Complete prevalence and indicators of cancer cure: enhanced methods and validation in Italian population-based cancer registries

Abstract

ObjectivesTo describe the procedures to derive complete prevalence and several indicators of cancer cure from population-based cancer registries.Materials and methodsCancer registry data (47% of the Italian population) were used to calculate limited duration prevalence for 62 cancer types by sex and registry. The incidence and survival models, needed to calculate the completeness index (R) and complete prevalence, were evaluated by likelihood ratio tests and by visual comparison. A sensitivity analysis was conducted to explore the effect on the complete prevalence of using different R indexes. Mixture cure models were used to estimate net survival (NS); life expectancy of fatal (LEF) cases; cure fraction (CF); time to cure (TTC); cure prevalence, prevalent patients who were not at risk of dying as a result of cancer; and already cured patients, those living longer than TTC at a specific point in time. CF was also compared with long-term NS since, for patients diagnosed after a certain age, CF (representing asymptotical values of NS) is reached far beyond the patient’s life expectancy.ResultsFor the most frequent cancer types, the Weibull survival model stratified by sex and age showed a very good fit with observed survival. For men diagnosed with any cancer type at age 65–74 years, CF was 41%, while the NS was 49% until age 100 and 50% until age 90. In women, similar differences emerged for patients with any cancer type or with breast cancer. Among patients alive in 2018 with colorectal cancer at age 55–64 years, 48% were already cured (had reached their specific TTC), while the cure prevalence (lifelong probability to be cured from cancer) was 89%. Cure prevalence became 97.5% (2.5% will die because of their neoplasm) for patients alive >5 years after diagnosis.ConclusionsThis study represents an addition to the current knowledge on the topic providing a detailed description of available indicators of prevalence and cancer cure, highlighting the links among them, and illustrating their interpretation. Indicators may be relevant for patients and clinical practice; they are unambiguously defined, measurable, and reproducible in different countries where population-based cancer registries are active.

Cure indicators and prevalence by stage at diagnosis for breast and colorectal cancer patients: A population‐based study in Italy

“…The completeness, accuracy, and representativeness of the cancer registries' data on incidence and survival in Italy represent an important strength of the study 2,3,10,14 . The size of the study population and the duration of follow‐up (≥15 years for all CRs used in the modeling) both contributed to the reliability of the estimation of long‐term survival, prevalence, and cure indicators.…”

“…The cure models we used may have potential limitations as well. The lack of standardized and widely accepted methods for estimating cancer cure indicators 21,43,44 suggests the need for caution in the international comparisons and interpretation of results for cancer cure indicators 10 . In particular, our mixture cure model includes only two patient groups, while non‐oncological mortality in cancer patients may be higher than the overall mortality in the general population 3,43 due to a combination of factors, such as long‐term adverse effects of treatments and genetic, environmental or lifestyle influences.…”

“…In addition, TTC is inevitably arbitrary and susceptible to the choice of the CNS threshold to fix a low risk of recurrence, death, or the margin of clinical relevance, and the methodological approach used, in particular for cancer types such as BC, which have a non‐negligible long‐term excess mortality rate. Nevertheless, the threshold used (i.e., the 5‐year CNS >95%) is clinically relevant and widely reproducible, as well as the CF estimated using “standard” mixture cure models, allowing comparability between countries 10,21–23,39 . It should be emphasized that, as NS estimates, CF and TTC are considerably less reliable for older patients (e.g., ≥75 years) 45 .…”

“…Indicators were also estimated by sex and subtypes (i.e., colon, rectum) for CRC. The methodological details on the study population, definitions, models used and assumptions, and their validation can be found in a recent paper 10 . All models are specific for each cancer type, stage, and sex, and include period of diagnosis and age group as covariates.…”

“…Parameters were estimated using the SAS NLIN procedure. The goodness of fit of “model‐based” NS to “observed” NS was evaluated by likelihood ratio tests and by visual comparison, for each cancer type and stage, period of diagnosis, sex, and age group (i.e., covariates) 10 …”