Disease-free periods and overall survival time in these groups were examined using Kaplan-Meier graphs and this website log-rank tests (SPSS for Windows version 14.0, Chicago, IL, USA). The degree of linear relationship between pairs of variables measured on a continuous scale was summarized using correlation (r) and a test for zero slope in a corresponding linear regression model. Kruskal-Wallis’ test was used to test the null hypothesis of equal cisplatin sensitivity for the cell lines. For comparison of 18F-FDG uptake between the cell lines, the following multiple linear regression model was used:FDG = c1 + b1V + c2I2 + b2I2V + c3I3 + b3I3V + c4I4 + b4I4V + c5I5 + b5I5V + c6I6 + b6I6V
where the dependent variable 18F-FDG is 18F-FDG uptake and the independent variables are: V = Number of viable cells five dummy variables contrasting cell lines 2–6 to cell line 1: Ij = 1 if cell line = j, j = 2–6 Ij = 0 otherwise and five interaction parameters (products): IjV = V if cell line = j, j = 2–6 IjV = 0 otherwise This linear model has 12 parameters with the following interpretation: c1: Intercept for the reference cell line
(1) b1: Slope for the reference cell line (1) cj: Intercept difference between cell line j and the reference cell line, j = 2–6 bj: Slope difference between cell line j and the reference cell line, j = 2–6 In this modelling framework, an F-test was used to test the null hypothesis of equal 18F-FDG uptake for the cell lines at a fixed number of of viable see more cells. The packages SPSS 14.0 (Chicago, IL, USA) and Stata 10.0 (StataCorp 2007, College Station, TX, USA) were used for statistical analysis. Results Patients: primary tumour characteristics and clinical course Six new permanent squamous cell carcinoma lines in vitro
were established from 18 HNSCCs, which constitutes an overall success rate of 33%. The overall survival of the patients as a function of the propensity of their tumours to grow in vitro, GKT137831 manufacturer calculated from date of diagnosis, is shown in Figure 1. The outcome for the patients from whom cell lines could be established was worse than for the other patients; the median overall survival being 8 vs. 78 months (p = 0.009;logrank test), and the fraction of 5-year survival 0 vs. 67%. The mean disease-free survival time was 13 months for the patients whose tumours grew as cell lines, compared with 80 months for those whose cancers did not grow in vitro (p = 0.056). No differences were observed in the two groups regarding tumour site, TNM status, stage, grade, ploidity or DNA indices (data not shown). Figure 1 Overall survival of the patients stratified by propensity of their tumours to grow in vitro. Survival times were calculated from date of diagnosis. Four patients were still alive (survival >100 months) when this analysis was carried out.