| BackgroundSurvival analysis is a statistical method which connects the outcome with survival time. In survival analysis, survival rate is the most important indicator in describing the survival law, which is the main research content in survival analysis. Survival rate is often measured by observed survival rate and relative survival rate. Relative survival rate can reflect the net survival on the assumption that the cancer is only reason of the cause of death. So, it always is the commonly choice to describe the survival situation of cancer patients.In the survival analysis for cancer patients, there are ending unknown survival data in the end of follow-up, namely censored data. When fitting model for the data of this kind, the commonly method is Cox proportional hazard model and Weibull model. Semiparametric models like Cox model and parameter model like Weibull model all have some limitations. Scholars have been looking for a new method which can make up this deficiency. In 2001, Royston and Parmar presented flexible parametric models for the first time. This method expanded the traditional model in three aspects. First, the method use time spline function to establish parametric model for the baseline distribution, secondly it can provide proportional odds and probability function, and allowing regression coefficient to vary with time. In 2007 Nelson extended the models to relative survival rate and time-dependent effects could be incorporated. In 2009, Lambert improved the model. The flexible parametric survival model incorporated the standard model and relative survival model, adding the time-dependent effects in model. A more stable model will be obtained by using the restricted cubic spline function. The survival and hazard curves will be more smoother compared with the standard model and the calculation speed will be faster.Now the approach of flexible parametric survival model has been used to estimate survival rates, publish the papers and reports in European. In domestic cancer registries, this new method of survival analysis has not been applied. ObjectivesThis study fitted flexible parametric survival model using the stomach cancer datasets of Linzhou city Henan province. Estimate the survival rate, hazard rate and excess mortality rate. Quantitative the differences between relative survival rate, excess mortality rate. Then evaluate the prognosis of patients with stomach cancer and provide basis for cancer prevention.Methods1. All incidence data for stomach cancer were drawn from Linzhou Cancer Registry. With the closing date of May 15,2008, all incidence data were linked to death database. Cases in each database were matched. Records with unsure outcomes were followed up to get accurate information. Dates of migration for settlers were recorded. Those records that were identified as duplicate cases or those belong to death certificate only were excluded.2. The data was sorted and divided into three groups according to the calendar year of registry. Then the data was divided into four groups according to the age of cancer patients diagnosed.3. Fitting flexible parameter relative survival model, calculate relative survival rate, hazard rate, excess hazard rate of different calendar year of diagnosis, different age groups and different gender groups. Plot relative survival rate, hazard rate and excess hazard rate curves versus follow-up time to different registry year of diagnosis, different age groups and gender group. Compare the difference between groups.Results1. There were 11185 cases included in this study. Those records that were identified as duplicate cases or those belong to death certificate only were excluded. As a result, only 10837 cases can effectively analysis, including 7116 male (65.66%) and 3721 female (34.34%).2. For stomach cancer from Linzhou, the 1-year,2-year,4-year,5-year and 10-year relative survival rates were 49.35%,37.26%,32.08%,31.34% and 30.92%, respectively. The relative survival rates of man at the same time points were 51.03%, 38.16%,32.85%,31.42%and 31.05%, respectively. The relative survival rates of woman at the same time points were 49.16%,36.45%,31.03%,29.87%and 28.59%, respectively. The 1-year,2-year,4-year,5-year and 10-year relative survival rates of patients diagnosis in 1988-1994 were 45.19%,34.54%,29.88%,27.59%and 26.58%, respectively. The relative survival rates of patient diagnosis in 1995-1999 at the same time points were 49.29%,38.46%,34.48%,32.49%and 31.71%, respectively. The relative survival rates of patient diagnosis in 2000-2004 at the same time points were 52.17%,42.18%,40.14%,39.27%and 37.43%, respectively. The relative survival rates of patient less than 50 years old at the same time points were 56.02%,43.16%, 38.05%,37.49%and 36.18%, respectively. The relative survival rates of patients 51-60 years old were 54.21%,42.10%,36.19%,35.14% and 34.12%, respectively. The relative survival rates of patients 61-70 years old were 53.15%,41.36%,35.47%, 34.51% and 33.97%, respectively. The relative survival rate of patients 71 years or older group were 47.05%,36.42%,28.46%,28.04% and 27.86% respectively.3. In fitting flexible parameter survival model, the hazard rates of the patients of 51-60 years old,61-70,71 years or older, female patients,1995-1999 registered group and 2000-2004 registered group were 1.064,1.341,2.044,1.030,0.800 and 0.623, respectively. In fitting flexible parameter relative survival model, the excess hazard rates for five groups were 0.948,0.988,1.266,1.087,0.775 and 0.594 respectively. In fitting time-dependent effects model, the coefficients of five variable were-0.029,0.024,0.141,0.065-0.259 and-0.524 respectively. The hazard rates of time-dependent effect were 0.971,1.025,1.151,1.067,0.772 and 0.592 respectively.4. For patients of different age groups, the 0.5-year,1-year,5-year and 10-year mortality rate ratio were 1.84,1.73,1.48 and 1.41, respectively. For different gender patients, the 0.5-year,1-year,5-year and 10-year mortality rate ratio were 1.13,1.10, 1.07 and 1.03, respectively. The mortality rate ratio of 1995-1999 registered group and 1988-1994 registered group in four time points were 0.78,0.89,0.80 and 0.73, respectively. The mortality rate ratio of 2000-2004 registered group and 1988-1994 registered group in four time points were 0.69,0.74,0.65 and 0.59, respectively.5. There are differences between groups in excess mortality rate. The difference between age groups, gender and years of registry can reach up to 700,60,180 and 475 per 1000 person years. The maximum value all appeared in six months to one year. There are also differences of relative survival rate between groups, the difference between age groups increase slowly. The difference between gender and years of registry are growing fast after diagnosis, after one year the difference is stable. The difference between male and female is stable in 2%. The difference between 1995-1999 registry group and 1988-1994 registry group is stable in 12%. The difference between 2000-2004 registry group and 1988-1994 registry group is stable in 24%. The difference between 2000-2004 registry group and 1995-1999 registry group is stable in 9.6%.Conclusions1. The prognosis of patients diagnosed younger were significantly better than those diagnosed older, the prognosis of male were better than female. The prognosis of late registration group was significantly better than early registration group.2. Rapidly reduction of survival rate, rapid change and maximum value of mortality rate and excess hazard rate all appeared in the period from diagnosis to one year later. It is obvious that this period is extremely critical to improve the survival of the patients.3. By using restricted cubic spline function, the flexible parameter survival model is more stable, the survival and hazard function curves are smoother than the standard model. The complex time-dependent effects are simple to fit and results are showed in curves. Using the flexible parameter survival model we can obtain useful prediction more easily, operation will be simpler and the weakness of traditional survival analysis methods will be made up. |
PDF Full Text Request