized incidence rates (ASIR), Kaplan–Meier survival rates (SR), and limited-duration prevalence proportions (PP) of 25 histology-specific CNS tumour groups that were classified based on site and histology. During 2010–2017, 45,115 patients were diagnosed with 47,085 primary CNS tumours, of which 19.0% were unclassified. The average annual ASIR was 21.48/100,000 person-years and did not vary by sex. The ASIR increased with age, particularly for meningioma, unclassified tumours, and glioblastoma. The eight-year PP was 102.1/100,000 persons (index date 1 January 2018). The most common histology was meningioma (ASIR: 5.19; PP: 31.6). The overall five-year SR among 51,310 patients diagnosed during 2008–2017 was 57.2% (95% CI: 56.8–57.7%). SRs varied by tumour behaviour, histology, and patient age, with the lowest SR among glioblastoma patients (5-year SRs ranged from 1.3–25.7%). For non-malignant tumours, the 5-year SRs ranged from 37.4–100%. We provide the most up-to-date histology-specific surveillance estimates for primary CNS tumours in Canada.
Canada’s decentralized healthcare system may lead to regional disparities in survival among Canadians diagnosed with central nervous system (CNS) tumours. We identified 50,670 patients diagnosed with a first-ever primary CNS tumour between 2008 and 2017 with follow-up until 31 December 2017. We selected the four highest incidence histologies and used proportional hazard regression to estimate hazard ratios (HRs) for five regions (British Columbia, Prairie Provinces, Ontario, Atlantic Provinces and the Territories), adjusting for sex, tumour behaviour and patient age. Ontario had the best survival profile for all histologies investigated. The Atlantic Provinces had the highest HR for glioblastoma (HR = 1.26, 95% CI: 1.18–1.35) and malignant glioma not otherwise specified (NOS) (Overall: HR = 1.87, 95% CI:1.43–2.43; Pediatric population: HR = 2.86, 95% CI: 1.28–6.39). For meningioma, the Territories had the highest HR (HR = 2.44, 95% CI: 1.09–5.45) followed by the Prairie Provinces (HR = 1.52, 95% CI: 1.38–1.67). For malignant unclassified tumours, the highest HRs were in British Columbia (HR = 1.45, 95% CI: 1.22–1.71) and the Atlantic Provinces (HR = 1.40, 95% CI: 1.13–1.74). There are regional differences in the survival of CNS patients at the population level for all four specific histological types of CNS tumours investigated. Factors contributing to these observed regional survival differences are unknown and warrant further investigation.
Rare cancers pose significant diagnostic challenges, leading to more tests and higher healthcare expenditures (HEs). Understanding the financial implications of diagnosing rare cancers is crucial, particularly in Canada, where overall HEs are high (12% of the GDP in 2023). We investigated the pre-diagnostic hospital–based HE for patients with malignant central nervous system (CNS) tumours and compared it to patients with common cancers across Canadian provinces, using in-patient and ambulatory care data (2010–2014) from the Canadian Institute for Health Information. Pre-diagnostic HE was calculated as the change in total HE (in-patient and out-patient) during the 12 months before diagnosis, calculated as the HE within this period minus the average annual HE estimated over the two preceding years. Comparison groups included pediatric patients diagnosed with leukemia and patients aged over 15 diagnosed with colorectal cancer and lung cancer. We used quantile regression to estimate the adjusted effect of diagnosis with a CNS tumour on pre-diagnostic HE. The results indicated that HE for CNS patients was higher compared to those with common cancers. The top three factors contributing to HE variation were encounter type (in-patient/out-patient), province (Alberta/Ontario), and comorbidities (yes/no). Further investigation is warranted to understand the drivers of the cost differences.
Health care in Canada is managed by provinces/territories. We investigated regional differences in survival among Canadians diagnosed with central nervous system (CNS) tumours. We identified 50,670 patients diagnosed with a first-ever primary CNS tumour between 2008 and 2017 with follow-up until December 31, 2017. We selected the four highest incidence histologies and used proportional hazards regression to estimate hazard ratios (HRs) for five regions (British Columbia, Prairie provinces, Ontario, Atlantic provinces and the Territories), adjusting for sex, tumour behaviour and patient age. Ontario had the best survival profile for all histologies investigated. The Atlantic provinces had the highest HR for glioblastoma (HR=1.26, 95% CI:1.18-1.35) and malignant glioma not otherwise specified (NOS) (Overall: HR=1.87, 95% CI:1.43-2.43; Pediatric population: HR=2.86, 95% CI:1.28-6.39). For meningioma, the Territories had the highest HR (HR=2.44, 95% CI:1.09-5.45) followed by the Prairie provinces (HR=1.52, 95% CI:1.38-1.67). For malignant unclassified tumours, the highest HRs were in British Columbia (HR=1.45, 95% CI:1.22-1.71) and the Atlantic provinces (HR=1.40, 95% CI:1.13-1.74). There are regional differences in the survival of CNS patients at the population level for all four specific histological types of CNS tumours investigated. Factors contributing to these observed regional survival differences are unknown and warrant further investigation.
Purpose. Machine learning(ML) has revolutionized data processing and analysis, with applications in health showing great promise. However, ML poses privacy risks, as models may reveal information about their training data. Developing frameworks to assess/mitigate privacy risks is essential, particularly for health data custodians responsible for adhering to ethical and legal standards in data use. In September 2022, we conducted a systematic review/meta-analysis to estimate the relative effects of factors hypothesized to contribute to ML privacy risk, focusing on membership inference attacks (MIA). Methods. Papers were screened for relevance to MIA, and selected for the meta-analysis if they contained attack performance(AP) metrics for attacks on models trained on numeric data. Random effects regression was used to estimate the adjusted average change in AP by model type, generalization gap and the density of training data in each region of input space (partitioned density). Residual sum of squares was used to determine the importance of variables on AP. Results. The systematic review and meta-analysis included 115 and 42 papers, respectively, comprising 1,910 experiments. The average AP ranged from 61.0% (95%CI:60.0%-63.0%; AUC)-74.0% (95%CI:72.0%-76.0%; recall). Higher partitioned density was inversely associated with AP for all model architectures, with the largest effect on decision trees. Higher generalization gap was linked to increased AP, predominantly affecting neural networks. Partitioned density was a better predictor of AP than generalization gap for most architectures. Conclusions. This is the first quantitative synthesis of MIA experiments, that highlights the effect of dataset composition on AP, particularly on decision trees, which are commonly used in health.