Risk prediction of radiation-induced hepatic toxicity 1 complications in patients with hepatocellular 2 carcinoma 3

Featured Application: This study shows that the dose-volume factors are effective in the risk of radiation-induced hepatic toxicity complications in patients with hepatocellular carcinoma. Abstract: We use dose-volume factors to predict the risk of radiation-induced hepatic toxicity (RIHT) complications in patients with hepatocellular carcinoma (HCC) for controlling the low tolerance of liver organs to radiation and reducing the incidence of radiation-induced hepatic toxicity complications. This study retrospectively collected 114 patients who underwent Intensity Modulation Radiation Therapy (IMRT) for hepatocellular carcinoma between 2014 and 2017. The total number of patients was 69 after excluding normal liver organs whose volume did not reach of RIHT can be less than 50%. It can control the low tolerance of liver organs to radiation and reduce 49 the incidence of hepatotoxic complications induced by radiotherapy techniques. to the Induced Hepatic Toxicity,

50%. A total liver receiving a dose volume of 35 Gy should be less than 54.75% so that the probability of RIHT can be less than 50%. It can control the low tolerance of liver organs to radiation and reduce algorithms, logistic regression (LR) and NTCP models. The area under the receiver operating characteristic curve (AUC), accuracy (ACC) and negative predictive value (NPV) were used to compare the system performance. The flow diagram of this study is shown in Figure 1.  (2) 144 ( ( = 1) 1 − ( = 1) ) = 0 + 1 1 + 2 2 + ⋯ + The individual models were evaluated using three evaluation methods to analyze the best model 145 for inducing RIHT. The three evaluation methods were: ACC, as shown in (4); AUC, as shown in (5);

162
distribution problems. The model estimates the probability of suffering from RIHT and presents the 163 outcome as a probability whose value is limited between 0 and 1 with a threshold value of 0.5. If the 164 probability is greater than 0.5 then the patient has the complication of RIHT, otherwise they do not.

165
The model is described in (10)    liver factors indicate that the impact is significant in the order age, NLV30 Gy, and NLV35 Gy. The non-177 zero factor coefficients in total liver factors indicate that the impact is significant in the order age,

178
TLV35 Gy, and TLV40 Gy. The LASSO factor selection ranking is shown in Figure 3. According to the selected predictors, they are substituted into LR to establish a probability verification model for 180 selecting the best combination of factors that induce RIHT.
In Figure 3A, 3C, coefficients were actually pushed to zero with lasso penalty for feature selection of normal liver factors and total liver factors, respectively. In Figure

187
The optimal value of lambda was found using minimization form of the cross-validation error  The LR model for predicting the risks of normal liver and total liver factors is shown in Table 4 195 For predicting the risks of normal liver, the combined prediction model using age and NLV30 Gy  Abbreviation: NLV5-50 Gy = normal liver volume receiving 5 -50 Gy, TLV5-50 Gy = total liver volume receiving 5 -   Table 5. Abbreviation: NLV5-50 Gy = normal liver volume receiving 5 -50 Gy, TLV5-50 Gy = total liver volume receiving 5 -

213
The univariate dose-response fitted curve of normal liver (using NLV30 Gy) for the incidence of

221
RIHT patients treated with IMRT is shown in Figure 4. The parameters for the univariate NTCP 222 regression model were calculated using the percentage and the absolute of the normal liver volume 223 that received more than 30 Gy. According to the model curve, the tolerance of NLV30 Gy producing a 224 50% complication rate (TD50) was determined to be 54.75% in volume in HCC patients treated with

226
The univariate dose-response fitted curve of total liver (using TLV35 Gy) for the incidence of RIHT 227 patients treated with IMRT is shown in Figure 5. The parameters for the univariate NTCP regression 228 model were calculated using the percentage and the absolute value of the normal liver volume that 229 received more than 35 Gy. According to the model curve, the tolerance of TLV30 GY producing a 50% 230 complication rate 231 (TD50) was determined to be 87.40% in volume in HCC patients treated with IMRT. The tolerance 232 corresponding to a 25% incidence of complications (TD25) was 47.40%.

235
In this study, the data revealed that the combined prediction model using age and NLV30 Gy 236 binomial factors has the best performance for predicting the risks of normal liver with NPV reaching 237 0.73, and using age and TLV35 Gy binomial factors has the best performance for predicting the risks of 238 total liver with NPV reaching 0.74.

239
Previous studies have reported a wide range of incidences of RIHT.

289
Radiation-induced hepatotoxicity is one of the most severe dose-limiting toxicities of HCC 290 patients receiving RT. Although most RIHT cases are usually self-limiting and can be treated with 291 supportive therapy, this complication may lead to a decline in liver reserve, moreover, in severe cases, 292 may lead to liver failure and death. The prognosis of patients with liver cancer is related to the extent 293 of tumor and liver remaining reserves. Therefore, when patients with HCC treated with RT, it is 294 important not only to maintain a low dose to the liver, but also to deliver an effective radiation dose 295 to the tumor. For avoiding RIHT, it is required both a better understanding of the biological 296 properties of RIHT and the parameters that predict the occurrence of RIHT.

298
For patients with HCC receiving RT, to control the liver organs' low tolerance to radiation and 299 reduce the incidence of RIHT, we recommend that a normal liver receiving a dose volume of 30 Gy 300 should be less than 54.75% in volume so that the probability of RIHT can be less than 50%. The overall 301 liver is recommended to receive a dose volume of 35 Gy and should be less than 87.40% in volume,

302
which can make the probability of RIHT less than 50%.