An easy model for prediction of human renal clear cell carcinoma: curve fitting for three kidney tumors observed for over 10 years.
Lin Yao,Lei Zhang,Xuesong Li,Zhisong He,Liqun Zhou
DOI: https://doi.org/10.3760/cma.j.issn.0366-6999.20132589
IF: 6.133
2014-01-01
Chinese Medical Journal
Abstract:The current researches on natural history of small renal masses show that small renal tumors grow slowly, and few metastasize.1–4 As the length of follow-up is short in these studies (the longest mean follow-up time is 48 months), changes in tumor size in vivo cannot be observed enough. In addition, the growth model of tumor is linear, and the average growth rate is used to reflect tumor growth in these studies. However, most malignant tumor growth curves were not linear. It may not be correct to assess the growth of tumors by linear models. To identify the relationship between tumor size and growth rate, we analyzed three renal tumor cases in this study which had undergone over 10 years of surveillance. We made curve fittings of the three cases and analyzed the growth of renal tumors based on time in vivo. METHODS Patients We searched for patients under active surveillance for over 10 years from the renal tumor database of the Peking University Urology Institute. Patients with enhancing solid renal tumors were included, and cases with no pathological outcomes and benign outcomes were removed. Hereditary renal cell carcinoma was not in discussion. Three cases with renal cell carcinoma were ultimately chosen. Research methods Tumor was assessed by senior radiologists with CT or MRI. Beginning of active surveillance was defined when renal solid tumors were found in the images. Our study excluded renal cystic tumor (e.g., Bosniak IV renal tumor). Patients were evaluated by imaging every 3-6 months until the operation. The method of evaluation was to measure the longest and shortest diameters of tumor to identify the stage by urinary radiologists. The size of the tumor was evaluated by the longest diameter in images. The equation to calculate the volume of tumor is V=XY(X+Y)/2×0.523 (X, the longest diameter; Y, the shortest diameter). The equation to calculate the VDT is VDT=(T-T0)×log 2/log(V/V0)5 (T and T0, the time of the last and first image evaluation; V and V0, the volume of the last and first image evaluation). Statistical analysis Growth model (L=ea×T+b) was adopted according to the feature of the recorded data. Curve fitting of each case was accomplished by the growth model and the curve linearized after logarithmic transformation. Regression analysis was completed with the initial size of tumor and observation time as independent variables. P<0.05 was statistically significant. RESULTS The length of AS in the three cases is all over 10 years (patient A 137 months, B 155 months, C 150 months). Compared to patient C, the initial tumor sizes of patients A and B are smaller (1.30 cm and 1.60 cm vs. 3 cm) and growth rates are slower (0.32 cm/y and 0.25 cm/y vs. 0.7 cm/y). The VDT in all three cases is over 2 years. All three cases have CT images when first diagnosed and before surgery (Figure 1). The ultimate diameters are 5.00, 4.80, and 11.80 cm for patients A, B, and C, respectively. Pathological outcomes are all renal clear cell carcinoma; patients A and B had G2, patient C had G1.Figure 1.: CT of patient A when first diagnosed as RCC and before operation (A1, A2). Specimen of patient A after partial nephrectomy (A3). CT of patient B when first diagnosed as RCC and before operation (B1, B2). Specimen of patient B after partial nephrectomy (B3). CT of patient C when first diagnosed as RCC and before operation (C1, C2). Specimen of patient C after partial nephrectomy (C3).Time point and the longest diameter of tumor A are illustrated in scatter diagrams (Figure 2). The figure shows that these two variables are distributed as growth model (L=ea×T+b). We used SPSS to make a curve fitting, selecting all curve models. Comparing all coefficients of determination, cubic model is the highest (R2=0.993), while compound, exponential, logistic, and growth models are all 0.980; but growth model is the closest to the distribution of data. The formula of tumor A growth is L=e0.011T+0.126, R2=0.980, P<0.001. Patients B and C go in the same way. They are also distributed as the growth model. Finally, the formula for tumor B growth is obtained as L=e0.007T+0.374 (R2=0.984, P<0.001); for tumor C it is L=e0.009T+1.063 (R2=0.971, P<0.001).Figure 2.: Fitted growth curve of the three cases.We linearized the curve of the three cases by logarithmic transformation. Through linear mixed model and considering the correlation of the measurement results from different patients, STATA was used to perform regression analysis with the initial tumor size (L0) and observation time (T) as independent variables. Coefficient of T was 0.009 039 7 (95% CI 0.008 621 8-0.009 457 5, P<0.001). Coefficient of ln L0 was 1.101 908 (95% CI 0.638 876 5-1.564 94, P<0.001). It is suggested that ultimate tumor size has a significant relation with initial tumor size (P<0.001) and the length of tumor growth (P<0.001). Growth model of renal clear cell carcinoma was lnL= 1.101 908×lnL0+0.009 039 7×T-0.147 877 7. DISCUSSION The relation between tumor size and tumor growth rate is still controversial. Chawla et al3 and Li et al1 showed that there is no significant relation between tumor initial size and growth rate. However, Crispen et al4 report through 173 renal tumor cases that the smaller the tumor is, the faster the tumor grows. Tumors are not all malignant in these retrospective studies, and the average growth rate is used to reflect the growth of tumor. The relationship between the size of tumors and instant growth speed was not analyzed. Thus, the relation between size and tumor growth rate cannot be reflected factually in those studies. In our study, the growth speed in every diameter interval is analyzed. The growth rate is slow when the diameter is 1-4 cm in the three cases. However, it increases with the diameter in the interval 1-7 cm. This demonstrates that the biochemical property is fine when the tumor is small. The newborn vessel increases with the tumor growing. This may cause the acceleration of tumor growth. When the tumor is large enough, the inner tissue lacks oxygen and does not have enough nutriment. This gives rise to a deceleration of tumor growth. There are two major limits in this report. One is the small sample size, which could lead to bias. The other is the fact that the three patients could live more than 10 years shows that the biological characteristics of these three tumors are particularly unique; whether they could be able to represent the overall renal cell carcinoma or not is worth discussing. We made curve fitting for three cases of renal cell carcinoma which had been observed for more than 10 years. It is suggested that the growth model is fit for the curves. The growth model of renal clear cell carcinoma is lnL=1.101 908×lnL0+0.009 039 7×T-0.147 877 7. The ultimate size of the tumor has a significant relationship with initial size (P<0.001) and duration of growth (P<0.001). However, more cases are needed to prove the model. Acknowledgments: We acknowledge and thank all the people who helped make this study possible.