Cox Proportional Hazards (CPH) analysis is the standard for survival analysis in oncology. Recently, several machine learning (ML) techniques have been adapted for this task. Although they have shown ...to yield results at least as good as classical methods, they are often disregarded because of their lack of transparency and little to no explainability, which are key for their adoption in clinical settings. In this paper, we used data from the Netherlands Cancer Registry of 36,658 non-metastatic breast cancer patients to compare the performance of CPH with ML techniques (Random Survival Forests, Survival Support Vector Machines, and Extreme Gradient Boosting XGB) in predicting survival using the Formula: see text-index. We demonstrated that in our dataset, ML-based models can perform at least as good as the classical CPH regression (Formula: see text-index Formula: see text), and in the case of XGB even better (Formula: see text-index Formula: see text). Furthermore, we used Shapley Additive Explanation (SHAP) values to explain the models' predictions. We concluded that the difference in performance can be attributed to XGB's ability to model nonlinearities and complex interactions. We also investigated the impact of specific features on the models' predictions as well as their corresponding insights. Lastly, we showed that explainable ML can generate explicit knowledge of how models make their predictions, which is crucial in increasing the trust and adoption of innovative ML techniques in oncology and healthcare overall.
Breast cancer is the most commonly diagnosed cancer worldwide, and its burden has been rising over the past decades. In this article, we examine and describe the global burden of breast cancer in ...2020 and predictions for the year 2040.
Estimates of new female breast cancer cases and deaths in 2020 were abstracted from the GLOBOCAN database. Age-standardized incidence and mortality rates were calculated per 100,000 females by country, world region, and level of human development. Predicted cases and deaths were computed based on global demographic projections for the year 2040.
Over 2.3 million new cases and 685,000 deaths from breast cancer occurred in 2020. Large geographic variation across countries and world regions exists, with incidence rates ranging from <40 per 100,000 females in some Asian and African countries, to over 80 per 100,000 in Australia/New Zealand, Northern America, and parts of Europe. Smaller geographical variation was observed for mortality; however, transitioning countries continue to carry a disproportionate share of breast cancer deaths relative to transitioned countries. By 2040, the burden from breast cancer is predicted to increase to over 3 million new cases and 1 million deaths every year because of population growth and ageing alone.
Breast cancer is the most common cancer worldwide and continues to have a large impact on the global number of cancer deaths. Global efforts are needed to counteract its growing burden, especially in transitioning countries where incidence is rising rapidly, and mortality rates remain high.
•With over 2.3 million new cases and 685,000 deaths in 2020, breast cancer is the most commonly diagnosed cancer worldwide.•Most cases occur in transitioned countries yet transitioning countries have disproportionate share of breast cancer deaths.•The future burden of breast cancer is predicted to increase to over 3 million new cases and 1 million deaths in 2040.
...hospitals might have postponed diagnostic evaluation or have longer turnaround times for diagnostic evaluation because many hospital-based resources are being allocated to tackle COVID-19. ......national screening programmes for breast, colorectal, and cervical cancer are temporarily halted as of March 16, 2020, to alleviate the demand on the health-care system due to COVID-19. ...misconceptions were eliminated about a heightened risk of contracting COVID-19 in a health-care setting because of inadequate policies for infection control at the institutional level and resource constraints in the delivery of essential oncological care.
Summary Background Investigators of registry-based studies report improved survival for breast-conserving surgery plus radiotherapy compared with mastectomy in early breast cancer. As these studies ...did not present long-term overall and breast cancer-specific survival, the effect of breast-conserving surgery plus radiotherapy might be overestimated. In this study, we aimed to evaluate 10 year overall and breast cancer-specific survival after breast-conserving surgery plus radiotherapy compared with mastectomy in Dutch women with early breast cancer. Methods In this population-based study, we selected all women from the Netherlands Cancer Registry diagnosed with primary, invasive, stage T1–2, N0–1, M0 breast cancer between Jan 1, 2000, and Dec 31, 2004, given either breast-conserving surgery plus radiotherapy or mastectomy, irrespective of axillary staging or dissection or use of adjuvant systemic therapy. Primary outcomes were 10 year overall survival in the entire cohort and breast cancer-specific survival in a representative subcohort of patients diagnosed in 2003 with characteristics similar to the entire cohort. We estimated breast cancer-specific survival by calculating distant metastasis-free and relative survival for every tumour and nodal category. We did multivariable Cox proportional hazard analysis to estimate hazard ratios (HRs) for overall and distant metastasis-free survival. We estimated relative survival by calculating excess mortality ratios using life tables of the general population. We did multiple imputation to account for missing data. Findings Of the 37 207 patients included in this study, 21 734 (58%) received breast-conserving surgery plus radiotherapy and 15 473 (42%) received mastectomy. The 2003 representative subcohort consisted of 7552 (20%) patients, of whom 4647 (62%) received breast-conserving surgery plus radiotherapy and 2905 (38%) received mastectomy. For both unadjusted and adjusted analysis accounting for various confounding factors, breast-conserving surgery plus radiotherapy was significantly associated with improved 10 year overall survival in the whole cohort overall compared with mastectomy (HR 0·51 95% CI 0·49–0·53; p<0·0001; adjusted HR 0·81 0·78–0·85; p<0·0001), and this improvement remained significant for all subgroups of different T and N stages of breast cancer. After adjustment for confounding variables, breast-conserving surgery plus radiotherapy did not significantly improve 10 year distant metastasis-free survival in the 2003 cohort overall compared with mastectomy (adjusted HR 0·88 0·77–1·01; p=0·07), but did in the T1N0 subgroup (adjusted 0·74 0·58–0·94; p=0·014). Breast-conserving surgery plus radiotherapy did significantly improve 10 year relative survival in the 2003 cohort overall (adjusted 0·76 0·64–0·91; p=0·003) and in the T1N0 subgroup (adjusted 0·60 0·42–0·85; p=0·004) compared with mastectomy. Interpretation Adjusting for confounding variables, breast-conserving surgery plus radiotherapy showed improved 10 year overall and relative survival compared with mastectomy in early breast cancer, but 10 year distant metastasis-free survival was improved with breast-conserving surgery plus radiotherapy compared with mastectomy in the T1N0 subgroup only, indicating a possible role of confounding by severity. These results suggest that breast-conserving surgery plus radiotherapy is at least equivalent to mastectomy with respect to overall survival and may influence treatment decision making for patients with early breast cancer. Funding None.
Abstract Introduction We present a comprehensive overview of most recent European trends in population-based incidence of, mortality from and relative survival for patients with cancer since the mid ...1990s. Methods Data on incidence, mortality and 5-year relative survival from the mid 1990s to early 2000 for the cancers of the oral cavity and pharynx, oesophagus, stomach, colorectum, pancreas, larynx, lung, skin melanoma, breast, cervix, corpus uteri, ovary, prostate, testis, kidney, bladder, and Hodgkin’s disease were obtained from cancer registries from 21 European countries. Estimated annual percentages change in incidence and mortality were calculated. Survival trends were analyzed by calculating the relative difference in 5-year relative survival between 1990–1994 and 2000–2002 using data from EUROCARE-3 and -4. Results Trends in incidence were generally favorable in the more prosperous countries from Northern and Western Europe, except for obesity related cancers. Whereas incidence of and mortality from tobacco-related cancers decreased for males in Northern, Western and Southern Europe, they increased for both sexes in Central Europe and for females nearly everywhere in Europe. Survival rates generally improved, mostly due to better access to specialized diagnostics, staging and treatment. Marked effects of organised or opportunistic screening became visible for breast, prostate and melanoma in the wealthier countries. Mortality trends were generally favourable, except for smoking related cancers. Conclusion Cancer prevention and management in Europe is moving in the right direction. Survival increased and mortality decreased through the combination of earlier detection, better access to care and improved treatment. Still, cancer prevention efforts have much to attain, especially in the domain of female smoking prevalence and the emerging obesity epidemic.
The aim was to investigate whether pathologic complete response (pCR) in the breast is correlated with absence of axillary lymph node metastases at final pathology (ypN0) in patients treated with ...neoadjuvant systemic therapy (NST) for different breast cancer subtypes.
Pathologic complete response rates have improved on account of more effective systemic treatment regimens. Promising results in feasibility trials with percutaneous image-guided tissue sampling for the identification of breast pCR after NST raise the question whether breast surgery is a redundant procedure. Thereby, the need for axillary surgery should be reconsidered as well.
Patients diagnosed with cT1-3N0-1 breast cancer and treated with NST, followed by surgery between 2010 and 2016, were selected from the Netherlands Cancer Registry. Patients were compared according to the pathologic response of the primary tumor with associated pathologic axillary outcome. Multivariable analysis was performed to determine clinicopathological variables correlated with ypN0.
A total of 4084 patients were included for analyses, of whom 986 (24.1%) achieved breast pCR. In clinically node negative patients (cN0), 97.7% (432/442) with breast pCR had ypN0 compared with 71.6% (882/1232) without breast pCR (P < 0.001). In clinically node positive patients (cN1), 45.0% (245/544) with breast pCR had ypN0 compared with 9.4% (176/1866) without breast pCR (P < 0.001). The odds of ypN0 was decreased in case of clinical T3 stage (OR 0.59, 95% CI 0.40-0.87), cN1 (OR 0.03, 95% CI 0.02-0.04) and ER+HER2- subtype (OR 0.30, 95% CI 0.20-0.44), and increased in case of breast pCR (OR 4.53, 95% CI 3.27-6.28).
Breast pCR achieved after NST is strongly correlated with ypN0 in cN0 patients, especially in ER+HER2+, ER-HER2+, and triple negative subtypes. These results provide data to proceed with future clinical trials to investigate if axillary surgery can be safely omitted in these selected patients when image-guided tissue sampling identifies a breast pCR.
Oncological care was largely derailed due to the reprioritisation of health care services to handle the initial surge of COVID-19 patients adequately. Cancer screening programmes were no exception in ...this reprioritisation. They were temporarily halted in the Netherlands (1) to alleviate the pressure on health care services overwhelmed by the upsurge of COVID-19 patients, (2) to reallocate staff and personal protective equipment to support critical COVID-19 care, and (3) to mitigate the spread of COVID-19. Utilising data from the Netherlands Cancer Registry on provisional cancer diagnoses between 6 January 2020 and 4 October 2020, we assessed the impact of the temporary halt of national population screening programmes on the diagnosis of breast and colorectal cancer in the Netherlands. A dynamic harmonic regression model with ARIMA error components was applied to assess the observed versus expected number of cancer diagnoses per calendar week. Fewer diagnoses of breast and colorectal cancer were objectified amid the early stages of the initial COVID-19 outbreak in the Netherlands. This effect was most pronounced among the age groups eligible for cancer screening programmes, especially in breast cancer (age group 50-74 years). Encouragingly enough, the observed number of diagnoses ultimately reached and virtually remained at the level of the expected values. This finding, which emerged earlier in age groups not invited for cancer screening programmes, comes on account of the decreased demand for critical COVID-19 care since early April 2020, which, in turn, paved the way forward to resume screening programmes and a broad range of non-critical health care services, albeit with limited operating and workforce capacity. Collectively, transient changes in health-seeking behaviour, referral practices, and cancer screening programmes amid the early stages of the initial COVID-19 epidemic in the Netherlands conjointly acted as an accelerant for fewer breast and colorectal cancer diagnoses in age groups eligible for cancer screening programmes. Forthcoming research is warranted to assess whether the decreased diagnostic scrutiny of cancer during the COVID-19 pandemic resulted in stage migration and altered clinical management, as well as poorer outcomes.
Many countries had to suspend their colorectal cancer (CRC) screening programme as a result of the COVID-19 pandemic. This eventually may lead to postponed diagnoses of premalignant lesions and CRC, ...resulting in increased incidence or more advanced CRCs rates. This study aimed to assess the impact of the COVID-19 pandemic on incidence and stage distribution of CRCs in the Netherlands, by monitoring CRC diagnoses and stage distribution in the months before, during and after the first COVID-19 wave. Data on incidence and stage distribution of CRCs of individuals aged 55–75 years in 25 hospitals in the Netherlands were extracted from the Netherlands Cancer Registry. The observed incidence after the suspension (March 2020–December 2020) was compared to the expected incidence in the same period. In the period April to June 2020, we observed the largest decrease in the total incidence of CRC. We found that 48% of the decrease was due to stage I, 23% due to stage II, 23% due to stage III and 5% due to stage IV. After gradually resuming screening mid May 2020, we observed an increase in CRC diagnoses from July 2020 onwards. As of October 2020, the observed number of diagnoses was higher than the expected number. As the decrease was mainly limited to stage I CRCs, it seems that the temporary suspension of the CRC screening programme due to the COVID-19 pandemic will have a minimal long-term impact on stage distribution and CRC mortality.
•A decrease in CRC diagnoses was observed during the first COVID-19 wave.•This seems to be related to the suspension of the Dutch CRC screening programme.•The reduction was mainly limited to stage I CRCs.•A catch-up can be achieved by temporarily expanding colonoscopy capacity.•The impact of the temporary suspension of CRC screening seems to be limited.
Purpose
We performed a systematic review to document the spatial location of local recurrences (LR) after mastectomy.
Methods
A PubMed search was conducted in August 2019 for the following terms: ...breast Title/Abstract AND cancer Title/Abstract AND recurrence Title/Abstract AND mastectomy Title/Abstract. The search was filtered for English language. Exclusion criteria included studies that did not specify the LR location or studies reporting LR associated with inflammatory breast cancer, or other breast cancers such as phyllodes tumours, lymphoma or associated with sarcoma/angiosarcoma.
Results
A total of 3922 titles were identified, of which 21 publications were eligible for inclusion in the final analysis. A total of 6901 mastectomy patients were included (range 25–1694). The mean LR proportion was 3.5%. Among the total of 351 LR lesions, 81.8% were in the subcutaneous tissue and the skin, while 16% were pectoral muscle recurrences.
Conclusion
Local recurrences are mostly located within the subcutaneous tissue and the skin, assumed to result from unrecognized/subclinical tumour foci left behind after mastectomy, surgical implantation of tumour cells in the wound/scar and/or tumour emboli within the subcutaneous lymphatics. Pectoral muscle recurrences are less frequent and may be attributed to residual disease along the posterior surgical margin and/or lymphatic involvement.
Abstract
Background
Risk-prediction tools allow classifying individuals into risk groups based on risk thresholds. Such risk categorization is often used to inform screening schemes by offering ...screening only to individuals at increased risk of harmful events. Adding information concerning an individual’s risk development over time would allow assessing not just who to screen but also when to screen. This paper illustrates the value of personalised, time-dependent risk predictions to optimize risk-based screening schemes.
Methods
In a simulation analysis, two different time-dependent risk-based screening approaches are compared to another risk-based, but time-independent approach regarding their impact on screening efficiency. For this purpose, 81 scenarios featuring 5000 patients with five consecutive annual risk estimations for a hypothetical disease
D
are simulated, using different parameters to model disease progression and risk distribution. This simulation analysis is validated using a real-world clinical case study based on German breast cancer patients and the INFLUENCE-nomogram for locoregional breast cancer recurrence.
Results
If individual risk estimations were used to personalise screening for a disease D aiming at detecting a 90% of curable cases, more than 20% of screening examinations could be avoided relative to a conventional uninformed approach, depending on the simulated scenario. Whereas an individual but time-independent approach is associated with acceptable saving potentials in case of a relatively homogenous risk distribution, the time-dependent approaches are superior when the complexity of a scenario increases. With slowly progressing diseases, risk-accumulation over time needs to be considered to achieve the highest screening efficiency on population level, for rapidly progressing diseases, an interval-specific approach is superior. The possible benefits of time-dependent risk-based screening were confirmed in the real-world clinical case study.
Conclusions
Appropriate approaches to use time-dependent risk predictions may considerably enhance screening efficiency on individual and population level. Therefore, predicting risk development over time should be supported by future prediction tools and be incorporated in decision algorithms.
PDF
