Predicting second breast cancer among women with primary breast cancer using machine learning algorithms, a population-based observational study.

Breast cancer survivors often experience recurrence or a second primary cancer. We developed an automated approach to predict the occurrence of any second breast cancer (SBC) using patient-level data and explored the generalizability of the models with an external validation data source. Breast cancer patients from the cancer registry of Zurich, Zug, Schaffhausen, Schwyz (N = 3213; training dataset) and the cancer registry of Ticino (N = 1073; external validation dataset), diagnosed between 2010 and 2018, were used for model training and validation, respectively. Machine learning (ML) methods, namely a feed-forward neural network (ANN), logistic regression, and extreme gradient boosting (XGB) were employed for classification. The best-performing model was selected based on the receiver operating characteristic (ROC) curve. Key characteristics contributing to a high SBC risk were identified. SBC was diagnosed in 6% of all cases. The most important features for SBC prediction were age at incidence, year of birth, stage, and extent of the pathological primary tumor. The ANN model had the highest area under the ROC curve with 0.78 (95% confidence interval [CI] 0.750.82) in the training data and 0.70 (95% CI 0.61-0.79) in the external validation data. Investigating the generalizability of different ML algorithms, we found that the ANN generalized better than the other models on the external validation data. This research is a first step towards the development of an automated tool that could assist clinicians in the identification of women at high risk of developing an SBC and potentially preventing it.

Trends in prostate cancer incidence between 1996 and 2013 in two Swiss regions by age, grade, and T-stage.

PURPOSE: To investigate differences in prostate cancer incidence between two distinct Swiss regions from 1996 to 2013 stratified by age group, grade, and T-stage. METHODS: The dataset included 17,495 men living in Zurich and 3,505 men living in Ticino, diagnosed with prostate cancer between 1996 and 2013. We computed age-standardized incidence rates per 100,000 person-years using the European Standard Population. Trends were assessed using JoinPoint regression analysis Software. RESULTS: Age-standardized incidence rates were generally higher in Zurich compared to Ticino but the difference decreased over time. Incidence rates increased significantly up to 2002 in Zurich and 2007 in Ticino and then decreased. A statistically significant increase was observed for men aged < 65 years, for grade 3 tumors, and for T-stage 2 and 3 tumors. The largest decrease was seen for grade 1 tumors. Furthermore, the incidence of tumors of unknown grade or T-stage decreased significantly in both regions. CONCLUSIONS: The trends in prostate cancer incidence rates were similar in both regions, although on a higher level in Zurich compared to Ticino. However, the difference decreased over time. The distribution of T-stage and grade did not explain the difference in incidence rates. Different use of opportunistic screening may play a role.

Quality indicators of clinical cancer care for prostate cancer: a population-based study in southern Switzerland.

BACKGROUND: Quality of cancer care (QoCC) has become an important item for providers, regulators and purchasers of care worldwide. Aim of this study is to present the results of some evidence-based quality indicators (QI) for prostate cancer (PC) at the population-based level and to compare the outcomes with data available in the literature. METHODS: The study included all PC diagnosed on a three years period analysis (01.01.2011-31.12.2013) in the population of Canton Ticino (Southern Switzerland) extracted from the Ticino Cancer Registry database. 13 QI, approved through the validated Delphi methodology, were calculated using the "available case" approach: 2 for diagnosis, 4 for pathology, 6 for treatment and 1 for outcome. The selection of the computed QI was based on the availability of medical documentation. QI are presented as proportion (%) with the corresponding 95% confidence interval. RESULTS: 700 PC were detected during the three-year period 2011-2013: 78.3% of them were diagnosed through a prostatic biopsy and for 72.5% 8 or more biopsy cores were taken. 46.5% of the low risk PC patients underwent active surveillance, while 69.2% of high risk PC underwent a radical treatment (radical prostatectomy, radiotherapy or brachytherapy) and 73.5% of patients with metastatic PC were treated with hormonal therapy. The overall 30-day postoperative mortality was 0.5%. CONCLUSIONS: Results emerging from this study on the QoCC for PC in Canton Ticino are encouraging: the choice of treatment modalities seems to respect the international guidelines and our results are comparable to the scarce number of available international studies. Additional national and international standardisation of the QI and further QI population-based studies are needed in order to get a real picture of the PC diagnostic-therapeutic process progress through the definition of thresholds of minimal standard of care.

Impact of subtypes and comorbidities on breast cancer relapse and survival in population-based studies.

OBJECTIVE: To study the impact of subtypes and comorbidities on breast cancer (BC) relapse and survival in the heterogeneous patients of the real world. METHODS: We identified patients diagnosed with BC between January 2003 and December 2005 from six population-based Swiss cancer registries. Clinicopathologic data was completed with information on locoregional and distant relapse and date and cause of death for over 10-years. We approximated BC subtypes using grade and the immunohistochemical panel for oestrogen, progesterone and human epidermal growth factor 2 (HER2) receptor status. We studied factors affecting relapse and survival. RESULTS: Luminal A-like subtype represented 46% of all newly diagnosed BC (N = 1831), followed by luminal B-like (N = 1504, 38%), triple negative (N = 436, 11%) and HER2 enriched (N = 204, 5%). We observed regional disparities in subtype prevalence that contribute to explain regional differences in survival formerly described. Disease relapse and BC specific mortality differed by subtype and were lower for luminal A like tumours than for other subtypes for any stage at diagnosis. After a median follow-up of 10.9 years, 1311 (33%) had died, half of them 647 (16%) due to another disease, showing the importance of comorbidities. Omission of systemic therapies in selected patients was not associated with poorer BC specific survival, BC subtype and life expectancy playing a role. CONCLUSIONS: Information on tumour subtype is necessary for an adequate interpretation of population-based BC studies. Measures of comorbidity or frailty help in the evaluation of quality of care in the highly heterogeneous patients of the real world.

Quality indicators of colorectal cancer care in southern Switzerland: results from a population-based study.

AIM OF THE STUDY: Assessing the quality of cancer care (QoCC) has become increasingly relevant to providers, regulators and purchasers of healthcare worldwide. The aim of this study was to assess adherence to validated quality indicators (QIs) for colorectal cancer (CRC) in a population-based setting, and to compare results with the available literature. METHODS: All colorectal cancers diagnosed between 1 January 2011 and 31 December 2012 were identified from the files of the population-based Ticino Cancer Registry, southern Switzerland. We computed 12 core QIs, approved by use of the validated Delphi methodology and for which all the necessary medical documentation was available or only minor data collection was still needed to complete the analysis: three for diagnosis, two for pathology and seven for treatment (surgery, oncology and radiotherapy). Each QI was analysed as proportion (%) with 95% confidence interval, following the approach "available case analysis". RESULTS: A total of 474 colorectal cancers were identified: 86.9% patients were diagnosed after they reported symptoms, 90.2% had preoperative colonoscopy, 8.7% underwent emergency surgery, 97.2% had a surgical resection with tumour-free margins, and for 86.6% at least 12 lymph nodes were examined. The overall 30-day postoperative mortality was 3.6% and 66.7% of locally advanced rectal cancers benefited of neoadjuvant radiotherapy ± chemotherapy. CONCLUSIONS: Our study showed the feasibility of assessing QoCC using Cancer Registry population-based data. Results according to the clinical domain of pathology, surgery, oncology and radio-oncology in southern Switzerland are generally positive and encouraging, sometimes more favourable in comparison with other international studies, except the very low proportion of patients with a diagnosis based on opportunistic screening (8.6%). Considering the lack in the literature of population-based studies, further national and international reports are urgently needed for comparative analysis as well as standardisation of QI definition is absolutely necessary for inter-regional comparative goals.

Quality indicators of clinical cancer care (QC3) in colorectal cancer.

OBJECTIVES: Assessing the quality of cancer care (QoCC) has become increasingly important to providers, regulators and purchasers of care worldwide. The aim of this study was to develop evidence-based quality indicators (QIs) for colorectal cancer (CRC) to be applied in a population-based setting. DESIGN: A comprehensive evidence-based literature search was performed to identify the initial list of QIs, which were then selected and developed using a two-step-modified Delphi process involving two multidisciplinary expert panels with expertise in CRC care, quality of care and epidemiology. SETTING: The QIs of the clinical cancer care (QC3) population-based project, which involves all the public and private hospitals and clinics present on the territory of Canton Ticino (South Switzerland). PARTICIPANTS: Ticino Cancer Registry, The Colorectal Cancer Working Group (CRC-WG) and the external academic Advisory Board (AB). MAIN OUTCOME MEASURES: Set of QIs which encompass the whole diagnostic-treatment process of CRC. RESULTS: Of the 149 QIs that emerged from 181 sources of literature, 104 were selected during the in-person meeting of CRC-WG. During the Delphi process, CRC-WG shortened the list to 89 QI. AB finally validated 27 QIs according to the phase of care: diagnosis (N=6), pathology (N=3), treatment (N=16) and outcome (N=2). CONCLUSIONS: Using the validated Delphi methodology, including a literature review of the evidence and integration of expert opinions from local clinicians and international experts, we were able to develop a list of QIs to assess QoCC for CRC. This will hopefully guarantee feasibility of data retrieval, as well as acceptance and translation of QIs into the daily clinical practice to improve QoCC. Moreover, evidence-based selected QIs allow one to assess immediate changes and improvements in the diagnostic-therapeutic process that could be translated into a short-term benefit for patients with a possible gain both in overall and disease-free survival.