Baseline tumor Lipiodol uptake after transarterial chemoembolization for hepatocellular carcinoma: identification of a threshold value predicting tumor recurrence

Retrospective single institution database review was performed of patients with primary HCC who underwent Lipiodol-TACE from April 2007 to August 2013. Institutional review board approval was obtained. Inclusion criteria were i) documented non-contrast computed tomography (CT) of the liver within one week after TACE, ii) at least one documented triphasic contrast CT including non-contrast, arterial and delayed phase imaging on follow-up for evaluation of tumor recurrence, iii) lesion diameter ≥ 10 mm, and iv) lesions stained entirely and homogeneously with Lipiodol after TACE. Exclusion criteria were i) patients with more than 6 lesions, ii) lesions stained partially or heterogeneously with Lipiodol; those lesions were not suitable for the Lipiodol uptake measurement because of the nonuniform Lipiodol concentration and possibility of incomplete treatment, iii) lesions treated with additional particulate embolic materials, iv) lesions for which a second TACE or any additional treatment including surgery, radiofrequency ablation (RFA), radiation therapy, or systemic therapy (e.g., sorafenib) was performed before the initiation of follow up. Forty-six tumors in 30 patients meeting criteria were evaluated.

Selective catheterization of lobar, segmental or subsegmental hepatic arteries supplying tumor was performed under standard fluoroscopic guidance using 5-6 Fr base catheters in the first order aortic branch (celiac or superior mesenteric artery) followed by microcatheter selection of higher order branches as appropriate. A Lipiodol-chemotherapy emulsion (doxorubicin or epirubicin, 50mg) was prepared by admixture in a 1:1 or 2:1 ratio using a three-way stopcock and injected into selected hepatic arteries through the microcatheter. The endpoint of injection was stasis of blood flow in the feeding artery or visualization of trans-sinusoidal portal branches. The volume of Lipiodol injected ranged from 1 to 10 mL per lesion. Lipiodol was used as the sole embolic material; no additional embolization with other particulate embolic materials was performed. Although this method corresponds to “chemo-lipiodolization” in European guidelines16 or “Lip-TAI (transcatheter arterial infusion)” in Japanese guidelines17, “TACE” was employed as a unified term in this study. The embolic particles were not routinely used in our institutional practice because our institutional data (not shown) demonstrated patients’ survival outcome was not significantly different using particulate embolization.

Patient data including age, sex, etiology of liver disease, Child-Pugh classification, and Barcelona Clinic Liver Cancer (BCLC) staging were collected from the electronic medical record. Procedural details recorded included selectivity of treatment (lobar vs. subsegmental/segmental), volume of Lipiodol delivered and amount of chemotherapeutic delivered. Image evaluation was performed using a Digital Imaging and Communication in Medicine (DICOM) viewer (IMPAX; Agfa, Mortsel, Belgium). Lipiodol uptake was measured using the Hounsfield units (HU) of the target lesion on a non-contrast CT taken within one week after TACE and on every follow-up CT. The HUs on the CT within one week after TACE represented the baseline Lipiodol uptake. The follow-up interval varied case by case. The HUs were measured by hand drawing a region of interest (ROI) along the lesion periphery at the axial slice with the tumor maximum diameter, using the contrast-enhanced pre-procedure diagnostic scan as a reference. ROI analysis was performed with consistent contrast window level (WL) and width (WW) (WL/WW = 50/350). The diameter of the ROI on the baseline CT was used as the lesion size, since only lesions demonstrating entire and homogeneous uptake were included in the analysis. Lipiodol washout was defined as decreased density of Lipiodol in a part of the tumor or uniformly within the entire tumor, and evaluated using HU analysis on every available follow-up CT. The degree of washout was quantitatively evaluated by calculating the HU difference between the baseline CT and the follow-up CT at which washout was initially detected. If the HUs within the tumor on the last follow-up CT were greater than at baseline CT, as was occasionally seen with contracting tumors after treatment, the quantitative washout was recorded as 1.0 HU. The washout rate was calculated by dividing the quantitative washout by the follow up time period (HU/month). Tumor recurrence was defined as the appearance of new arterial enhancement within or along the margins of a completely treated lesion on the follow-up triphasic contrast CT. Evaluation was terminated when additional treatment was performed or when recurrence was identified. Two board-certified radiologists independently performed all image evaluation, with discrepancies decided by consensus with a third radiologist. Mean values of two measurements were used in all analyses.

Numerical data were summarized as mean ± standard deviation (SD) and the categorical variables were shown as frequency (percentage). Interobserver reliability of measurements was assessed by Cronbach’s alpha18, and standard error of the measurement was calculated using the formula: SEM=SD×1−rxx $SEM = SD \times \sqrt {1 - r_{xx} }$ , where SEM is standard error of measurement, SD is standard deviation and rxx demonstrates the reliability of the test represented by Cronbach’s alpha19. Analyses for predictors of tumor response and estimation of the cutoff lipiodol uptake value predicting tumor response were performed on lesion-by-lesion basis. Uni- and multivariate Cox proportional hazard models were created to evaluate the predictability of baseline Lipiodol uptake for tumor response. In addition to the baseline Lipiodol uptake, other variables such as lesion size, treatment selectivity and washout rate were subjected to the univariate model. Variables with a P value < 0.200 were entered in the multivariable model. Independent samples t-test was used to compare the mean baseline Lipiodol uptake between the tumors with and without recurrence, as well as between the degrees of treatment selectivity. The median recurrence-free interval was defined, and tumor recurrence before the median recurrence-free interval was defined as early recurrence. Baseline Lipiodol uptake and washout rates were compared between tumors with and without early recurrence using the independent samples t-test. Receiver operating characteristic (ROC) curve analysis was used to determine the optimal cutoff point for baseline Lipiodol uptake predicting tumor recurrence, and for Lipiodol washout rates predicting early tumor recurrence. Natural logarithm values of the numerical exposure variables were used in Cox regression models, and P values < 0.05 were considered statistically significant. All statistical analyses were done using Stata IC version 13.1 for Windows.

No recurrences were seen in tumors with uniformly retained Lipiodol at follow up imaging. Of the 29 tumors with visible washout on follow-up imaging, 19 (65.5%) demonstrated recurrence. In univariate analysis, lower baseline Lipiodol uptake and greater Lipiodol washout rate were statistically significant predictors of earlier time to tumor recurrence (Table 2, P = 0.002 and < 0.0001, respectively). Performing a non-selective (lobar) TACE was also significantly associated with earlier time to recurrence compared to selective (segmental/ subsegmental) TACE (Table 2, P = 0.001). In multivariate Cox regression analysis, lower baseline Lipiodol uptake and greater Lipiodol washout rate were both significant predictors of earlier time to tumor recurrence (Table 2, P = 0.001 and < 0.0001, respectively).

Table 2

Results of cox regression analysis to predict factors for tumor recurrence

The mean baseline Lipiodol uptake of tumors without recurrence was significantly higher than in tumors demonstrating recurrence (Figure 1; 426.3 ± 148.8 HU vs. 189.1±137.5 HU, p < 0.0001). Tumors treated with selective TACE demonstrated significantly higher baseline Lipiodol uptake compared with tumors treated with lobar TACE (Figure 1; 401.6 ± 32.4 HU vs. 190.9 ± 26.4 HU, p < 0.0001). Baseline Lipiodol uptake was significantly lower in tumors recurring before the median recurrence-free interval (e.g., early tumor recurrence) compared to tumors recurring after the median recurrence-free interval and tumors without recurrence (Figure 2; 208.5 ± 154.8 HU vs. 373.2 ± 200.3 HU, P = 0.013). By comparison, the rate of Lipiodol washout at follow up was significantly greater in early tumor recurrence compared to tumors recurring after the median recurrence-free interval and tumors without recurrence (Figure 2; 103.1 ± 129.1 HU/month vs. 25.7 ± 26.6 HU/month, P=0.044).

Figure 1

Figure 2

Using the area under the ROC curve for post-TACE baseline Lipiodol uptake, an estimated threshold Lipiodol uptake value of 270.2 HU was identified below which there was a high probability of tumor recurrence and above which there was a high probability of tumor response over the mean follow-up time of 5.6 months with a sensitivity of 95% and specificity of 93% (Figure 3). Comparison of the cumulative hazards of tumor recurrence over the follow-up period also showed significantly higher hazards of recurrence in tumors with Lipiodol uptake < 270.2HU after TACE (Figure 4; P < 0.0001). ROC curve analysis for follow up Lipiodol washout rate revealed a value of 37.8 HU per month above which there was a high probability of early tumor recurrence with a sensitivity of 78% and specificity of 74% (Figure 5). The cumulative hazards for tumor recurrence were significantly higher in tumors with a washout rate > 37.8 HU per month (Figure 6; P = 0.001).

Figure 3

Figure 4

Figure 5

Figure 6