Maximum utilization of the life table method in analyzing survival
Abstract
We have illustrated the life table method for computing survival rates with 5-year survival data for cancer patients, emphasizing the advantage gained by including survival information on cases which entered the series too late to have had the opportunity to survive a full 5 years. The advantage is measured in terms of reduction in standard error of the survival rate. For the five series of patients in this paper, the reduction in standard error ranged from one-third to two-thirds.
References (12)
- M. Merrell et al.
Determination of Prognosis in Chronic Disease, Illustrated by Systemic Lupus Erythematosus
J. Chron. Dis.
(1955) Manual for Cancer Programs
Bull. Am. Coll. Surgeons
(1953)- R. Pearl
Introduction to Medical Biometry and Statistics
(1923) - A.B. Hill
Principles of Medical Statistics
(1956) - J. Berkson et al.
Calculation of Survival Rates for Cancer
- M.H. Griswold et al.
Cancer in Connecticut, 1935–1951
Cited by (2452)
Mix-supervised multiset learning for cancer prognosis analysis with high-censoring survival data
2024, Expert Systems with ApplicationsHigh censoring phenomenon usually occurs in cancer prognosis analysis, which, however, would introduce bias for model construction and limit generalization performance. In this paper, we first explore and identify an appropriate censoring range for cancer prognosis evaluation, upon which we present a mix-supervised multiset learning framework to cope with high-censoring data. Specifically, we construct multiple subsets with the specified censoring proportion, followed by a multiset representation learning method to learn subset-specific representations, which equips with adversary integrality preservation and dependency limitation constraints to ensure the unbiasedness of subsets and eliminate the redundancy among subsets, respectively. Furthermore, a mix-supervised multiset fusion model is proposed to estimate the relative survival risk, in which teacher model can make full use of the survival time of uncensored samples and the prognosis-related attributes of censored ones to generate reliable pseudo-labels and latent-space for student model. We evaluate the proposed method on three public datasets, and extensive experimental results demonstrate its superiority.
Statistical analysis of analytical results near detection limits with illustrations using elemental determinations in medical and biological samples by total reflection X-ray fluorescence
2024, Spectrochimica Acta - Part B Atomic SpectroscopyThe following paper presents statistical analysis of measurement results close to the detection limit of an analytical technique and being random-left censored data. For such observations the measured value is not known precisely but is restricted by the detection limit. In the paper, the idea of random-left censored observations and data analysis are presented generally and next it is shown how to include censored observations into the statistical analysis with the use of available statistical analysis software (survival analysis procedure). Statistical procedure is presented on the examples of the X-ray spectroscopic analysis of element concentrations in medical and biological samples. Censored observations are included in data analysis and quartiles are estimated. Additionally, two- and multigroup comparison is presented for the censored observations. The accuracy of the presented procedure is demonstrated and discussed on the basis of Monte-Carlo simulations.
CoxNAM: An interpretable deep survival analysis model
2023, Expert Systems with ApplicationsSurvival analysis is widely used in medicine, engineering, economics and other fields as an effective method to model the relation between the time of an event of interest occurring and related features. However, traditional survival analysis models lack the ability to capture nonlinearity. In addition, most nonlinear survival analysis models, especially deep learning-based methods, lack interpretability, which limits the practical application of these models. For these gaps, we proposed an interpretable deep survival analysis model named CoxNAM. This model is based on the Cox proportion hazards model and uses neural additive model to predict the hazard function. We also used the backpropagation algorithm to train the model based on the corresponding loss function. When performing a survival analysis, we can obtain the survival functions, shape functions of features, and the importance of related features while predicting the probability of the occurrence of the event of interest. We conducted numerical experiments on two synthetic datasets and one public breast cancer dataset to verify the performance of the model, at the same time, we compared the interpretability with the SHAP framework on the two synthetic datasets and the results demonstrated the effectiveness of the proposed model’s interpretation. We also applied the model for prognostic analysis of gastric cancer patients to illustrate its application. The experimental results indicate that the proposed model performs better on C-index than the classic statistical survival analysis model (i.e., Cox proportional hazards model) and machine learning-based survival analysis models (i.e., random survival forest and DeepSurv), and it can also provide the importance of features related to the time of the occurrence of events of interest and the effect of the feature values on the results. The proposed method shows promising performance and realistic interpretability. The model can potentially be extended to survival analysis problems in multiple domains for relevant decision-making.
Machine learning for survival analysis in cancer research: A comparative study
2023, Scientific AfricanSurvival analysis is at the basis of every study in the field of cancer research. As every endeavor in this field aims primarily and eventually to improve patients’ survival time or reduce the potential for recurrence. This article presents a summary of some cancer survival analysis techniques and an up-to-date overview of different implementations of Machine Learning in this area of research. This paper also presents an empirical comparison of selected statistical and Machine Learning approaches on different types of cancer medical datasets.
In this paper we explore a selection of recent articles that: review the use of Machine Learning in cancer research and/or benchmark the different Machine Learning techniques used in cancer survival analysis. This search resulted in 12 papers that were selected following certain criteria. Our aim is to assess the importance of the use of Machine Learning for survival analysis in cancer research, compared to the statistical methods, and how different Machine Learning techniques may perform in different settings in the context of cancer survival analysis. The techniques were selected based on their popularity. Cox Proportional Hazards with Ridge penalty, Random Survival Forests, Gradient Boosting for Survival Analysis with a CoxPh loss function, linear and kernel Support Vector Machines were applied to 10 different cancer survival datasets. The mean Concordance Index and standard deviation were used to compare the performances of these techniques and the results of these implementations were summarized and analyzed for noticeable patterns or trends. Kaplan-Meier plots were used for the non-parametric survival analysis of the different datasets.
Cox Proportional Hazards delivers comparable results with Machine Learning techniques thanks to the Ridge penalty and the different methods for dealing with tied events but fails to produce results in higher dimensional datasets. All techniques benchmarked in the study had comparable performances. The use of prognostic tools when there is a mismatch between the patients and the populations used to train the models may not be advisable since each dataset provides a differently shaped survival curve even when presenting a similar cancer type.
Time trends in mortality of oesophageal cancer in Finland over 30 years
2023, European Journal of Surgical OncologyOesophageal cancer survival is reported by epidemiological studies, but knowledge on survival trends regarding different histologies and operative treatment status is lacking.
Data from all patients diagnosed with oesophageal cancer in Finland in 1987–2016 was collected from national registries. 1-, 3- and 5-year survival rates were examined stratified by histology (adenocarcinoma (OAC) and squamous cell carcinoma (OSCC)) and treatment strategy (surgery, no surgery). Hazard ratios (HR) with 95% confidence intervals (CI) for death were provided by multivariable Cox regression, adjusted for confounders.
Of the 9102 patients, 3140 had OAC (1074 [34%] oesophagectomies), and 3778 had OSCC (870 [23%] oesophagectomies). Men were overrepresented in both OAC (77%) and OSCC (55%). The proportion of oesophagectomies decreased in both histologies. From 1987 to 1991 to 2012–2016, 5-year survival increased from 11% to 22% in OAC and from 7% to 13% in OSCC. For patients undergoing oesophagectomy, the corresponding increases were from 20% to 49% in OAC and from 11% to 54% in OSCC, and non-operated patients from 5% to 8% and from 5% to 7%, respectively. Earlier calendar period, older age and comorbidity were associated with mortality in both histologies. Female sex was a protective factor for patients operated for OSCC (HR 1.56 (95% CI 1.33–1.83), men versus women).
The prognosis of oesophageal cancer has improved in Finland over the last 30 years in both main histological types. The survival of patients undergoing oesophagectomy has drastically improved, while the prognosis of patients not undergoing surgery is slowly improving but remains poor.
Survival of patients with cervical cancer in India – findings from 11 population based cancer registries under National Cancer Registry Programme
2023, The Lancet Regional Health - Southeast AsiaCancer survival data from Population Based Cancer Registries (PBCR) reflect the average outcome of patients in the population, which is critical for cancer control efforts. Despite decreasing incidence rates, cervical cancer is the second most common female cancer in India, accounting for 10% of all female cancers. The objective of the study is to estimate the five-year survival of patients with cervical cancer diagnosed between 2012 and 2015 from the PBCRs in India.
A single primary incidence of cervical cancer cases of 11 PBCRs (2012–2015) was followed till June 30, 2021 (n = 5591). Active follow-ups were conducted through hospital visits, telephone calls, home or field visits, and public databases. Five-year Observed Survival (OS) and Age Standardised Relative Survival (ASRS) was calculated. OS was measured by age and clinical extent of disease for cervical cancers.
The five-year ASRS (95% CI) of cervical cancer was 51.7% (50.2%–53.3%). Ahmedabad urban (61.5%; 57.4%–65.4%) had a higher survival followed by Thiruvananthapuram (58.8%; 53.1%–64.3%) and Kollam (56.1%; 50.7%–61.3%). Tripura had the lowest overall survival rate (31.6%; 27.2%–36.1%). The five-year OS% for pooled PBCRs was 65.9%, 53.5%, and 18.0% for localised, regional, and distant metastasis, respectively.
We observed a wide variation in cervical cancer survival within India. The findings of this study would help the policymakers to identify and address inequities in the health system. We re-emphasise the importance of awareness, early detection, and increase the improvement of the health care system.
The National Cancer Registry Programme is funded through intra-mural funding by Indian Council of Medical Research, Department of Health Research, India, Ministry of Health & Family Welfare.