Phase III HEAT Study Adding Lyso-Thermosensitive Liposomal Doxorubicin to Radiofrequency Ablation in Patients with Unresectable Hepatocellular Carcinoma Lesions Free

Liver cancer is the third leading cause of cancer-related deaths worldwide, with hepatocellular carcinoma (HCC) making up 70% to 90% of all liver cancers (1). HCC lesions with a maximum diameter (d_max) of ≤3 cm are generally curable by surgery, liver transplantation, or radiofrequency ablation (RFA). For multinodular lesions of >3 cm d_max, the likelihood of cure is reduced and standard guidelines recommend palliative treatment such as transarterial chemoembolization (TACE) to prolong survival (2, 3). Recently introduced guidelines (4), however, propose that patients with certain intermediate-size tumors can achieve better survival when treated more aggressively than is currently recommended in the widely used Barcelona Clinic Liver Cancer (BCLC) treatment algorithm (2).

RFA has been performed safely in lesions with d_max up to 7 cm (5–8). For lesions of >3 cm d_max, multiple overlapping ablation cycles are required to ablate the tumor plus a 360° 1-cm margin (5, 6, 9, 10). However, when multiple ablations are performed in large tumors, RFA alone is at risk for leaving viable tumor cells in the margins of overlapping ablation zones (11). This increases the possibility of rapid recurrence at the original site and elsewhere, due to vascular spread. Thus, an adjuvant therapy to RFA might improve patient outcomes.

Doxorubicin monotherapy has long been used as palliative therapy for HCC (12). Lyso-thermosensitive liposomal doxorubicin (LTLD, ThermoDox®; Celsion Corporation), which consists of doxorubicin contained within a heat-sensitive liposome, was designed for adjuvant use with RFA. After LTLD administration by intravenous (IV) infusion, the liposomes selectively localize within and around tumor tissues (13) because of their heightened permeability and retention properties (14). At normal body temperatures, doxorubicin remains encapsulated within the liposomes. When the tumor areas are heated to ≥40°C, however, the doxorubicin is rapidly released from the heat-sensitive liposomes and quickly diffuses into local tissues. Heated LTLD produces doxorubicin tumor concentrations up to 25-fold greater than free (nonliposomal) doxorubicin administered at the same doses (15). This increased drug concentration within the tumors can likely be attributed not only to the actions of the heat-sensitive liposomes, but also to the intrinsic properties of the doxorubicin itself, which has been shown to produce increased intracellular drug levels and increased activity when combined with hyperthermia in numerous in vitro studies (16–18).

In a phase I liver cancer trial, the MTD of LTLD was found to be 50 mg/m² (19). About 90% of the liposomal doxorubicin plasma area under the curve (AUC_0–∞) occurred during the first 3 hours following infusion (19), establishing this period as optimal for RFA. An LTLD + RFA dose–response relationship (P = 0.04) was found for time to treatment failure, with a median of 80 days for doses <50 mg/m² versus 374 days for doses ≥50 mg/m² (20). Both this promising phase I efficacy and the synergy between LTLD and heat warranted further examination of LTLD combined with RFA. The current phase III HEAT Study examined whether RFA + LTLD could improve survival outcomes for patients with unresectable HCC lesions of 3 to 7 cm d_max.

The HEAT Study was sponsored by Celsion Corporation and registered with ClinicalTrials.gov (NCT00617981). The protocol was approved by the institutional review board and/or independent ethics committee at each participating site (Supplementary Table S1) before any study-related procedures were performed. In addition, patients had to provide informed consent before they were enrolled and any study-related procedures were performed. The HEAT Study was conducted according to Good Clinical Practice, the Declaration of Helsinki, and local laws. An independent and unblinded Data Monitoring Committee (DMC) reviewed trial integrity and accumulating safety and efficacy data throughout the trial.

Males or females ≥18 years old were eligible if they had one to four unresectable HCC lesions, with at least one lesion ≥3.0 cm d_max but none >7.0 cm d_max (only one lesion could have a 5–7 cm d_max). The anticipated ablation volume could be no larger than removal of three hepatic segments or 30% of total liver volume, whichever was less. Patients also had to be appropriate candidates for RFA. Patients were randomized if they met American Association for the Study of Liver Disease (AASLD) criteria for the diagnosis of HCC (21). Patients also had to have Child–Pugh Class A or B liver disease without current encephalopathy or ascites. Left ventricular ejection fraction had to be ≥50%.

Patients were excluded from the trial if they were scheduled for liver transplantation, were HIV positive, had any prior HCC treatment, prior exposure to doxorubicin, extrahepatic metastasis, concurrent malignancy (except treated squamous cell carcinoma of the skin or basal cell carcinoma of the skin), portal or hepatic vein tumor invasion/thrombosis, evidence of hemachromatosis, history of contrast-induced nephropathy, international normalized ratio (INR) >1.5 times the upper normal limit, platelet count <75,000/mm³, neutrophil count <1,500/mm³, hemoglobin <10.0 g/dL, serum creatinine ≥2.5 mg/dL, serum bilirubin >3.0 mg/dL, serum albumin <2.8 g/dL, or any serious illness within the prior 6 months (e.g., congestive heart failure, myocardial infarction, life-threatening cardiac arrhythmia, cerebral vascular accident).

The HEAT Study was a randomized, double-blind, dummy-controlled trial of the efficacy and safety of RFA + LTLD compared with RFA alone for the treatment of unresectable HCC lesions up to 3 to 7 cm d_max. A centralized computer-generated randomization scheme was prepared before accrual. Randomization and analysis were stratified by largest lesion d_max (3–5 cm and >5–7 cm) and RFA approach (laparoscopic, open surgical, and percutaneous). Eligible patients were randomized 1:1 to the two arms using a web-based interactive response technology system.

To prevent any hypersensitivity reactions to liposomes, patients received a blinded premedication regimen. Twenty-four hours prior to study drug infusion, patients took 20 mg of oral dexamethasone (RFA + LTLD arm) or matching dummy capsule (control arm). Blinded IV premedication was administered within 30 minutes before starting study treatment infusion. Patients in the RFA + LTLD arm received 20 mg dexamethasone, H1 antihistamine (e.g., 50 mg diphenhydramine or 10 mg chlorpheniramine), and H2 antihistamine (e.g., 50 mg ranitidine or 20 mg famotidine). Patients in the control arm received dummy IV premedication of sodium chloride 0.9% or 5% dextrose in water (D5W). The sponsor provided masking materials to maintain blinding, and the unblinded site representative (i.e., pharmacist) prepared the materials to cover IV bags and tubing.

Patients then received a similarly blinded 30-minute intravenous infusion of either 50 mg/m² LTLD or D5W. Only RFA devices approved by the FDA were used (Angiodynamics, Boston Scientific, and Covidien). Because computational modeling showed that beginning RFA 15 minutes after initiation of LTLD infusion maximized bioavailable doxorubicin in the tumor margin (data not shown), RFA was initiated at least 15 minutes after start of study drug infusion and was completed within 3 hours after infusion initiation. Duration of RFA was determined by the treating investigators based on individual considerations; no minimum RFA duration was mandated.

Patients had contrast CT imaging studies of the chest, abdomen, and pelvis within 14 days before treatment to confirm the presence of evaluable HCC and at day 28 posttreatment to confirm initial complete response. The protocol defined a complete response as a complete disappearance of viable tumor on follow-up scan when compared with the pretreatment images. Further CT scans were done at months 3, 5, 7, 9, and 12, and then every 3 months until progression. All protocol-specified CT images were assessed centrally by an independent, blinded, central imaging laboratory using criteria based on those used by Chopra and colleagues (22). CT scans were also assessed by each study site, but the central read was considered definitive.

Patients with incomplete ablations after initial treatment were retreated with the same assigned treatment. The second ablation occurred at least 21 days after initial ablation but no later than 14 days after first postablation CT scan assessment.

After radiologic progression, patients who were not treatment failures could have up to five repeat treatments for a recurrent lesion or for newly identified local or distant intrahepatic lesions. All repeat treatments were the same as initial treatment, with no crossover allowed.

All statistical analyses and reporting were performed using SAS System Version 9.4. Accounting for an annual dropout rate of 10% and slower-than-expected accrual, at least 700 patients needed to be randomized (350 per arm) to achieve 80% power to detect a progression-free survival (PFS) HR of 0.75 (25% risk reduction) in the RFA + LTLD group compared with the RFA-alone group with a 5% two-sided type I error. The primary PFS analysis required 380 events, and one preplanned interim PFS analysis was to be performed after the 700-patient enrollment was complete and ≥50% of the PFS events had been observed. Survival follow-up was to continue until 372 deaths had been observed for the primary overall survival (OS) analysis, to achieve the same study power for OS for a target HR of 0.75.

The primary endpoint was PFS, based on independent radiologic assessment, which was measured from the date of randomization to the first date on which one of the following occurred: local recurrence (within 1 cm of the original lesion) following an initial complete response, new distant intrahepatic tumor, extrahepatic tumor, or death from any cause. An incomplete ablation following two study treatments (treatment failure) was also considered a PFS event. Patients alive with no disease progression as of the PFS cut-off date (December 28, 2012) were censored at the date of last tumor assessment.

Secondary endpoints were OS, patient-reported outcomes (PRO), time to local recurrence, and safety. OS was measured from randomization to death or, if alive at the end of the study, censored at the time of last contact. PROs were measured from randomization to clinically significant symptom deterioration, defined as ≥4-point increase from baseline in the eight-item Functional Assessment of Cancer Therapy-Hepatobiliary Symptom Index (23). Patients whose conditions did not deteriorate while on study were censored at the last date the instrument was administered. Time-to-event endpoints were analyzed in the intent-to-treat (ITT) population using the stratified log-rank test, with the HR and 95% confidence intervals (CIs) based on the Cox proportional hazards model and the associated Kaplan–Meier survival estimates. In addition, multivariate analyses using the Cox proportional hazards model were performed to compare the treatment effect when adjusted for various baseline characteristics.

Adverse events (AEs) and laboratory abnormalities were summarized by treatment group and worst grade for all patients who received ≥1 dose of study drug. AE assessments began at signing of informed consent and continued until 1 month following final study treatment. AEs assessed as related to study treatment by the investigator were recorded until patient discontinuation or 12 months after study drug administration, whichever came first.

Studies completed after the conclusion of the HEAT Study demonstrated that doxorubicin delivery to the tumor margin increases with RFA dwell time in computational modeling of RFA + LTLD (Supplementary Fig. S1). These results were later confirmed in preclinical models demonstrating that increasing ablation duration increases local drug concentration and the tissue volume exposed to drug (24). Hypothesis-generating post hoc subset analyses were performed to investigate whether RFA duration was associated with efficacy. These analyses were not prespecified in the statistical analysis plan.

A total of 701 patients were randomized to RFA alone (n = 347) or RFA + LTLD (n = 354; Fig. 1) at 55 centers in Asia, four centers in Europe, and seven centers in North America. The first and last patients were randomized in March 2008 and May 2012, respectively.

Baseline characteristics are summarized in Table 1. Most of the 701 patients in the ITT population were male (75.6%) and <65 years of age (58.3%). A major portion of patients were Asian (90.3%), with few Caucasian patients (9.7%) and no black patients. Furthermore, 64.6% had a solitary lesion and 82.0% had tumors with a d_max of 3 to 5 cm. The number of Asian patients appears to be increased in the subset population. However, considering the number of population characteristics evaluated and the small size of the non-Asian cohort, not only in this subset but in the study overall, it is difficult to determine whether there is a true interaction between race and extended dwell time or if this is a chance occurrence. Otherwise, baseline characteristics between arms were similar.

Post hoc analyses of patients receiving prolonged dwell time focused on those with solitary lesions who had dwell times ≥45 minutes. This subset had 147 patients in the RFA-alone arm and 138 patients in RFA + LTLD arm. Baseline characteristics between the two arms within this subset were similar (Table 1). The characteristics of this subset and the ITT population were also similar, except that all patients in the subset had solitary lesions.

As indicated in Supplementary Table S2, >96% of patients in both arms received treatment, all as randomized. Fifty-five patients (7.8%) had retreatment for an incomplete initial ablation. After receiving treatment, >94% of patients in both arms achieved complete ablation. The vast majority of patients (>90% in both arms) underwent a percutaneous RFA approach.

All patients were evaluated for efficacy. All efficacy data were based on final assessments made as of December 28, 2012, except for OS, in which interim OS results were based on follow-up as of July 2014 and final OS results were based on follow-up as of August 2016.

The study did not meet its primary objective of PFS improvement in the ITT population with the addition of LTLD to RFA compared with RFA alone (Table 2). In the ITT analysis, the PFS HR was 0.96 (95% CI, 0.79–1.18; P = 0.71; Table 2, Fig. 2A) and the interim OS HR was 0.95 (95% CI, 0.76–1.20; P = 0.67; Table 2; Supplementary Fig. S2A). Although the primary endpoint was not met, the DMC recommended continuing survival follow-up. The final OS HR was 0.98 (95% CI, 0.80–1.20; P = 0.82; Table 2; Fig. 2B).

Multivariate Cox regression analyses were run separately for all three key efficacy parameters: OS, investigator PFS, and independent radiologic review committee (IRRC) PFS. All nine potential prognostic factors presented (Fig. 3; Supplementary Fig. S3) were considered for inclusion in each of the three models. On the basis of stepwise selection, only two factors remained in each model (main factor P ≤ 0.05). For all three efficacy parameter models, the remaining factors were the same: number of baseline tumors (1 vs. 2+) and RFA duration (<45 vs. ≥45 minutes; Supplementary Table S3A). Once the models were established, the interaction of these two prognostic factors with treatment was tested separately (two independent tests in each of the three efficacy parameter models) to identify potential predictive factors. None of the two-way interaction tests were significant (P > 0.10; Supplementary Table S3B). In addition, a full interaction model for each of the three efficacy parameters was tested with both prognostic factors, both two-way interactions with treatment, as well as the interaction between the two prognostic factors, and the three-way interaction term (treatment × number of baseline tumors × RFA duration). In these full interaction models, the only significant term (P < 0.10) was the three-way interaction term in the OS and the investigator PFS models (Supplementary Table S3C).

A prespecified subgroup analysis of the ITT population showed no interim OS benefit with the addition of LTLD to RFA in most of the subgroups analyzed (Fig. 3A). Exceptions were in two relatively small subgroups: patients undergoing open surgery RFA (HR, 0.24; 95% CI, 0.08–0.79; n = 38) and patients ages 65–74 (HR, 0.61; 95% CI, 0.37–0.98; n = 182); both of these subgroups favored combination therapy.

The HEAT Study population fits into both BCLC A and B stages. All patients with solitary lesion disease in this study are BCLC A, and all patients with multinodular disease are BCLC B. When analyzed by stage, the ITT population showed no OS benefit in either subgroup (Supplementary Table S4). The median survival for BCLC A with RFA alone and RFA with LTLD was 59.5 months and 63.2 months, respectively (P = 0.37). For BCLC B, median survival with RFA alone and RFA with LTLD was 44.3 months and 37.0 months, respectively (P = 0.30).

The hypothesis-generating examination of the subset receiving ≥45 minutes RFA dwell time revealed an apparent prolongation of median PFS compared with the subset receiving a shorter dwell time, regardless of whether they received concomitant LTLD (Table 2). The improvement in PFS outcomes for the combination arm from this subset, however, did not reach statistical significance (P = 0.15; Table 2; Fig. 2C). The addition of LTLD to RFA within the prolonged dwell time subset did indicate a significant interim OS benefit, with an HR of 0.63 (95% CI, 0.41–0.96; P < 0.05; Table 2; Supplementary Fig. S2B). This benefit was maintained at the final OS analysis, with an HR of 0.65 (95% CI, 0.45–0.94; P < 0.05; Table 2; Fig. 2D).

A subgroup analysis of the prolonged dwell time subset showed an interim OS benefit trend favoring the addition of LTLD to RFA in most of the subgroups analyzed, with three subgroups demonstrating a statistically significant benefit of combination therapy, including those receiving open surgery RFA (HR, 0.14; 95% CI, 0.03–0.72; n = 22), those <65 years old (HR, 0.55; 95% CI, 0.31–0.97; n = 173), and male patients (HR, 0.51; 95% CI, 0.31–0.84; n = 208; Fig. 3B).

Unlike the prolonged dwell time subset, the shortened dwell time subset (consisting of patients with solitary lesions and RFA dwell time <45 minutes) did not demonstrate PFS or OS benefit in the combination arm (PFS HR, 1.15; 95% CI, 0.75–1.79, P = 0.52; interim OS HR, 1.07; 95% CI, 0.67–1.71, P = 0.78; final OS HR, 1.24; 95% CI, 0.81–1.89, P = 0.32; Table 2). Moreover, the shortened dwell time subset showed no improvement in median PFS or OS values compared with the ITT population.

The study defined treatment failure as incomplete ablation following two treatments with RFA and study medication, which is the complement parameter to complete response. These data are presented in Supplementary Table S5 and show that initial complete response was achieved in >95% of patients treated in both arms of the ITT population, as well as in >95% of patients across both lesion size cohorts (3–5 cm and 5–7 cm).

Treatment-related AEs are defined as AEs that are recognized on or after the date of the first dose and throughout study duration and related to study drug. The overall incidence of treatment-related AEs was 35% in the RFA-alone arm and 83% in the RFA + LTLD arm (Table 3). AEs that were reported as related to combination therapy were spread across all grades and were primarily blood and gastrointestinal disorders. Treatment-related grade ≥3 AEs that occurred at a frequency ≥5% with combination therapy included neutropenia and leukopenia.

Three deaths due to treatment-related AEs occurred in the combination therapy arm, and one occurred in the RFA-alone arm. Two in the combination arm were caused by hepatic function abnormality, and one was caused by hypotension. The death in the RFA-alone arm was caused by cardiorespiratory arrest.

An overview of safety shows that 90% of patients receiving RFA alone and 96% of patients receiving RFA + LTLD experienced at least one treatment-emergent AE (Supplementary Table S6). Treatment-emergent AEs are defined as AEs that are recognized on or after the date of the first dose and throughout study duration. Treatment-emergent AEs leading to discontinuation were uncommon in both arms (<1%, RFA alone; 2%, RFA + LTLD). Deaths resulting from any SAEs had a frequency of 2% in both arms (n = 6, RFA alone; n = 7, RFA + LTLD). The 19 SAEs experienced by these 13 patients (one patient had seven SAEs leading to death) were four events of hemorrhage (upper gastrointestinal hemorrhage, peritoneal hemorrhage, gastroesophageal varices hemorrhage, postprocedural hemorrhage), three events of hepatic dysfunction (two instances of hepatic function abnormal, hepatic failure), three events of infection (two instances of septic shock, abdominal infection), two events of respiratory disorders (chronic obstructive pulmonary disease, aspiration), two events of cardiac disorders (myocardial infarction, cardiorespiratory arrest), and single events of cerebral ischemia, hypotension, multiorgan failure, stress ulcer, and portal hypertensive gastropathy.

Supplementary Table S7 itemizes the treatment-emergent AEs that affected ≥5% of patients and/or the grade ≥3 AEs that affected ≥1% of patients. The known major toxicities of doxorubicin are cardiac dysfunction and neutropenia (25). The AE profile of LTLD in this study was similar to that of conventional doxorubicin, except that no congestive heart failure was observed. Reversible neutropenia was the most important toxicity, with 47% of patients in the RFA + LTLD arm experiencing neutropenia rated as grade ≥3 and 19% rated as serious (Supplementary Table S6). Nine cases of febrile neutropenia, which were generally reversible, were reported. Leukopenia was observed at lower frequencies, with 24% of patients in the RFA + LTLD arm experiencing leukopenia rated as grade ≥3 and 6% rated as serious. Unlike the toxicity profile of a pegylated formulation of liposomal doxorubicin, no hand–foot syndrome was observed with LTLD.

In the solitary tumor subset, the frequency of treatment-emergent neutropenia or leukopenia events do not show a difference by dwell time; however, there is a higher incidence of blood bilirubin increased, aspartate aminotransferase increased, and alanine aminotransferase increased in the prolonged dwell time subset (Supplementary Table S8).

RFA is considered a treatment option for early-stage and very early-stage HCC, as indicated in the widely accepted BCLC system (3, 26). Although ablation is generally not recommended in intermediate-size HCC, it can be considered in select cases (27). Recent studies and newly developed guidelines have suggested that curative interventions should be further considered in patients with intermediate-size HCC due to the heterogeneity of the group (3, 28). Concerns regarding the limits of RFA in larger tumors include the high rates of incomplete ablations and the potential for thermal injury (29). Improvements in technique, tumor visualization and, in some cases, utilization of multiple probes may be helping to overcome these limitations.

We believe the HEAT Study demonstrates that intermediate-size HCC (3–7 cm) can be safely ablated with the expectation of a satisfactory outcome. In this study, initial complete response was achieved in >94% of patients treated, including the 7.8% who required a completion treatment. Even in the subset of patients with relatively large lesions (5–7 cm), the treatment failure rate was <5%. These figures, however, do not account for local recurrences observed during the follow-up period, which occurred in 113 of 701 patients (16.1%) and might indeed have been due to persistent viable tumor undetected by imaging.

The role of LTLD was not established in this study; however, an important subgroup has emerged that deserves further investigation. Patients with solitary lesions appear to have benefited more than patients with multiple lesions, which may be explained by the rapid pharmacokinetics of LTLD (19) along with prolonged RFA heating. Longer heating appears to increase local tissue concentration of doxorubicin released from LTLD (24) and heating may also trigger a local T-cell response (30). In patients with multiple lesions, the time needed to reposition the probe and initiate the ablation would lose the high serum concentration of LTLD due to relatively rapid clearance.

We hypothesized that RFA + LTLD efficacy would be superior to RFA alone among patients with unresectable intermediate-size HCC lesions. Instead, median PFS and interim OS were similar in both study arms. In trying to understand these results, we noted a wide variation in RFA dwell times between the 66 active centers. The study did not require minimum dwell times because the manufacturers of the FDA-approved RFA devices used in this study do not recommend minimum dwell times nor do guidelines exist. Beyond the criterion for RFA adequacy, which was ablation of each target lesion plus a 360° 1-cm tumor margin, no attempt was made to manage RFA dwell time in the HEAT Study.

We now hypothesize that dwell time is relevant for LTLD activity because the target and surrounding tissues remain at ≥40°C, even between ablation cycles. A preclinical study in which healthy pigs received LTLD and RFA demonstrated that doxorubicin concentration within target tissue increased with duration of hyperthermia (24). Thus, an exploratory post hoc analysis was performed to determine whether dwell time may have influenced the outcomes of this study.

Computational modeling examined the kinetics of drug concentration within tumor tissue. Within liver tumor tissue, computational modeling demonstrated not only that doxorubicin concentration increases with heating duration but that the majority of the maximum concentration is delivered in the first 45 minutes, with the remaining 25% delivered in the next 75 minutes (31). Thus, these findings suggest that the cut-off dwell time best associated with efficacy was ≥45 minutes, leading to the use of a 45-minute cutoff for the analyses of patients in this study.

Although patients in the shortened dwell time subset showed no benefit in PFS and OS with the addition of LTLD to RFA, outcomes appeared to be improved in the prolonged dwell time subset. OS was significantly improved in the combination arm of this subset (median not reached vs. 60.2 months, HR, 0.65; 95% CI, 0.44–0.94), and PFS showed a trend toward improvement (22.7 versus. 16.7 months, HR, 0.78; 95% CI, 0.56–1.09) in the same arm. The prolonged dwell time OS data are strengthened by the fact that most subgroups examined favored the combination therapy arm, particularly those subgroups with ≥20 patients.

By themselves, the HEAT Study post hoc subset analysis results can only be hypothesis generating. The multivariate Cox regression model, which investigated the effect of numerous potential prognostic factors on OS in the ITT population, suggested that RFA dwell time was an important prognostic factor. In addition, animal research and computational modeling studies support the hypothesis that the longer the target tissue is heated, the greater the doxorubicin tissue concentration (24, 31). The results of the post hoc analyses and these preclinical studies are all consistent with LTLD's heat-based mechanism of action.

We conclude that the need for a minimum duration of heating (i.e., RFA dwell time) for LTLD efficacy likely explains the equivocal HEAT Study results in the ITT population, because inadequate heating duration may have resulted in subtherapeutic doxorubicin concentrations in local tumor tissue. Among all HEAT Study patients with a solitary lesion, one-third had an RFA dwell time <45 minutes. Moreover, the remaining patients with >1 lesion also may have had insufficient time to heat all tumors for ≥45 minutes each, potentially resulting in subtherapeutic doxorubicin concentrations in one or more lesions.

Neither time to local recurrence nor time to clinically significant symptom deterioration could be calculated in a meaningful way, because the majority of patients progressed before reaching these endpoints; hence, the majority was censored at that time. However, neither the number of local recurrences nor the number of clinically significant symptom deteriorations was improved with the addition of LTLD in this population.

The safety profile of RFA + LTLD was similar to the RFA-alone arm, with the exception of reversible myelosuppression. This is not unexpected, because free doxorubicin also causes myelosuppression (25). While neutropenia was the most frequent grade ≥3 AE, febrile neutropenia affected <2% of patients. Nevertheless, investigators should monitor patients carefully for neutropenia, which is typically observed 2 weeks after LTLD administration. The safety results in this study showed no indication of increased cardiac toxicity with the addition of LTLD, in contrast with the safety profile of free doxorubicin, which has a well-described increased risk of cumulative cardiotoxicity (25, 32). These data suggest that the toxicity profile of RFA + LTLD is manageable. Finally, an investigation of the potential impact of increased RFA dwell time on toxicity showed higher incidence of grade 3/4 elevations in aspartate aminotransferase, alanine aminotransferase, and blood bilirubin in those patients treated with dwell time ≥45 minutes but none of these were graded as serious.

In the end, this was a negative study. The post hoc analyses are clearly hypotheses generating only and require clinical validation. This is now being performed in the prospective phase III OPTIMA study, designed to control for RFA dwell time. We are hoping to learn from the current study and impact future development of LTLD in HCC.

The HEAT Study had a number of limitations. First, RFA treatment lacked rigorous control, both in the device used and duration of dwell time. Preclinical studies conducted after the HEAT Study suggest dwell time positively affects local doxorubicin concentration, so an inadequate dwell time likely adversely affects outcomes. This limitation may explain why the addition of LTLD to RFA had no impact on outcomes in the overall population. Dexamethasone, which was administered in the LTLD arm to prevent type I hypersensitivity, could have blunted the post-ablation inflammatory and immune T-cell response (30). This could be particularly true in patients who did not receive a sufficient thermal dose and local LTLD release. Another limitation of the study is its post hoc analysis, as the dwell time subsets are nonrandomized and therefore firm conclusions regarding the efficacy of combination therapy in these subsets cannot be made.

In conclusion, we believe that RFA can potentially fit into treatment guidelines for patients with BCLC A intermediate-size HCC who have adequate liver function and physical status. Furthermore, based on our subgroup analysis, we hypothesize that adjuvant treatment with LTLD and adequate duration of intrahepatic heating should lead to therapeutic doxorubicin tumor concentrations and, subsequently, to improved outcomes with combination therapy among select patients with unresectable HCC. The RFA literature recommends a minimum of four ablation cycles to ablate a >3.0-cm d_max tumor together with its 360° 1-cm margin (5, 6). In the HEAT Study, RFA dwell times ≥45 minutes were achieved by administering four ablation cycles. To test the new hypothesis that LTLD + RFA with a dwell time ≥45 minutes will improve outcomes in patients with unresectable HCC, we initiated the double-blind, randomized, controlled OPTIMA Study (ClinicalTrials.gov NCT02112656) in 550 patients with a solitary HCC lesion of 3 to 7 cm d_max. The OPTIMA Study differs from the HEAT Study in optimizing both RFA (by specifying ≥4 overlapping ablation cycles) and doxorubicin tumor tissue concentration (by heating a solitary target area ≥45 minutes to concentrate a therapeutic amount of doxorubicin in tumor tissue).

M. Sherman, R.S. Finn, and R. Lencioni are consultant/advisory board members for Celsion Corporation. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.S. Lee, M. Omata, L. Makris, N. Borys, R. Poon, R. Lencioni

Development of methodology: A. Vecchione, M.H. Chen, J.S. Lee, L. Makris, N. Borys, R. Poon, R. Lencioni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.Y. Tak, S.-M. Lin, Y. Wang, J. Zheng, A. Vecchione, S.Y. Park, M.H. Chen, S. Wong, R. Xu, C.-Y. Peng, Y.-Y. Chiou, G.-T. Huang, J. Cai, B.J. Abdullah, J.S. Lee, J.Y. Lee, J.Y. Choi, J. Gopez-Cervantes, M. Sherman, R.S. Finn, M. O'Neal, R. Poon, R. Lencioni

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W.Y. Tak, S.Y. Park, C.-Y. Peng, J.S. Lee, M. Sherman, R.S. Finn, L. Makris, R. Poon, R. Lencioni

Writing, review, and/or revision of the manuscript: W.Y. Tak, S.-M. Lin, S.Y. Park, M.H. Chen, S. Wong, C.-Y. Peng, Y.-Y. Chiou, B.J. Abdullah, J.S. Lee, J. Gopez-Cervantes, M. Sherman, R.S. Finn, M. O'Neal, L. Makris, N. Borys, R. Poon, R. Lencioni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.Y. Tak, Y. Wang, R. Xu, Y.-Y. Chiou, G.-T. Huang, J.S. Lee, J.Y. Choi, L. Makris, N. Borys

Study supervision: W.Y. Tak, Y. Wang, R.S. Finn, M. Omata, N. Borys

Other (one of the main participants of the HEAT study; conducted most of the studies in China): R. Xu

The authors would like to acknowledge Jennifer Klem, PhD for writing assistance and contribution to the manuscript. The HEAT Study was sponsored and supported by Celsion Corporation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.