gms | German Medical Science

Introduction: The question of whether and when one is cured from cancer is important for patients, health care providers, and others (e.g., insurance or credit companies). Yet survival probability is the most used measure to statistically describe the survival prospects after a cancer diagnosis. Cure from cancer can depend on many different factors such as sex, age, tumour site, stage at diagnosis and treatment, but also on general health status and socio-demographics. Using population-based data from cancer registries, we want to answer the question “how likely am I to be cured and when will it be?”, also stratified for available prognostic variables.

Methods: Statistical “cure from cancer” is defined as having the same mortality probability as a sex- and age-matched individual from the general population.

We extracted and received data on all cancers diagnosed between 2004 and 2019 from the cancer registry of Schleswig-Holstein. We used the R package “CuRe”. It contains two main functions to model time to cure with many additional specification options with varying flexibility. We modelled time to cure and the proportion being cured using various input parameters and fixed clinical parameters such as cancer site and compared the results depending on input parameters.

Results: In general, higher stage at diagnosis and older age prolong time to cure and lower the number of individuals that reach cure. We observed differences depending on the function and its specifications. E.g., for pancreatic cancer in the simplest model the proportion of people diagnosed in stage I reaching cure is around 30% and about 10 years after diagnosis above 90% of patients can expect cure for stages I-III. Meanwhile, in the most flexible function with one degree of freedom for the additional specifications the proportion reaching cure is above 50% right after being diagnosed with stage I and more than 90% of patients in stage I-III can expect to be cured less than 5 years after diagnosis. The time to cure tends to be shorter and the cure proportion higher for the more flexible functions, at least when only using one degree of freedom for the additional variation possibilities. This is more pronounced if the probability of death is lower, i.e. there isn’t a lot of variation for the very deadly stage IV pancreatic cancer, but for early stage the variation seems very high.

Using more degrees of freedom in the models tends to prolong the modelled time to cure.

Conclusion: Depending on the function used the time to cure might depend on year of diagnosis (with varying degrees of freedom). Time to cure models can be applied to German cancer registry data, but both the time to cure as well as the proportion of cancer patients that can expect cure after diagnosis depend on the function and specifications chosen. Next, we will analyse the data from all over Germany.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.