Background
Cancer is one of the most frequent diseases worldwide and is one of the main causes of hospital admissions. According to GLOBOCAN data, there were a total of 18.1 million new cancer cases and 9.6 million cancer deaths in 2018 [
1]. In both sexes combined, colorectal cancer (CRC) is the fourth most commonly diagnosed cancer and the fifth leading cause of cancer-related death [
1]. Advances in the diagnosis and treatment of cancer patients have generally increased patient survival rates. Indeed, according to the American Cancer Society (ACS), the 5-year survival rate after a cancer diagnosis is around 65% for all tumours, as well as for CRC in particular [
2]. EUROCARE-5 data indicates that in Europe, the survival rate 5 years after diagnosis exceeds 50% for most tumour sites [
3]. According to this report, the age-standardised 5-year survival rate for colon cancer was 57 and 55.8% for rectal tumours [
3].
Despite these figures, very few prospective studies have evaluated the recurrence or mortality rates in survivors of CRC [
4]. Although controversial, a patient with an oncological process diagnosis is currently considered ‘cured’ when they survive 5 years after diagnosis [
5], with these patients often being referred to as ‘long-term survivors’. Thus, in CRC, standard follow-up strategies are usually performed periodically during this period [
6]. However, other authors suggest that this ‘cure’ is not guaranteed for patients who survive the first 5 years after diagnosis and that follow-up strategies should be modified according to the risk factors presented by each patient [
4]. Thus, it would be useful to use a rigorous methodology to determine, in greater detail, the probability (or percentage) of cured patients, i.e. long-term CRC survivors, as well as the variables associated with the prognosis of both these and ‘non-cured’ patients in order to establish more appropriate follow-up strategies.
Kaplan–Meier curves and the Cox proportional hazard models are the statistical methods most commonly used to analyse all-cause mortality in cancer studies, while a competing-risk analysis is preferred to determine cause-specific mortality and its associated factors [
7]. However, cure models, which are still not often used, provide an alternative statistical tool to estimate the cure rates of cancer patients and analyse the differences between those individuals who are long and short-term survivors, as well as to identify covariates associated differently with short or long-term progosis [
8]. Although some authors have highlighted the usefulness of cure models as an analysis strategy which could provide especially useful information for quantifying the improvement in survival figures in CRC [
9], very few publications are available in this regard. These studies show cure rates of around 50% and a median survival rate for uncured patients of about 1 year after diagnosis [
9‐
13]. No such studies have been undertaken in Spain, and furthermore, most of this work has been carried out using population registries and/or do not include clinical information other than the age at diagnosis, sex, and disease stage, and therefore, have not used these models to explore the impact of other covariates on cure rates or survival time.
Data evaluating the long-term health-related quality of life (HRQoL) of CRC survivors are also scarce, although work studying the first 5 years after their diagnosis is more common [
14]. Recent work suggests that these patients present a HRQoL similar to that of the general population, although some factors such as intestinal dysfunction can contribute to their deterioration, even 15 years after their initial diagnosis [
14]. Furthermore, these results should be contrasted with those obtained in other cohorts and in different locations. Therefore, this study was planned with the aim of determining the long-term prognosis in CRC patients, to characterise long-term survivors and their HRQoL, and demonstrate the utility of statistical cure models in the study of CRC survival. Results from this study may be useful to determine the cure rate (proportion of long-term survivors) in CRC, the survival rates of ‘uncured’ patients and associated factors, and to develop a predictive model to identify long-term survivors from among CRC patients. In addition, it will also allow us to describe HRQoL and the prevalence of symptoms in these patients.
Discussion
CRC is a major public health problem worldwide and is one of the most common tumour types in terms of its incidence and associated mortality rates [
1]. Advances in the diagnosis and treatment of CRC has caused its mortality rates to decrease in recent years, thus increasing the number of long-term CRC survivors [
3]. These survivors experience the normal issues related to aging, along with the physical and emotional effects of a cancer diagnosis and of its treatments [
25]. Therefore, one goal remains further improvement of the probability of ‘curing’ this disease while also improving the life expectancy for these cured patients.
Thus, large epidemiological studies are important to check the effects of diagnosis and treatment improvements in terms of patient survival, to identify prognostic factors, and to detect subgroups who could need more frequent follow-up surveillance. Therefore, the primary endpoint of this proposed study will be to provide an accurate, updated estimation of long-term survival in CRC patients, as well as to identify variables that may be associated with the probability of a cure and with survival time in uncured patients.
Traditional statistical methods (such as Kaplan–Meier curves or the Cox proportional hazards model) commonly used to estimate survival in CRC patients, implicitly assume that all patients with the same global diagnosis of CRC are at risk of developing the event of interest (e.g., death from the tumour) if they are followed for a sufficiently long period of time. This hypothesis is reasonable when overall mortality is analysed, but falls short when analysing specific-cause mortality or disease-free survival because some patients will never die from the CRC cancer or suffer a recurrence of the tumour. Therefore, alternative survival models taking this into account should be considered.
Statistical cure models could be a useful alternative in this context, even though they are not often currently used in clinical research [
26]. Cure models assume that a fraction of the patients will be cured by the treatment and will never be at risk for suffering an event related to the specific disease again (e.g., CRC death, recurrence, or metastases). Therefore, these models might provide a better estimation of the ‘cure fraction’ while also modelling the average time to the occurrence of a new event among uncured patients as well as associated prognosis factors. These models therefore allow the clinical determinants of the cure and the variables associated with survival to be analysed [
27].
To the best of our knowledge, this is the first study in Spain employing cure models to analyse the long-term prognosis of CRC patients. Furthermore, no other published studies have evaluated the impact of clinical variables other than age, sex, disease stage, or location on long-term CRC survival by using a similar methodology. The research published to date are mainly cancer-registry population studies that do not include clinically important variables such as those registered in hospital-based cohort studies like this one. Other strengths of this project are its relatively large sample size and the long-term follow-up periods considered.
Health care for long-term survivors must include strategies not only for the early diagnosis of recurrences and new neoplasms, but also to detect the long-term medical and psychological effects of cancer diagnosis and treatments [
25]. Thus, characterisation of long-term survivors and analysis of their outcomes will help researchers to assess the adequacy of the medical care provided to them and to optimise the health resources invested in these patients. The secondary aims of this project will be to provide information on the symptoms and HRQoL of long-term CRC patients. These results will allow us to confirm whether, as other authors have indicated, HRQoL returns to normal 1 year after diagnosis [
28]. This data could be important to CRC patients, for example, in planning scheduled follow-up visits to screen not only for medical issues, but also for the late effects of treatments on patient HRQoL and symptoms.
Finally, this research is not without limitations. Firstly, it is a single-centre study which includes a sample from only one hospital in Spain. Secondly, some of the measurements were obtained from clinical records, and so the possibility of information bias could not be discarded. This could limit the generalisability of the results and so, future studies including other populations are warranted. Nonetheless, this study reflects the outcomes of real-life practice in a specialised hospital. Moreover, because it is a single-centre study, the procedures, metrics, and variables collected were homogeneous. In summary, this study had a large sample size and long follow-up time and its results are expected to help identify the needs and clinical situation of long-term survivors of CRC and will be useful for proposing new models of care for the follow-up of these patients.
Acknowledgements
In memory of Salvador Pita-Fernández. We hope to be able to translate all his energy, enthusiasm, expertise and experience in the development of this project.
The authors would like to thank all the patients in the cohort and their families for generously contributing their time to the development of this study. We would also like to thank the co-authors of previous research projects, without which this one could not be possible.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.