Pretransplant Therapy for Hepatocellular Carcinoma
Pretransplant Therapy for Hepatocellular Carcinoma
Using a prospectively maintained transplant database, we performed a retrospective review of all adult patients (18 years of age and older) who underwent LT for HCC and received at least 1 pretransplant LRT at the University of California, Los Angeles, from 1994 to 2013. The primary objective was to determine the extent of pathologic tumor necrosis in the liver explant and its effect on HCC recurrence after LT. Recipient demographics [age, sex, primary end-stage liver disease diagnosis, body mass index (BMI), diabetes, and hypertension], laboratory [physiologic MELD, total cholesterol, alpha fetoprotein (AFP), neutrophil-lymphocyte ratio (NLR)], radiologic (number of lesions, maximum tumor diameter, cumulative tumor diameter, radiographic response to LRT), and treatment-specific factors (number and type of LRT) were analyzed to determine predictors of pathologic response. This study was approved by the UCLA Institutional Review Board.
Contrast-enhanced computed tomography or magnetic resonance imaging was utilized to characterize pretransplant disease extent including lesion number, diameter, and radiologic size criteria [MC, University of California, San Francisco (UCSF) criteria]. Pretransplant LRT included transarterial modalities including bland and chemoembolization (TACE), TA techniques including radio frequency or microwave ablation, and percutaneous ethanol injection. Because only a minority of recipients receiving transarterial embolization underwent bland embolization, all transarterial therapies were categorized as TACE for the purposes of analysis. The number and type of LRTs were recorded and the time from last LRT to LT. Radiologic assessment of the contrast-enhanced computed tomography or magnetic resonance imaging performed subsequent to the final LRT and before LT categorized patients as having no residual/viable tumor (no arterially enhancing lesion/nodularity, stable ablation zone, or embolization defect), possible residual/viable tumor (questionable area of arterial enhancement/nodularity), and definite residual/viable tumor (clear demonstration of a persistent arterially enhancing nodule or a new or increasing lesion).
All liver explants were examined by an experienced hepatopathologist. The extent of tumor necrosis was categorized as none, partial, or complete. An explant was categorized as having a cPR if there was no viable carcinoma in the entire specimen, and any and all lesions were completely necrotic. The presence of even a small focus of viable carcinoma resulted in a partial necrosis categorization; subsequently, patients with no necrosis or partial necrosis were jointly categorized as "No cPR."
Continuous variables were compared using the Wilcoxon rank sum test or Student t test and summarized as means and standard deviation or medians and interquartile ranges (IQRs). Categorical variables were compared using χ test/Fisher exact test and summarized as percentages and frequencies. Ordinal variables were compared between groups using the Wilcoxon rank sum test of trend. Comparisons were made among recipients with and without cPR. Survival curves were computed using the Kaplan-Meier methods and compared using log-rank tests. The survival endpoints for overall, recurrence-free, and disease-specific survival were all-cause mortality, HCC recurrence or all-cause mortality, and mortality due to HCC recurrence, respectively. Adjusted risks of recurrence were calculated controlling for the competing risk of non-HCC mortality and compared using the Fine and Gray test.
A multivariate logistic regression was performed to evaluate the effect of 4 recipient demographic factors (age, sex, diagnosis, and BMI), 5 laboratory (MELD, total cholesterol, immediate pretransplant AFP, maximum pretransplant AFP, and NLR), 5 radiologic (number of lesions, cumulative radiologic tumor diameter, maximum radiologic tumor diameter, MC, and UCSF criteria), and 3 intermediary variables (radiologic assessment after LRT, post-LRT AFP, and time from last LRT to LT) on achieving cPR. Final models were selected using backward stepwise search with P value of less than 0.25 as the retention criterion and summarized as odds ratios (OR) and 95% confidence intervals (CIs). A P value of 0.05 or less was considered statistically significant. In a separate analysis, the effect of the number and type of pretransplant LRT on the probability of achieving cPR was analyzed by univariate (unadjusted) and multivariate (adjusted) analysis, which controlled for recipient MELD, radiologic cumulative tumor diameter, and pretransplant AFP. To assess the relationship between treatment type and number on the odds of having cPR, we created a categorical variable with the following 8 mutually exclusive categories so as not to impose additivity: 1 TA/not TACE, 2 TAs/not TACE, 3 or more TAs/not TACE, 1 TACE/not TA, 2 TACEs/not TA, 3 or more TACEs/not TA, 1 TACE and 1 TA, and 1 or more TACE/1 or more TA (>2 LRTs). To evaluate whether there was a dose-dependent relation between increasing number of treatments and the odds of having cPR, we performed tests of trend using linear contrasts under the aforementioned logistic model.
For the purposes of multivariate regression, missing values were singly imputed using the Markov Chain Monte Carlo method. Linearity was confirmed by fitting splines under the final logistic model. The variables for immediate pretransplant AFP, maximum pretransplant AFP, NLR, cumulative tumor diameter, and maximum tumor diameter were modeled on the log-transformed scale where they displayed normal distribution, compared with a skewed distribution on the original scale. The concordance statistic (C statistic) was reported as a measure of model performance.
Materials and Methods
Using a prospectively maintained transplant database, we performed a retrospective review of all adult patients (18 years of age and older) who underwent LT for HCC and received at least 1 pretransplant LRT at the University of California, Los Angeles, from 1994 to 2013. The primary objective was to determine the extent of pathologic tumor necrosis in the liver explant and its effect on HCC recurrence after LT. Recipient demographics [age, sex, primary end-stage liver disease diagnosis, body mass index (BMI), diabetes, and hypertension], laboratory [physiologic MELD, total cholesterol, alpha fetoprotein (AFP), neutrophil-lymphocyte ratio (NLR)], radiologic (number of lesions, maximum tumor diameter, cumulative tumor diameter, radiographic response to LRT), and treatment-specific factors (number and type of LRT) were analyzed to determine predictors of pathologic response. This study was approved by the UCLA Institutional Review Board.
Contrast-enhanced computed tomography or magnetic resonance imaging was utilized to characterize pretransplant disease extent including lesion number, diameter, and radiologic size criteria [MC, University of California, San Francisco (UCSF) criteria]. Pretransplant LRT included transarterial modalities including bland and chemoembolization (TACE), TA techniques including radio frequency or microwave ablation, and percutaneous ethanol injection. Because only a minority of recipients receiving transarterial embolization underwent bland embolization, all transarterial therapies were categorized as TACE for the purposes of analysis. The number and type of LRTs were recorded and the time from last LRT to LT. Radiologic assessment of the contrast-enhanced computed tomography or magnetic resonance imaging performed subsequent to the final LRT and before LT categorized patients as having no residual/viable tumor (no arterially enhancing lesion/nodularity, stable ablation zone, or embolization defect), possible residual/viable tumor (questionable area of arterial enhancement/nodularity), and definite residual/viable tumor (clear demonstration of a persistent arterially enhancing nodule or a new or increasing lesion).
All liver explants were examined by an experienced hepatopathologist. The extent of tumor necrosis was categorized as none, partial, or complete. An explant was categorized as having a cPR if there was no viable carcinoma in the entire specimen, and any and all lesions were completely necrotic. The presence of even a small focus of viable carcinoma resulted in a partial necrosis categorization; subsequently, patients with no necrosis or partial necrosis were jointly categorized as "No cPR."
Statistical Analysis
Continuous variables were compared using the Wilcoxon rank sum test or Student t test and summarized as means and standard deviation or medians and interquartile ranges (IQRs). Categorical variables were compared using χ test/Fisher exact test and summarized as percentages and frequencies. Ordinal variables were compared between groups using the Wilcoxon rank sum test of trend. Comparisons were made among recipients with and without cPR. Survival curves were computed using the Kaplan-Meier methods and compared using log-rank tests. The survival endpoints for overall, recurrence-free, and disease-specific survival were all-cause mortality, HCC recurrence or all-cause mortality, and mortality due to HCC recurrence, respectively. Adjusted risks of recurrence were calculated controlling for the competing risk of non-HCC mortality and compared using the Fine and Gray test.
A multivariate logistic regression was performed to evaluate the effect of 4 recipient demographic factors (age, sex, diagnosis, and BMI), 5 laboratory (MELD, total cholesterol, immediate pretransplant AFP, maximum pretransplant AFP, and NLR), 5 radiologic (number of lesions, cumulative radiologic tumor diameter, maximum radiologic tumor diameter, MC, and UCSF criteria), and 3 intermediary variables (radiologic assessment after LRT, post-LRT AFP, and time from last LRT to LT) on achieving cPR. Final models were selected using backward stepwise search with P value of less than 0.25 as the retention criterion and summarized as odds ratios (OR) and 95% confidence intervals (CIs). A P value of 0.05 or less was considered statistically significant. In a separate analysis, the effect of the number and type of pretransplant LRT on the probability of achieving cPR was analyzed by univariate (unadjusted) and multivariate (adjusted) analysis, which controlled for recipient MELD, radiologic cumulative tumor diameter, and pretransplant AFP. To assess the relationship between treatment type and number on the odds of having cPR, we created a categorical variable with the following 8 mutually exclusive categories so as not to impose additivity: 1 TA/not TACE, 2 TAs/not TACE, 3 or more TAs/not TACE, 1 TACE/not TA, 2 TACEs/not TA, 3 or more TACEs/not TA, 1 TACE and 1 TA, and 1 or more TACE/1 or more TA (>2 LRTs). To evaluate whether there was a dose-dependent relation between increasing number of treatments and the odds of having cPR, we performed tests of trend using linear contrasts under the aforementioned logistic model.
For the purposes of multivariate regression, missing values were singly imputed using the Markov Chain Monte Carlo method. Linearity was confirmed by fitting splines under the final logistic model. The variables for immediate pretransplant AFP, maximum pretransplant AFP, NLR, cumulative tumor diameter, and maximum tumor diameter were modeled on the log-transformed scale where they displayed normal distribution, compared with a skewed distribution on the original scale. The concordance statistic (C statistic) was reported as a measure of model performance.
