The improved cure fraction for esophageal cancer in Linzhou city

Esophageal cancer is the eighth most common cancer around the world and the fifth most common cancer in China [1, 2]. There is a large geographical variation with an incidence rate roughly 3 times higher in the incidence rates in low urbanization areas than high urbanization areas in China [3]. Linzhou city has high incidence of esophageal cancer in China. An organized screening of esophageal cancer in population was carried out since 2005. Survival of esophageal cancer in Linzhou was seen to increase over the past few decades and was higher than the average level of China [4‐6]. Population-based studies of esophageal cancer survival are meaningful for clinicians, patients and health administrators.

In the population-based cancer registries, the survival for patients is traditionally measured using the cumulative survival function (cause-specific survival or relative survival) [7, 8]. However, for patients and physicians, the cure fraction may be of great interest. Cure models can be used to calculate both the cure proportion and the survival function of “uncured” patients [9]. When looking at temporal changes, more detailed view of the changes observed in the survival of cancer patients can be obtained with cure model than traditional methods.

We study the survival of patients of esophageal cancer in Linzhou city from 2003 to 2012, using cure model.

There were 8067 patients (56% males and 44% females) with esophageal cancer in this study. The average age at diagnosis was 62 years. The relative survival rate was 36.3%, 33.5%, 26.8% and 17.8% in age group 19–49, 50–59, 60–69 and 70–99 years respectively during 2003–2004. At the year of 2011–2012 the rate was severally 62.5%, 49.0%, 51.1 and 41.7% and there was 72.2%, 46.3%, 90.7% and 34.3% improvements respectively compared with year of 2003–2004. Demographic characteristics and survival indicator were shown in Table 1.

First, models were fitted by sex separately. The fitted curves of the estimated relative survival rate for male and female can be seen in Fig. 1. The empirical life table estimates calculated with the Ederer II method are also shown in the figure. It can be seen that the flexible parametric cure model fits the data well. After 5 years of follow-up the curves of relative survival rate tend to reach a plateau.

During 2003–2004 period, the cure proportion was 33.7%, 31.4%, 21.8% and 11.0% in age group 19–49, 50–59, 60–69 and 70–99 (Table 2, Fig. 2) and at the year of 2011–2012 the proportion was 60.6%, 49.6%, 48.7% and 28.9% separately. There was constant improvement of cure proportion with time going on in all age groups. At the year of 2011–2012, 79.8%, 58.0%, 123.4% and 162.7% improvements of cure proportion were seen in age group 19–49, 50–59, 60–69 and 70–99 years compared with year of 2003–2004.

At the same time, median survival of ‘uncured’ patients changed little during different period. There were separately 0.01, − 0.03, 0.06 and 0.03 years difference in median survival time for age group 19–49, 50–59, 60–69 and 70–99 from 2003 to 2012. The time at which 90% of the ‘uncured’ patients are dead, which is another indicator for measuring survival of the ‘uncured’ has little variation in every age group (Table 2, Fig. 3).

The 5 year relative survival rate was increased in each age group from 2003 to 2012. The cure fraction has increased greatly over time for esophageal cancer in Linzhou city during 2003–2012 with very little variation of median survival of uncured group at the same time. The improvement in survival was mainly attribute to an increasing proportion of ‘cured’ patients.

When studying trends with cure models considering the cure fraction and the survival time of the uncured patients at the same time, better answer will be probably distinguished from many competing explanations [18]. Higher cure proportion and longer survival time of uncured represents a general improvement. Higher cure proportion and shorter survival time of uncured represents a selective improvement: improvements in treatment enabled uncured patients with a longer survival time to be cured. Stable cure proportion and longer survival time of uncured might occur due to better palliative care or new diagnostic techniques introduced which can move up the diagnosed date but without affecting the date of death, that is, lead time bias. Higher cure proportion and unchanged survival time of uncured might occur following introduction of some diagnostic procedure, like screening, which included more patients who have no excess risk.

An organized screening of esophageal cancer was carried out for people aged 40–69 since 2005 in Linzhou city. From 2006 to 2012, 22,869 people took part in the screening of esophageal cancer [19]. There were higher proportion of early-stage esophageal cancer cases and 5 year survival rate of esophageal cancer in endoscopy screening group in Linzhou city [20, 21]. According to the screening procedure, local people aged 40–69 will be organized to screen with endoscope for esophageal cancer. The detected high-grade dysplasia and cancer of esophagus in the screening will be treated and the detected mid-grade dysplasia of esophagus will be checked 1 year later. So there should be some high-grade dysplasia or cancer of esophagus diagnosed by screening procedure in age group over 70 years. Hence the survival of esophageal cancer aged over 70 years can also be affected by the screening. An advantage of cure model is that the lead time bias has no effect on the cure fraction, so we can distinguish the true clinical benefit and lead time bias [18]. In this study, greatly improved cure fraction within short period showed a screening effect and a success in the screening of esophageal cancer in Linzhou city.

Although cure models have given us a better understanding of the survival of patients, we should keep in mind that there are some potential limitations. First, the model will still give the estimates of the cure fraction even when statistics cure is not reached. However, for esophageal cancer it seems not a problem. The cumulative relative survival for esophageal cancer was plateau after about 4 years from the date of diagnosis [4], indicating that the living patients at that time can be considered statistically cured. We have followed-up information at least 4 years which should be enough for reach a point of cure. Second, the traditional cure model may produce biased estimates in older age group due to less flexible to get the shape of the survival distribution [22]. However, the flexible parametric cure model has been shown to perform good even when older groups was included in the model [10]. In this paper, further study of cure fraction in different TNM stage cannot be done for the absence of stage variable in the data provided by Linzhou cancer registry which only collected the essential information of cancer registry during year 2003–2012.

Cure model analysis of cancer survival in China has not been reported ever. In this study, the flexible parametric cure model was used to analyze the survival of esophageal cancer in Linzhou city. And we found the improvement of survival in Linzhou city during 2003–2012 was mainly due to an increasing cure proportion. Huge improvement of cure fraction within short period is likely due to the organized screening of esophageal cancer in Linzhou city.