A nonrandomized cohort and a randomized study of local control of large hepatocarcinoma by targeting intratumoral lactic acidosis

Background: Previous works suggested that neutralizing intratumoral lactic acidosis combined with glucose deprivation may deliver an effective approach to control tumor. We did a pilot clinical investigation, including a nonrandomized (57 patients with large HCC) and a randomized controlled (20 patients with large HCC) study. Methods: The patients were treated with transarterial chemoembolization (TACE) with or without bicarbonate local infusion into tumor. Results: In the nonrandomized controlled study, geometric mean of viable tumor residues (VTR) in TACE with bicarbonate was 6.4-fold lower than that in TACE without bicarbonate (7.1% [95% CI: 4.6%–10.9%] vs 45.6% [28.9%–72.0%]; p<0.0001). This difference was recapitulated by a subsequent randomized controlled study. TACE combined with bicarbonate yielded a 100% objective response rate (ORR), whereas the ORR treated with TACE alone was 44.4% (nonrandomized) and 63.6% (randomized). The survival data suggested that bicarbonate may bring survival benefit. Conclusions: Bicarbonate markedly enhances the anticancer activity of TACE. Funding: Funded by National Natural Science Foundation of China. Clinical trial number: ChiCTR-IOR-14005319.


Introduction
We recently found that Lactic acidosis could effectively protect cancer cells against glucose starvation or deprivaiton (Wu et al., 2012;Xie et al., 2014). First, lactic acidosis dramatically reduces glycolysis rate with little wasting glucose to lactate, as such, a limited amount of glucose could support cancer cells for a relatively long time otherwise would be exhausted quickly. Second, when glucose was deprived, lactic acidosis transformed cancer cells to a 'dormant' state, via arresting cells at G0/ G1 phase, initiating autophagy, inhibiting apoptosis, etc. The protective function relies on co-presence of lactate and proton, depriving either of which would abolish the function (Wu et al., 2012;Xie et al., 2014). When converting lactic acidosis to lactosis by a base, the protective function is gone; similarly, removing lactate, acidosis conferred cancer cells with little resistance to glucose deprivation.
The significance of intratumoral lactic acidosis in tumor biology has been extensively revealed by many other investigators. Clinical studies showed that high level of lactate was a strong prognostic indicator of increased metastasis and poor overall survival (Gatenby and Gillies, 2004;Brizel et al., 2001;Walenta et al., 2000;Schwickert et al., 1995;Walenta et al., 1997;Yokota et al., 2007;Paschen et al., 1987). The work of Gillies and Gatenby group demonstrated that systematic and tumor pHe alkalization could inhibit carcinogenesis, tumor invasion and metastasis, and they also provided integrated models that can predict the safety and efficacy of buffer therapy to raise tumour pHe (Silva et al., 2009;Robey et al., 2009;Ibrahim-Hashim et al., 2012) and related theoretical work (Martin et al., 2012, Martin et al., 2011. Furthermore, many studies reported that lactic acidosis played multifaceted roles in skewing macrophages (Colegio et al., 2014) and inhibiting the function of cytotoxic T cells (Haas et al., 2015), altering cancer cell metabolism (Chen et al., 2008;Sonveaux et al., 2008), inducing chromosomal instability (Dai et al., 2013), and promoting tumor angiogenesis (Gatenby and Gillies, 2004;Végran et al., 2011).
According to the guideline of Barcelona Clinic Liver Cancer (BCLC) staging and treatment strategy, HCC larger than 3 cm in diameter is not suitable for curative therapy (surgical resection, liver transplantation, and ablation) and the recommended treatment is TACE (Forner et al., 2012;El-Serag, 2011;Knox et al., 2015). But sadly, it is recognized that TACE is not effective to treat large tumors (Sieghart et al., 2015). This leaves the patients with large HCC without choice of effective therapy, as also pointed out by Sieghart et al, "maximal restriction of patients selection for TACE would otherwise only improve the results of the treatment modality per se but again would leave those more advanced patients within the intermediate stage without treatment options." (Sieghart et al., 2015) TACE is for local control of the targeted tumor. T E X equation checkTACE kills HCC via 2 mechanisms, delivering concentrated anticancer drugs locally into tumor and occluding tumor feeding arteries to deprive nutrients to starve cancer cells. We would focus on the second mechanism. Occluding tumor feeding arteries effectively deprive nutrients including glucose. The problem is that embolization-created hypoxia condition would stimulate cancer cells to emit strong signals to initiate angiogenesis (Knox et al., 2015) to reestablish tumor vasculature to bypass the occluded tumor feeding arteries. If the tumor cells cannot be rapidly eliminated, tumor vasculature would be reestablished, and a certain amount of tumor (ranging from a few percent of the original tumor to eLife digest Surgery is the main treatment for liver cancer, but the most common liver cancercalled hepatocellular carcinoma -can sometimes become too large to remove safely. An alternative option to kill the tumor is to block its blood supply via a process called embolization. This procedure deprives the tumor cells of oxygen and nutrients such as glucose. However, embolization also prevents a chemical called lactic acid -which is commonly found around tumors -from being removed. Lactic acid actually helps to protect cancer cells and also aids the growth of new blood vessels, and so the "trapped" lactic acid may reduce the anticancer activity of embolization.
Previous works suggested that neutralizing the acidic environment in a tumor while depriving it of glucose via embolization could become a new treatment option for cancer patients. Chao et al. now report a small clinical trial that tested this idea and involved patients with large hepatocellular carcinomas. First, a group of thirty patients received the embolization treatment together with an injection of bicarbonate -a basic compound used to neutralize the lactic acid -that was delivered directly to the tumor. The neutralization killed these large tumors more effectively than what is typically seen in patients who just undergo embolization Chao et al. then recruited another twenty patients and randomly assigned them to receive either just the embolization or the embolization with bicarbonate treatment. This randomized trial showed that the tumors died more and patients survived for longer if they received the bicarbonate together with the embolization treatment compared to those patients that were only embolized. In fact, four patients initially assigned to, and treated in, the embolization-only group subsequently asked to cross over to, and indeed received, the bicarbonate treatment as well.
These data indicate that this bicarbonate therapy may indeed be effective for patients with large tumors that are not amenable to surgery. In future, larger clinical trials will need to be carried out to verify these initial findings.
even a larger tumor known as progressive disease) would survive and thrive. The lactate concentrations in HCC biopsies were around 20 mM (unpublished data), suggesting a lactic acidosis condition. After TACE, lactic acidosis would be trapped in embolized tumor and it would potentially attenuate the therapeutic efficacy of TACE. If it were true, locally infusion of bicarbonate to neutralize it would result in a severer necrosis in the embolized area.

Results
Local control of targeted large tumors as assessed by viable tumor residues (VTR) TACE is to control the targeted tumors rather than to systematically control the disease. As such, to measure the necrosis of the targeted tumors can quantitatively reflect the therapeutic efficacy of TACE. It is known that large size of HCC is a major obstacle to limit the therapeutic efficacy of TACE (Sieghart et al., 2015), the larger the tumor size, the poorer the efficacy of TACE. Thus, the large HCC is a perfect tumor model to test our hypothesis. If intratumoral lactic acidosis is responsible for the therapeutic limitation, locally neutralizing it should significantly improve the anticancer activity, which can be quantitatively measured by the viable tumor residues. To the best of our knowledge, assessment of the necrosis of tumor after the first treatment is probably most objective with least interferences of known and unknown factors. Unless otherwise indicated, the data presented below are the response to the first TACE treatment. We retrieved 27 patients with large HCC (ranging 5.0-14.6 cm) from the pool of patients treated with cTACE in the hospital between 2010 and 2012 according to the inclusion and exclusion criteria as specified in the Materials and methods section. The amount of viable tumor residues after cTACE treatment was calculated. We then recruited 30 patients for the TILA-TACE group using the same inclusion and exclusion criteria between 2012 and 2013. Most of patients' demographic and clinicopathologic characteristics between the two treatment groups cTACE and TILA-TACE were generally comparable (Table 1). However, patients in the cTACE group were more likely to have multifocal tumors.
The geometric mean of VTR after the first treatment was 45.1% (95% CI: 30.3%-67.0%) in the cTACE group, significantly greater than that in the TILA-TACE (7.1% [95% CI: 4.4%-11.5%]; p<0.0001) (Table 2 and Figure 1). We further evaluated whether treatment effects (VTR) were confounded by some covariates such as age, BCLC tumor stage, extra-hepatic metastasis, HBV DNA copy numbers, viable tumor volume before treatment, macrovascular invasion, and tumor multifocality using general linear models. In this study, none of these clinical covariates were significantly associated with VTR (data not shown). Adjustment of these variables did not appreciably alter the results ( Table 2). We then calculated the relative therapeutic improvement by TILA-TACE as described in Materials and methods. TILA-TACE achieved a 81.1% therapeutic improvement relative to cTACE.
We also categorized tumor responses to treatment into four categories from complete response (CR) to progressive disease (PD) according to EASL criteria. Complete or partial responses to treatment in the TILA-TACE group were significantly higher than that in the cTACE group. The percentage of CR, PR, SD, and PD in the cTACE group was 0%, 44.4%, 33.3%, and 22.2% (Figure 2A), respectively, compared with 23%, 77%, 0%, and 0%, respectively, in the TILA-TACE group ( Figure 2B). The total objective response rate (complete or partial responses) in the cTACE group was 44.4%, whereas it was 100% in the TILA-TACE group ( Figure 2C). The observed greater responses to treatment in the TILA-TACE than in the cTACE persisted after accounting for tumor volume prior to treatment using the proportional odds model (p<0.0001). TILA-TACE: targeting-intratumoral-lactic-acidosis TACE; *Adjusted for viable tumor volume prior to treatment and macrovascular invasion using the general linear regression.
No appreciable alterations in results were found after adjustment for other covariates such as age, BCLC tumor stage, extra-hepatic metastasis, HBV DNA copy numbers, and tumor multifocality.
Because of the marked therapeutic improvement by TILA-TACE, the sample size required for the subsequent RCT to evaluate tumor responses to treatment was rather small. When patient number was 10 for each group, the estimated power value reached 0.826. In 2014, we recruited and randomly assigned twenty patients with large HCC (5.0 -13.5 cm) into cTACE (n=10) or TILA-TACE (n=10). Again, most of the selected clinicopathologic characteristics were well matched between two treatment groups ( Table 3). After completion of the first treatment, the VTR in the TILA-TACE group was 5.5-fold lower than that in the cTACE group (4.6% [95% CI: 1.8%-11.4%] vs. 25.4% [95% CI: 10.1%-64.0%]; p=0.008) ( Figure 3 and Table 4). We also evaluated whether the treatment effect was confounded by aforementioned clinical parameters in the RCT. Among them, only viable tumor volume before treatment was significantly associated with VTR. Adjustment for tumor volume before treatment slightly changed the results, with the VTR in the TILA-TACE of 4.1% (2.0%-8.4%) vs. 28.1% (13.9%-56.8%) in the cTACE (p=0.0009). The relative therapeutic improvement by bicarbonate in the RCT was 80.1%, similar to that observed in the non-randomized cohort, 81.1%.  Table 1. The relative therapeutic improvement by TILA-TACE was 81.1%. Relative therapeutic improvement by bicarbonate is defined as [(m 1 -m 2 )/m 1 ] Â (100%), where m 1 is the mean of viable tumor residues after treatment (cTACE) and m 2 the mean residual after treatment (TILA-TACE), and the maximal therapeutic improvement is 100%. Differences in VTR between two groups were statistically significant (p<0.0001), as assessed using the general linear model after adjustment for viable tumor volume before treatment The online version of this article includes the following source data for figure 1: Source data 1. Calculation of viable volume residues of each patient in the nonrandomized study after the first round treatment. If assessed by tumor objective response rate according to the EASL criteria, in TILA-TACE group, out of 12 targeted tumors, 4 achieved CR and 8 PR (Figure 4), whereas in cTACE group, out of 11 targeted tumors, 1 achieved CR, 6 PR, 2 SD, and 2 PD (Figure 4), with P value of 0.003 after Figure 2 continued volume before treatment; SD, viable tumor volume larger than 50% of the viable tumor volume before treatment; PD, viable tumor volume larger than 100% after treatment. The online version of this article includes the following source data for figure 2: Source data 1. The criteria and classification of response to treatment in the nonrandomized study after the first round treatment.
In the randomized study, 4 patients initially assigned in and treated with cTACE group subsequently requested to cross over to TILA-TACE treatment. Although the cross-over was ethically warranted, this somehow blurred the overall survival difference between cTACE and TILA-TACE ( Figure 5B). In cTACE group, 3 deaths occurred in 6 patients who solely received cTACE treatment, and 4 patients who initially received cTACE treatment and subsequently crossed over to TILA-TACE Figure 3. Viable tumor residues after treatment with cTACE or TILA-TACE in the randomized controlled study. The relative therapeutic improvement by TILA-TACE was 80.1%. Differences in VTR between two arms were statistically significant (p=0.0009), as assessed using the general linear model after adjustment for viable tumor volume before treatment. The online version of this article includes the following source data for figure 3: Source data 1. Calculation of viable volume residues of each patient in the randomized study after the first round treatment. treatment were alive. In TILA-TACE group of 10 patients, 3 deaths occurred and 7 patients live. There was no apparent difference in overall survival between two treatment groups in the intent-totreat (ITT) analysis ( Figure 5B, left panel). However, a survival advantage appears in TILA-TACE treatment over cTACE treatment in the per-protocol (PP) analysis ( Figure 5B, right panel), but statistically not significant. Overall, the RCT was limited by the small sample size and the result of the PP analysis was potentially confounded by the crossover of patients from cTACE group to TILA-TACE.
After pooling all patients together, there was a significant difference of survival between cTACE and TILA-TACE group ( Figure 5C).

Adverse effects
The adverse effect between cTACE and TILA-TACE group were comparable (Tables 5 and 6).

Discussion
In this clinical investigation, we carried out 2 studies sequentially, the first one is a nonrandomized controlled study, which demonstrated a remarkable therapeutic improvement of TILA-TACE, based on which, a randomized controlled study was designed and carried out, and again it demonstrated a superior anticancer activity of TILA-TACE. The most striking point is that the numbers reflecting the therapeutic improvements by TILA-TACE from the 2 studies were nearly identical (81.1% and 80.1%). This confirms the consistency of anticancer activities of cTACE as well as TILA-TACE with respect to local control of large HCC.
We compared the ORR in our cTACE practice with those reported globally. The average objective tumor response to TACE is 35% (range, 16%-61%), as systematically reviewed by LIovet and Bruix for the Barcelona-Clinic Liver Cancer Group in 2002 (Llovet and Bruix, 2003). In 2012, Forner, LIovet, and Bruix summarized that more than 50% of patients had an objective response to TACE (Forner et al., 2012), suggesting that the objective tumor response to TACE in the 10-year period (2002-2012) had been increased for about 15%. The complete tumor response to TACE is rare (0-4.8%) (Jansen et al., 2005). Obviously, the results obtained from our cTACE practice were similar to those reported globally.
TACE is for local control of the targeted tumor. Many previous studies have confirmed that better local control was an independent prognostic indicator for patient survival (Kim et al., 2015;Jung et al., 2013;Kim et al., 2013;Shim et al., 2012;Riaz et al., 2011;Gillmore et al., 2011;Riaz et al., 2010). Kim et al and Shim et al (Kim et al., 2015;Shim et al., 2012) further demonstrated a clear prognostic difference between CR, PR, stable disease and progressive disease. The current study demonstrated that TILA-TACE achieved a remarkable improvement of local tumor control and suggested an early sign of improved survival for patients with large HCCs ( Figure 5A). It is noted that, during follow up, 16 patients in the TILA-TACE arm ( Figure 5A), eventually exhibited progressive disease, including 11 patients with new foci in the liver, 1 with new liver foci and lung metastasis, and 3 with lung metastasis, and 1 with bone metastasis, all of which may account for the death (Supplementary file 2). These observations suggest that fast CR and timely control of  The pattern of tumor response to TILA-TACE and cTACE. The difference between 2 groups was statistically significant (p=0.003), as assessed using proportional odds model. CR, complete necrosis; PR, viable tumor volume less than 50% of the viable tumor volume before treatment; SD, viable tumor volume larger than 50% of the viable tumor volume before treatment; PD, viable tumor volume larger than 100% after treatment.
The online version of this article includes the following source data for figure 4: Source data 1. The criteria and classification of response to treatment in the randomized study after the first round treatment.
recurrent tumors in the liver would likely improve the survival of patients, especially those with large tumor burden and low liver reserve. Nevertheless, the work is limited by the study design. The randomized controlled study designed in this investigation was for local tumor control, not for survival. Although the small sample size allowed us to evaluate the local tumor control, we did not expect that such small sample size would yield statistically significant data for cumulative survival. Evaluation of survival is a matter much more complicated than evaluation of local control. There are many more factors affecting the survival of  Table 1. p=0.0052. (B) Survival of patients described in Table 2. Left panel, all patients; right panel, patients who initially assigned to cTACE but subsequently crossed over to TILA-TACE treatment were excluded. p>0.05. (C) Cumulative Survival of patients pooled from Table 1 and 2. p=0.0133. The online version of this article includes the following source data for figure 5: Source data 1. The survival status of each patient at the cut-off date in the nonrandomized and the randomized studies.
patients than those affecting the local tumor control, e.g., tumor characteristics, liver function reserve, tumor staging, disease complications, vascular invasion, metastasis, etc., all of which must be well controlled. While we acknowledged these limitations of the present small RCT, the preliminary survival data allowed us to rationally calculate the sample size for a subsequent large RCT for evaluating the survival difference between cTACE and TILA-TACE. We are planning a large-scale RCT to further confirm the therapeutic advantage of TILA-TACE with respect to overall and progression-free survival of patients with large HCCs.
There was no significant difference of adverse effect between TILA-TACE and cTACE (Tables 3 and 4), i.e., TILA-TACE was as tolerable as cATCE. It was within our expectation, as locally administration of bicarbonate into tumor is safe.
Taken together, this pilot study demonstrated that bicarbonate infusion locally into HCC can markedly enhance anticancer activity of TACE, supporting the notion that neutralizing intratumoral lactic acidosis combined with glucose deprivation may deliver an effective approach to control tumor.

Patients and study design
The study was performed with patients' written informed consent and with the approval of hospital's Institutional Review Board. Patients' consent to publish is obtained. The study is composed of 2 parts (Reporting standard 1). In the first part, twenty seven patients treated with cTACE were retrospectively retrieved (February 2010 -March 2012) (Supplementary file 1) and the viable tumor residues were calculated, and 30 patients (January 2012 -September 2013) (Supplementary file 2) with large HCC were recruited and treated with TILA-TACE (targeting-intratumoral-lactic-acidosis TACE, in this modality, bicarbonate was infused into tumor to neutralizing intratumoral lactic acidosis). In the second part, the data generated from part 1 were used to estimate the sample size for a subsequent randomized controlled study. Twenty patients (March 2014 -August 2014) (Supplementary file 3) were randomly assigned to either TILA-TACE or cTACE treatment with ratio 1:1 (registration number: ChiCTR-IOR-14005319 in Chinense Clinical Trial Registry; protocol [Supplementary file 5] can be accessed at http://www.medresman.org/), according to the method of random number table. Patient inclusion criteria are: a diagnosis of HCC based on EASL or histological evidence, Barcelona Clinic Liver Cancer (BCLC) stage B or C, Child-pugh A or B, adult patients of age ! 20, hypervascular lesion as evaluated by triphasic MRI and digital subtraction Fever (!38.5) ‡ 13 out of 30 9 out of 27 * The adverse events monitored also include acute hepatic decomposition, irreversible hepatic decompensation, respiratory failure or decompensation, biliary stricture or obstruction, liver abscess, gastrointestinal bleeding, arterial thrombosis, arterial-portal shunting. These events were not observed in the patients. † Patients were retrospectively retrieved from our database. ‡ Mild fever occurred to all patients. Table 6. Adverse events in patients in the RCT.

Adverse events* TILA-TACE cTACE
Pain 2 out of 10 1 out of 10 Fever (!38.5) † 2 out of 10 3 out of 10 * The adverse events monitored also include acute hepatic decomposition, irreversible hepatic decompensation, respiratory failure or decompensation, biliary stricture or obstruction, liver abscess, gastrointestinal bleeding, arterial thrombosis, arterial-portal shunting. These events were not observed in the patients. † Mild fever occurred to all patients.
angiography (DSA). Patient exclusion criteria are: BCLC stage 0, A or D, Child pugh C, evidence of combined A-V shunt.

Sample size calculation in the RCT
Sample size calculation was determined by the values of power and alpha: power (0.8) and alpha (0.05) can properly tell the statistical significance and minimize the number of patients to receive false choice of therapy. In order to calculate sample size for RCT, we calculate the viable tumor residues of 30 patients treated with TILA-TACE and 27 patients treated with cTACE as described below (Evaluation of tumor response, Calculation of total tumor volume, viable tumor volume, and necrotic tumor volume). Based on these data, we assessed the minimal number of patients required in a subsequent RCT. We assume the probability of making a Type I error to be 0.05 (a, two-sided), and the probability of making a Type II error to be 0.20 (b level), so the power of this RCT is 0. 8 (1-b). The sample size was calculated according to Schouten's general formula: Where n 1 and n 2 represent the number of patients assigned to cTACE and TILA-TACE group, g=n 1 / n 2 , m 1 (cTACE) and m 2 (TILA-TACE) is the mean viable tumor residues, and s 1 and s 2 are the corresponding standard deviation, t=s 2 2 / s 1 2 , and Z is the normal deviate for alternative hypothesis at a level of significance. In practice, we used PASS software (version 11) to calculate the power and sample size, using the parameters mentioned above. As can been seen in Supplementary file 4, 9 patients in each group already satisfy the power and alpha value. For a pilot study, the sample size of 10 patients for each group was appropriate, as the estimated power value reached 0.826. Twenty patients were randomly allocated in a ratio of 1:1 to each group for the RCT.

Primary endpoint and secondary endpoint
As the study was designed for observing the local tumor control, the primary endpoint was tumor objective response rate to the first treatment, measured by viable tumor residue (VTR), and the secondary endpoint was the overall survival.

Definition of relative therapeutic improvement
Relative therapeutic improvement by bicarbonate is defined as [(m 1 -m 2 )/m 1 ] Â (100%), where m 1 is the mean percentage of viable tumor residues after treatment (cTACE) and m 2 the mean percentage of viable tumor residues after treatment (TILA-TACE). The maximum therapeutic improvement is 100%.

Quantitative calculation of viable tumor residues after treatment
The enhanced area and nonenhanced area in every slice of a tumor (2 mm thichness of each image if scanned by 3.0T MRI Discovery 750, GE Medical Systems or 3.5 mm thichness of each image if scanned by 3.0T MRI Signa Excite HD,GE Medical Systems), which was confirmed by 2 radiologists, was integrated according to MIPAV software (Partecke et al., 2011) (http://mipav.cit.nih.gov/). Summation of integrated enhanced and nonenhanced area of all slices of a tumor gave the viable and necrotic volumes of a tumor. Total tumor volume was the sum of viable and necrotic volume. Viable tumors were assessed by MRI according to EASL criteria. The enhanced and non-enhanced areas represent viable and necrotic tumors. MRI examinations were performed on a 3.0T MRI (Signa Excite HD,GE Medical Systems,USA) or 3.0T MRI (Discovery 750, GE Medical Systems, USA). We assessed the necrotic and non-necrotic volume using T1 post gadolinium. In some cases, when the T1 post gadolinium image of the lesions did not show typical enhancement, we confirmed these lesions using imaging sequence of DWI and T2W. European association for the study of the liver (EASL) criteria of tumor response to treatment The targeted-tumor response to treatment was assessed 30 days after the first treatment according to EASL criteria (Bruix et al., 2001) and defined as below: complete response (CR), no obvious viable residues; partial response (PR), viable residues <50%; stable disease (SD), viable tumor residues between >50% but 100%; and progressive disease (PD), viable tumors >100%.

Overall survival
Overall survival was measured from the date of the first treatment until the date of death or the final follow up visit.

Treatment
TILA-TACE was performed through the transfemoral route using a 5-Fr catheter (Shepherd-hook modified Angiographic Catheter, HANACO Medical, Tian Jin, China) that was advanced to celiac artery. The tumor feeding arteries were defined by digital subtraction angiography. Then, a coaxial microcatheter (2.8 Fr Marguerite II, ASAHI INTECC GMA CO., LTD, Nagoya, Japan) was selectively inserted through a 5-Fr catheter into the tumor feeding artery, into which, 5% sodium bicarbonate (5% Sodium Bicarbonate Injection, Hunan Kelun Pharmaceutical, Ltd., Hunan, China) was infused alternatively with doxorubicin-lipiodol emulsion and oxaliplatin/homocamptothecin with the dose adjusted to tumor size . The dose of bicarbonate ranged between 50 and 250 ml corresponding to tumor sizes between 5 and 14 cm. Finally, the artery was permanently embolized with PVA of proper sizes (Embosphere, BioSphere Medical, Paris, France) and microcoil (Tornado, COOK Medical, USA).
The following example may give a clearer description of the procedure: If a tumor (10 cm) is to be treated, according to our experience, we would prepare 150 ml 5% bicarbonate, 60 ml lipiodol doxorubicin emulsion (40 mg doxorubicin dissolved in 30 ml contrast medium and mixed with 30 ml lipiodol), oxaliplatin 150 mg in 20 ml 5% glucose, 40 mg homocamptothecin in 20 ml saline, PVA particles (100-300, 300-500, 500-700, or 700-900 mm), and microcoil. The TILA-TACE procedure would be as follows 1. Superselectively identify all tumor feeding arteries. 2. Estimate roughly the tumor volume covered by each tumor feeding artery. If there are 3 tumor feeding arteries, divide bicarbonate, doxorubicin-lipiodol emulsion, oxaliplatin, homocamptothecin into 3 parts. If they cover nearly the same volume of tumor, then divide the agents into 3 equal parts. If this is the case, the procedures will be as below: 3. In each tumor feeding artery, about 25 ml bicarbonate would be injected into tumor feeding artery. 4. Then, 3 ml doxorubicin-lipiodol emulsion, 1 ml oxaliplatin, 1 ml homocamptothecin, 3 ml bicarbonate, 20-40 PVA particles (PVA particle size would be chosen according to the diameter of artery), are sequentially injected. 5. Repeat step 4 until the tumor supported by this artery is totally filled (lipiodol oil injection was monitored under fluoroscopic guidance). This cycle may repeat several times. 6. Embolize the tumor feeding artery using PVA of suitable size to block blood stream. Microcoil is used to prevent washout of lipiodol by blood stream, if the diameter of the tumor feeding artery is large, e.g., the internal diameter is larger than 2 mm. 7. The fully occlusion of feeding artery is confirmed by DSA angiography. The deposit of lipiodol oil was assessed by C-arm CT. 8. Repeat step 3 to 7 on the second and the third feeding artery.
cTACE was performed the same as above, except no bicarbonate. Retreatment was based on the evidence of viable tumor residues. The average sessions of treatment were 4 (range 1-13). TACE, Analyzed the data, Critical revision, Contributed unpublished essential data or reagents; Bin Li, Jianjun Wu, Guangqiang Zhang, Performed TACE, Analysis and interpretation of data, Contributed unpublished essential data or reagents; Gong Yang, Critical revision, Analyzed the data; Xun Hu, Conceived the conception, Designed the study, Wrote the paper, Analyzed the data, Critical revision, Contributed unpublished essential data or reagents Author ORCIDs Xun Hu http://orcid.org/0000-0002-3784-371X

Ethics
Clinical trial registration registration number: ChiCTR-IOR-14005319 in Chinense Clinical Trial Registry; protocol can be assessed at http://www.medresman.org/; available as Supplementary file 5). Human subjects: The study was performed with patients' written informed consent and with the approval of hospital's Institutional Review Board (The Second Affiliated Hospital, Zhejiang University School of Medicine).