Estimating and modeling the cure fraction in population-based cancer survival analysis

In population-based cancer studies, cure is said to occur when the mortality (hazard) rate in the diseased group of individuals returns to the same level as that expected in the general population. The cure fraction (the proportion of patients cured of disease) is of interest to patients and is a useful measure to monitor trends in survival of curable disease. There are 2 main types of cure fraction model, the mixture cure fraction model and the non-mixture cure fraction model, with most previous work concentrating on the mixture cure fraction model. In this paper, we extend the parametric non-mixture cure fraction model to incorporate background mortality, thus providing estimates of the cure fraction in population-based cancer studies. We compare the estimates of relative survival and the cure fraction between the 2 types of model and also investigate the importance of modeling the ancillary parameters in the selected parametric distribution for both types of model.     

Estimating the cure proportion of malignant melanoma,an alternative approach to assess long term survival: A population-based study

ObjectivesA large proportion of patients with cutaneous malignant melanoma (CMM) do not experience excess mortality due to their disease. This group of patients is referred to as the cure proportion. Few studies have examined the possibility of cure for CMM. The aim of this study was to estimate the cure proportion of patients with CMM in a Swedish population.MethodsWe undertook a population-based study of 5850 CMM patients in two Swedish health care regions during 1996–2005. We used flexible parametric cure models to estimate cure proportions and median survival times (MSTs) of uncured by stage, sex, age and anatomical site.ResultsDisease stage at diagnosis was the most important factor for the probability of cure, with a cure proportion of approximately 1.0 for stage IA. While the probability of cure decreased with older age, the influence of age was smaller on the MST of uncured. Differences in prognosis between males and females were mainly attributed to differences in cure as opposed to differences in MST of uncured.ConclusionsThis population-based study showed approximately 100% cure among stage IA disease. Almost 50% of patients had stage IA disease and the high cure proportion for this large patient group is reassuring.     

Time-to-cure and cure proportion in solid cancers in France. A population based study

BackgroundIn cancer care, the cure proportion (P) and time-to-cure (TTC) are important indicators for practitioners, patients, and healthcare policy makers. The recent definition of TTC as the time at which the probability of belonging to the cured group reaches 95% was used for the first time.MethodsThe data stem from the common database of French cancer registries including 335,358 solid tumours diagnosed between 1995 and 2009 at 27 sites. P and TTC were estimated through a flexible parametric net survival cure model for each cancer site, sex, and age at diagnosis with acceptable assumption of cure (excess mortality rate ≤0.05).ResultsTTC was ≤5 years and P was >80% for skin melanoma and thyroid and testis cancers. It was 0 for testis cancer in men <55 and for thyroid cancer in men <45 and women <65. TTC was between 5 and 10 years for all digestive cancers except small intestine and all gynaecologic cancers except breast. It was ≥10 years in prostate, breast, and urinary tract. The range of P according to age and sex was 37–79% for urinary tract 72–88% for prostate and breast, 4–16% for pancreatic and 47–62% for colorectal cancer.ConclusionTime-to-cure was estimated for the first time from a large national database and individual probabilities of cure. It was 0 in the younger patients with testis or thyroid cancer and <12 years in most cancer sites. These results should help improve access to credit and insurance for patients still alive past the estimated TTCs.     

Population-level cure of colorectal cancer in Malta: An analysis of patients diagnosed between 1995 and 2004

AimThe aim of this study was to estimate the population-level ‘cure’ of Maltese colorectal cancer patients diagnosed between 1995 and 2004, and to estimate the median survival time for the ‘uncured’ patients.Methods and study populationAnalysis was conducted on 1470 cases registered by the Malta National Cancer Register between 1995 and 2004 and followed up to end of 2010. The mean age of the patients was 66.4 (95%CI 65.8–67.1), and the number of men and women were equal. Background mortality for 1995–2010 was extracted from publicly available life tables. A mixture model with Weibull survival distribution and identity link was used to model ‘cure’.ResultsThe overall ‘cured’ proportion for the patients diagnosed in 1995–1999 was 45.3% (95%CI 40.2–50.5) while the ‘cured’ proportion for the patients diagnosed in 2000–2004 was 52.3% (95%CI 47.2–57.5). Median survival time for the ‘uncured’ patients increased in the second calendar period from 1.25 years (95%CI 1.04–1.45) to 1.42 years (95%CI 1.15–1.76).ConclusionIn Malta, as in the rest of Europe, improvements have been made in short- and long-term survival over the 15-year period under study. To continue this improvement, differences by age that still persist must be investigated and efforts focused to reduce any gaps between Malta and other European countries.     

Loss in life expectancy and gain in life years as measures of cancer impact

There are a broad range of survival-based metrics that are available to report from cancer survival studies, with varying advantages and disadvantages. A combination of metrics should be considered to improve comprehensibility and give a fuller understanding of the impact of cancer. In this article, we discuss the utility of loss in life expectancy and gain in life years as measures of cancer impact, and to quantify differences across population groups. These measures are simple to interpret, have a real-world meaning, and evaluate impact over a life-time horizon. We illustrate the use of the loss in life expectancy measures through a range of examples using data on women diagnosed with cancer in England. We use four different examples across a number of tumour types to illustrate different uses of the metrics, and highlight how they can be interpreted and used in practice in population-based oncology studies. Extensions of the measures conditional on survival to specific times after diagnosis can be used to give updated prognosis for cancer patients. Furthermore, we show how the measures can be used to understand the impact of population differences seen across patient groups. We believe that these under-used, and relatively easy to calculate, measures of overall impact can supplement reporting of cancer survival metrics and improve the comprehensibility compared to the metrics typically reported.     

Statistical models for analyzing repeated quality measurements of horticultural products. Model evaluations and practical example

In the field of postharvest quality assessment of horticultural products, research on the development of non-destructive quality sensors, replacing destructive and often time consuming sensors, has spurred in the last decennium offering the possibility of taking repeated quality measures on the same product. Repeated measures analysis is gaining importance during recent years and several software packages offer a broad class of routines. A dataset dealing with the postharvest quality evolution of different tomato cultivars serves as practical example for the comparison and discussion of four different statistical model types. Starting from an analysis at each time point and an ordinary least squares regression model as standard and widely used methods, this contribution aims at comparing these two methods to a repeated measures analysis and a longitudinal mixed model. It is shown that the flexibility of such a mixed model, both towards the repeated measures design of the experiments as towards the large product variability inherent to these horticultural products, is an important advantage over classical techniques. This research shows that different conclusions could be drawn depending on which technique is used due to the basic assumptions of each model and which are not always fulfilled. The results further demonstrate the flexibility of the mixed model concept. Using a mixed model for repeated measures, the different sources of variability, being inter-tomato variability, intra-tomato variability and measurement error were characterized being of great benefit to the researcher.     

Estimating the proportion of identifiable individuals and group sizes in photographic identification studies

Simple Bayesian statistical models are introduced to estimate the proportion of identifiable individuals and group sizes in photographic identification, or photo‐ID, studies of animals that are found in groups. The models require a simple random photographic sampling of animals, where the photographic captures are treated as sampling with replacement within each group. The total number of images, including those that cannot be identified, and the number of images that contain identifiable individuals are used to make inference about the proportion of identifiable individuals within each group and as the population when a number of groups are sampled. The numbers of images for individuals within each group are used to make inference about the group size. Based on analyses of simulated and real data, the models perform well with respect to accuracy and precision of posterior distributions of the parameters. Widths of posterior intervals were affected by the number of groups sampled, sampling duration, and the proportion of identifiable individuals in each group that was sampled. The structure of the models can accommodate covariates, which may affect photographic efficiency, defined in this study as the probability of photographically capturing individuals.     

A cure‐rate model for Q‐learning: Estimating an adaptive immunosuppressant treatment strategy for allogeneic hematopoietic cell transplant patients

Cancers treated by transplantation are often curative, but immunosuppressive drugs are required to prevent and (if needed) to treat graft‐versus‐host disease. Estimation of an optimal adaptive treatment strategy when treatment at either one of two stages of treatment may lead to a cure has not yet been considered. Using a sample of 9563 patients treated for blood and bone cancers by allogeneic hematopoietic cell transplantation drawn from the Center for Blood and Marrow Transplant Research database, we provide a case study of a novel approach to Q‐learning for survival data in the presence of a potentially curative treatment, and demonstrate the results differ substantially from an implementation of Q‐learning that fails to account for the cure‐rate.     

Calibration plot for proteomics: A graphical tool to visually check the assumptions underlying FDR control in quantitative experiments

In MS‐based quantitative proteomics, the FDR control (i.e. the limitation of the number of proteins that are wrongly claimed as differentially abundant between several conditions) is a major postanalysis step. It is classically achieved thanks to a specific statistical procedure that computes the adjusted p‐values of the putative differentially abundant proteins. Unfortunately, such adjustment is conservative only if the p‐values are well‐calibrated; the false discovery control being spuriously underestimated otherwise. However, well‐calibration is a property that can be violated in some practical cases. To overcome this limitation, we propose a graphical method to straightforwardly and visually assess the p‐value well‐calibration, as well as the R codes to embed it in any pipeline. All MS data have been deposited in the ProteomeXchange with identifier PXD002370 ( http://proteomecentral.proteomexchange.org/dataset/PXD002370 ).     

Assessment of lead-time bias in estimates of relative survival for breast cancer

Relative survival ratios (RSRs) can be useful for evaluating the impact of changes in cancer care on the prognosis of cancer patients or for comparing the prognosis for different subgroups of patients, but their use is problematic for cancer sites where screening has been introduced due to the potential of lead-time bias. Lead-time is survival time that is added to a patient's survival time because of an earlier diagnosis irrespective of a possibly postponed time of death. In the presence of screening it is difficult to disentangle how much of an observed improvement in survival is real and how much is due to lead-time bias. Even so, RSRs are often presented for breast cancer, a site where screening has led to early diagnosis, with the assumption that the lead-time bias is small. We describe a simulation-based framework for studying the lead-time bias due to mammography screening on RSRs of breast cancer based on a natural history model developed in a Swedish setting. We have performed simulations, using this framework, under different assumptions for screening sensitivity and breast cancer survival with the aim of estimating the lead-time bias. Screening every second year among ages 40–75 was introduced assuming that screening had no effect on survival, except for lead-time bias. Relative survival was estimated both with and without screening to enable quantification of the lead-time bias. Scenarios with low, moderate and high breast cancer survival, and low, moderate and high screening sensitivity were simulated, and the lead-time bias assessed in all scenarios.     

Field Evaluation of Some Models Estimating the Seasonal Pattern of Airborne Ascospores of Venturia inaequalis

A 6-year study was carried out to evaluate the accuracy of some models in estimating airborne ascospores of Venturia inaequalis . The proportion of the season's ascospores trapped on each discharge event was compared with the proportion of mature ascospores, estimated by the New Hampshire model or by some related models. The models differed from each other in the degree-day cumulation, accounting or not for the leaf litter wetness caused by rainfall or by deposition of atmospheric humidity. The New Hampshire model did not fit spore trappings well: 59% of the actual values fell outside the range of the estimates, and 83% of them were overestimates. The wide discrepancy between reality and estimates resulted from the effect of dryness: when many consecutive rainless days occurred, the proportion of ascospores trapped was constantly lower than the model estimates, due to a slowed spore maturation. The effect of dryness was evident during the greater part of the ascospore maturity season, irrespective of the proportion of the season's ascospores that had just matured when the dry period began. Models accounting for leaf litter wetness significantly improved estimates. Therefore, in the Po Valley, the accuracy of the New Hampshire model can be improved by accumulating degree-days only when leaf litter is wet.     

Increasing incidence and survival of corpus uteri cancer in Estonia over the past two decades

BackgroundCorpus uteri cancer has become the fourth most common female cancer in Europe. In Estonia, the prevalence of obesity is increasing, and corpus uteri cancer survival has been relatively low. The aim of the study was to evaluate incidence, mortality and survival trends of corpus uteri cancer in Estonia by age, stage and histological subtypes with an emphasis on surgical treatment.MethodsEstonian Cancer Registry data on incident cases of corpus uteri cancer were used to examine incidence trends (1995–2016) and calculate relative survival ratios (RSR) (1996–2016). Cases were classified by morphology and FIGO stage. Causes of Death Registry data were used to analyse corrected mortality (1995–2017).ResultsA total of 4281 cases were diagnosed in 1996–2016. A significant increase was seen in age-standardized incidence from 2009, while mortality remained stable throughout the study period. Significant increases were observed for type I cancers and age groups ≥65 years. Overall age-standardized 5-year RSR improved from 70% in 1996–2002 to 78% in 2010–2016. Survival increased for type I cancers, all age groups and all stages (significantly for stage IV). The proportion of surgically treated cases increased significantly from 85% to 89%, with the largest increases seen in older age groups and later stages.DiscussionThe rising corpus uteri cancer incidence in Estonia is driven by the type I cancer trend. Survival gain for later stages and older age groups likely reflected more frequent surgical treatment. To reduce mortality, further efforts are necessary to ensure appropriate care for all patients.     

Estimating allowable take for an increasing bald eagle population in the United States

Effectively managing take of wildlife resulting from human activities poses a major challenge for applied conservation. Demographic data essential to decisions regarding take are often expensive to collect and are either not available or based on limited studies for many species. Therefore, modeling approaches that efficiently integrate available information are important to improving the scientific basis for sustainable take thresholds. We used the prescribed take level (PTL) framework to estimate allowable take for bald eagles (Haliaeetus leucocephalus) in the conterminous United States. We developed an integrated population model (IPM) that incorporates multiple sources of information and then use the model output as the scientific basis for components of the PTL framework. Our IPM is structured to identify key parameters needed for the PTL and to quantify uncertainties in those parameters at the scale at which the United States Fish and Wildlife Service manages take. Our IPM indicated that mean survival of birds >1 year old was high and precise (0.91, 95% CI = 0.90–0.92), whereas mean survival of first-year eagles was lower and more variable (0.69, 95% CI = 0.62–0.78). We assumed that density dependence influenced recruitment by affecting the probability of breeding, which was highly imprecise and estimated to have declined from approximately 0.988 (95% CI = 0.985–0.993) to 0.66 (95% CI = 0.34–0.99) between 1994 and 2018. We sampled values from the posterior distributions of the IPM for use in the PTL and estimated that allowable take (e.g., permitted take for energy development, incidental collisions with human made structures, or removal of nests for development) ranged from approximately 12,000 to 20,000 individual eagles depending on risk tolerance and form of density dependence at the scale of the conterminous United States excluding the Southwest. Model-based thresholds for allowable take can be inaccurate if the assumptions of the underlying framework are not met, if the influence of permitted take is under-estimated, or if undetected population declines occur from other sources. Continued monitoring and use of the IPM and PTL frameworks to identify key uncertainties in bald eagle population dynamics and management of allowable take can mitigate this potential bias, especially where improved information could reduce the risk of permitting non-sustainable take.     

The burden of rare cancer in Japan: Application of the RARECARE definition

BackgroundDespite the fact that rare cancer is a new target of cancer control in Japan, the incidence of rare cancers is unknown and there is no generally accepted definition of rare cancers in this country. With the aim of calculating incidences of rare cancers in Japan, we therefore adopted a definition and classification of rare cancers that had been published in the European Union (EU) in 2011.MethodsUsing incidence data between 1998 and 2007 submitted by 12 of population based cancer registries in Japan that met our quality criteria and drawing on the EU definition (incidence <6 per 100,000 per year), we estimated the incidences of 845 combinations of tumor sites and histological groups and thus identified the cancers that are rare in Japan.ResultsAfter identifying 193 combinations of tumor sites and histological groups that fit our criteria for rare cancers, we estimated their incidence to be about 75 per 100,000, which corresponds to about 94,800 new diagnoses in 2012 or approximately 15% of all cancer diagnoses. The categorization of rare and common cancers was almost the same in Japan as in EU.ConclusionsThe present study provides an indication of the size of the rare cancer burden in Japan and epidemiological information to explore this. We are expecting further discussion based on our results with stakeholders in order to construct a Japanese definition of rare cancers.     

A generalization of Gompertz law compatible with the Gyllenberg-Webb theory for tumour growth

We present a new extension of Gompertz law for tumour growth and anti-tumour therapy. After discussing its qualitative and analytical properties, we show, in the spirit of [16], that, like the standard Gompertz model, it is fully compatible with the two-population model of Gyllenberg and Webb, formulated in [14] in order to provide a theoretical basis to Gompertz law. Compatibility with the model proposed in [17] is also investigated. Comparisons with some experimental data confirm the practical applicability of the model. Numerical simulations about the method performance are presented.     

Impact of universal health care and screening on incidence and survival of Thai women with cervical cancer: A population-based study of the Chiang Mai Province

Universal Health Coverage (UHC) was implemented in Thailand in 2002. This study aims to compare cervical cancer incidence and survival before and after the implementation of UHC, including the national screening program, in the Chiang Mai population in Northern Thailand. Data of women diagnosed with in situ or malignant cervical cancer in Chiang Mai during 1998–2012 were used in our analysis. Annual age-standardized incidence rates (ASR) and age-adjusted relative survival (RS) were estimated for the following three diagnosis periods: period I: 1998–2002 (before UHC), period II: 2003–2007 (UHC implementation) and period III: 2008–2012 (after UHC). The ASR peaked in 2001 at 38 per 100,000, and then subsequently declined to 23 per 100,000 in 2012. The proportion of in situ and localized tumors increased in all age groups, while regional tumors declined. In all women (aged 15–89) with malignant cervical cancer or in situ, the 5-year RS in Period I, Period II and Period III was 73%, 74% and 77%, respectively; when only malignant cases were considered, the RS was 63%, 61% and 62%, respectively. In the screening target women (aged 30–59) with malignant or in situ tumors, the 5-year RS was 84%, 88% and 90%, respectively, in the three periods, while the RS was 71%, 74% and 75%, respectively, in only those with malignant cancers. The introduction of UHC including national cervical cancer screening program has likely reduced the magnitude and severity of cervical cancer and improved the survival of cervical cancer in the screening target age group.     

Estimating Population Growth and Recruitment Rates Across the Range of American Common Eiders

Sound management of bird populations rests upon an adequate understanding of their population dynamics. Our study evaluated recruitment and population growth rates of 14 American common eider (Somateria mollissima dresseri) colonies from Labrador, Nova Scotia, Quebec, Canada, and Maine, USA, during various periods between 1970 and 2019. We used Pradel mark-recapture models to estimate colony-specific growth rates and the relative contributions of survival and recruitment on growth. We also validated this approach using annual nest counts (~8,000 pairs) conducted between 2003 and 2019 during down harvest operations in 3 colonies located in the Saint Lawrence estuary in Quebec. There was generally a good agreement between estimates derived using the 2 approaches. We considered that capture-recapture data were suitable to estimate population trends of common eiders in other colonies, especially for colonies where accurate nest monitoring is impaired by dense vegetation. The breeding abundance declined at major colonies in Maine and Nova Scotia and increased or was stable in Quebec and Labrador. Female survival contributed the most to population growth, but variation in recruitment among colonies was more important than variation in survival to explain population growth. Management measures should thus strive to maximize local recruitment in colonies with declining populations. The assumption that apparent survival probabilities were homogeneous throughout an individual capture history was violated at several colonies in Quebec and Labrador. Using recaptures and band recoveries, we showed that the lower apparent survival for newly marked individuals compared to females that had been recaptured at least once was caused by a difference in site fidelity rather than true survival. But <1% of recaptured females dispersed to another colony for breeding, indicating that the lower site fidelity could be related to heterogeneity in capture probability among individuals. © 2021 The Wildlife Society.