Local control and recurrence patterns after stereotactic irradiation delivered in more than 4 fractions for hepatocellular carcinomas and liver metastases: a retrospective study

Abstract

Background

Stereotactic Body Radiation Therapy (SBRT) is a safe and effective treatment for liver metastases or hepatocellular carcinoma (HCC) with a dose–response relationship for local control (LC). Proximity to organs at risk (OAR) often requires dose de-escalation. This study evaluated LC and recurrence patterns in patients administered hepatic SBRT in more than 4 fractions due to dosimetric constraints.

Methods

This retrospective study included 33 patients treated with SBRT (Cyberknife®) in more than 4 fractions for HCC or liver metastases, between January 2011 and December 2019. Patients were ineligible for treatment in 3 or 4 fractions due to OAR proximity. Recurrence patterns were analysed according to the volume shared between recurrence and initial target or treatment isodose volumes.

Results

The primary dose ranged from 35 to 50 Gy delivered in 5 to 7 fractions for the treatment of HCC (39%) or liver metastases (61%) mainly secondary to colorectal cancer (40%).

LC rate was 64%, with 12 patients showing recurrence volume overlap with the initial target volume or treatment isodose. In-field recurrence occurred in only 12.5% of patients with most relapses being out-of-field. No grade ≥ 3 events were reported.

Conclusion

Despite dose reductions to spare OAR, SBRT showed satisfactory LC with low toxicity. Out-of-field recurrence remains the most common pattern identified and likely related to underlying disease. Prospective data are necessary to determine whether preserving dose while reducing planning target volume (PTV) coverage could enhance LC.

Trial registration

All patients were retrospectively registered, and this study is registered at the Health Data Hub site (number HDH414).

Background

Stereotactic Body Radiation Therapy (SBRT) and other local treatments such as radiofrequency ablation (RFA) or microwave ablation (MWA) are safe options for first-line curative treatment of hepatocellular carcinoma (HCC) [1, 2] in inoperable patients [3], although surgery remains the preferred option where possible [4, 5].

Global changes in the management of patients with oligometastatic disease have led to an increase in the use of SBRT for liver metastases [6].

The ideal dose and fractionation for liver-SBRT remains unclear although most recommendations agree on a prescription to the 80% isodose line from 30 to 50 Gy delivered in 3 to 5 fractions [5, 7].

Several studies have shown a dose–response relationship for SBRT in HCC with better response rate and overall survival (OS) when using biological effective dose (BED₁₀) ≥ 75 Gy [8] and equivalent dose in 2 Gray fractions (EQD2) ≥ 74 Gy [9]. For liver metastases this dose–response relationship has been demonstrated for BED₁₀ ≥ 100 Gy [10] with possible variations according to histological origin.

Fractionation changes are often required to comply with dose restrictions for organs at risk (OAR), especially hollow ones including the duodenum, stomach, or intestine. This type of dose/fractionation adjustment ensures safe treatment with acceptable levels of toxicity [11, 12]. However, a decrease in EQD2 and BED could worsen the local control (LC).

No consensus has been reported on volume coverage and whether it is preferable to maintain a high ablative dose in a smaller volume rather than changing the fractionation. One study found no association between the planning target volume coverage and better OS, progression free survival (PFS) or LC [13] .

To date, few studies have examined the pattern of recurrence after SBRT [14, 15] and none have involved patients treated with more than 4 fractions.

The aim of this study was therefore to evaluate the LC and describe the pattern of failure after hepatic SBRT in patients with primary or secondary liver cancer treated in more than 4 fractions due to dosimetric constraints.

Materials and methods

Study design and patient selection

This observational retrospective study was conducted in our centre and included all patients treated with SBRT for HCC or liver metastases, delivered in more than 4 fractions, between January 2011 and December 2019. All patients enrolled were ineligible for treatment with 45 Gy delivered in 3 fractions or with 40 Gy in 4 fractions with a tumour coverage ≥ 95%, due to the proximity of OAR and department protocols.

A multidisciplinary committee, comprising at least one radiation oncologist, surgeon, hepatologist, and interventional radiologist, reviewed and validated all treatment indications as well as contraindications to alternative local treatments. Additionally, a dedicated radiotherapy board approved the treatment dose.

This study was approved by the National Commission for Information Technology and Civil Liberties (CNIL) and registered with the French data protection authority.

Treatment

SBRT was delivered using the Cyberknife® system (Accuray Inc.). Gold fiducial markers were implanted percutaneously into the liver parenchyma near the tumour under computed tomography (CT) or ultrasound guidance. Real time tumour tracking was performed using the Synchrony Respiratory Tracking System. If markers could not be implanted, motion tracking was achieved using the Xsight Spine tracking system. Dedicated contrast-enhanced CT with 1 mm-thick slices, and magnetic resonance imaging (MRI) were performed in supine position 7 to 10 days after fiducial implantation.

Gross tumour volume (GTV) was delineated by the radiation oncologist and defined as the T1-enhanced tumour using image fusion of contrast planned CT and MRI. An isotropic margin was added to define the clinical target volume (CTV) and then the planning target volume (PTV).

OAR were delineated and could include healthy liver parenchyma, stomach, duodenum, intestine, colon, oesophagus, lungs, kidneys, spinal canal, and heart. The organs delineated depended on the topography of the lesion. Other concomitant local treatments were allowed.

To compare protocols differing in terms of dose prescription or fractionation, we calculated the BED₁₀ and the EQD2 using respectively the BED formula: BED = D x (1 + d/αβ) and the EQD2 formula: EQD2 = D × ((d + αβ)/(2 + αβ)); where D is the total dose, d is the dose per fraction and αβ is the α/β ratio of the tumour (defined at 10 Gy) [16] .

Follow-up

Patients were followed by the radiation oncologist, hepato-gastroenterologist, or medical oncologist every 3 months for the first two years, then every 6 months. Assessments included blood analysis every 3 months as well as Child [17] Pugh [18] score assessment for patients with cirrhosis. CT or MRI imaging was performed at 3–6 months to assess local and distant control. Follow-up images were analysed by an experienced radiologist.

Data collection

Demographics, clinical data, and treatment characteristics were retrieved from medical records. Acute toxicity was assessed according to the Common Terminology Criteria for Adverse Events (CTCAE) v5.0.

Regarding the recurrence pattern, data concerning all previous treatments were loaded into RayStation© (v11B, RaySearch Laboratories AB). Local recurrence (LR) was defined as progression of the treated lesion or development of a new intrahepatic lesion. Several volumes were identified or created by a radiation oncologist:

• initial GTV (GTVi),

• initial CTV (CTVi),

• initial PTV (PTVi),

• the volume receiving 80% of the prescribed dose (D_p) (V80%),

• the volume receiving 70% of the D_p (V70%),

• the volume receiving 50% of the D_p (V50%).

All follow-up CT scans or MRI of patients with LR were analysed: volumes of recurrence (GTVr) were then delineated using image fusion between the initial planning CT and imaging of the recurrence.

The overlap between the GTVr and the different isodose previously created (V80%, V70%, V50%) was analysed. For a patient with multiple relapses, the volumes of all lesions were delineated and compared to the initial treatment volumes. These different volume delineations are represented in Fig. 1.

Fig. 1

Volumes delineation. A, Planning-CT: initial Gross Tumour Volume (GTVi: red), initial Clinical Target Volume (CTVi: magenta), initial Planning Target Volume (PTVi: dark blue), isodose 80% (light blue), isodose 70% (cyan), isodose 50% (green). B, Recurrence MRI: volumes of recurrence (GTVr: red). C, Recurrence MRI: shared volume GTVr-PTVi (filled brown), shared volume GTVr-isodose 80% (filled yellow), GTVr (red), PTVi (dark blue), isodose 80% (light blue), isodose 70% (cyan), isodose 50% (green)

In accordance with various publications in other malignancies [19,20,21], recurrence patterns were defined as an in-field relapse if 80 to 100% of the GTVr overlapped with the different isodose volumes, as a marginal relapse if 20 to 80% of the GTVr overlapped, and as an out-of-field relapse if less than 20% of the GTVr overlapped. Unshared volume was defined as the development of a new lesion with no volume with the GTVi.

Study endpoints

The primary endpoint was the LC defined as the lack of recurrence within the initially treated volume. The secondary endpoints were pattern of recurrence, progression free survival (PFS), intrahepatic progression free survival (IHPFS), in-field recurrence free survival (IRFS) and acute toxicity.

Statistics

Qualitative parameters were described by frequency and percentage, and quantitative parameters by median and interquartile range. Survival analysis and tumour control were described using the non-parametric Kaplan–Meier method from the first day of treatment by SBRT. PFS was defined as the time from first day of treatment by SBRT to clinical relapse or death whichever occurred first; IHPFS as the time from first day of treatment by SBRT to either intrahepatic relapse or death whichever occurred first; and IRFS as the time from first day of treatment by SBRT to the first occurrence of in-field recurrence or death whichever occurred first. Analyses were performed with RStudio Software, version 2022.07.2 + 576.

Results

Patients

Thirty-three patients were treated with liver SBRT delivered in more than 4 fractions between January 2011 and December 2019 in our centre. The target lesions were HCC in 39% of cases and liver metastases in 61%, the latter or which the most common primary tumour origin was colorectal (40%), followed by breast, gastric, endometrial, head and neck, sarcoma, and other. Twenty patients (61%) had a history of previous liver treatment, most commonly surgery (60%). At the time of treatment, 94% of patients had no extrahepatic disease and 6% had 1 to 4 other extrahepatic lesions. Patient characteristics are described in Table 1.

Table 1 Patient characteristics

Treatment characteristics

Treatment characteristics are presented in Table 2.

Table 2 Treatment characteristics

The median lesion size was 24 [19—35] mm. The lesion had already been treated (surgery, transarterial chemoembolization (TACE), RFA or previous SBRT) in 27% of cases. The total dose prescription ranged from 35 to 50 Gy delivered in 5 to 7 fractions, with the main prescription scheme being 35 Gy delivered in 5 fractions, as administered to 49% of patients (BED₁₀: 59.5 Gy; EQD2: 49.58 Gy). Median GTV was 9 [4 – 22] cm3. Median PTV coverage was 96.7% [93.5—99.2] with 13 (39%) having less than 95% coverage. Dose restrictions due to OAR were not met in 61% of planned treatments. Such dose restrictions were more frequently observed in the duodenum and stomach where respectively 27% and 24% of the planned treatments exceeded the recommended dose limits.

Treatment safety

Toxicities are presented in Table 3.

Table 3 Acute toxicity and biochemical changes within 90 days of liver SBRT

No grade ≥ 3 treatment-related toxicity was reported. Only 4 (12%) patients experienced grade 2 nausea or vomiting. Among the patients with cirrhosis (n = 12), biological follow-up at 90 days showed no transaminase or alkaline phosphatase elevations greater than grade 1. Half of these patients showed no change in their Child Pugh score, and two (17%) showed an increase of 3 points. Three patients (25%) progressed to a Child–Pugh score greater than B7 within three months following treatment with CyberKnife.

Local control

LR and recurrence patterns are shown in Table 4.

Table 4 Hepatic recurrence

A total of 32 intrahepatic recurrences occurred in 22 of the 33 patients (67%), of which 8 (36%) were within the initially treated liver segment.

The median time from SBRT treatment to first intrahepatic recurrence was 6 [4 – 15] months. The median GTVr was 3.77 [2.07 – 9.56] cm3. Shared volume relapse occurred in 12 (36%) patients, 10 (30%) for whom it was overlapping with PTVi.

Among the 32 recurrences, 20 (62.5%) GTVr did not overlap with V80% isodose, 3 (9.4%) were considered out-of-field with < 20% shared volume, 5 (15.6%) were considered marginal, and 4 (12.5%) were in-field with more than 80% shared volume.

Six of the liver recurrences overlapping with the treatment isodose were HCC treated with BED₁₀ < 75 Gy and EQD2 < 74 Gy. Five other recurrences were colorectal metastases with two lesions treated with BED₁₀ < 100 Gy and three lesions BED₁₀ = 100 Gy. One recurrence was cholangiocarcinoma treated with BED₁₀ = 52.5 Gy.

One of these relapses occurred following treatment delivered in 6 fractions and the others following treatment delivered in 5 fractions.

Figure 2 (A, B, C) shows PFS, IHPFS and IRFS.

Fig. 2

PFS, IHPFS and IRFS. A—Progression Free Survival (PFS); 2.B—Intrahepatic Progression Free Survival (IHPFS); and 2.C—In-field recurrence free survival (IRFS); with red points indicating data at 6 months, 1 year, and 2 years post-treatment

Discussion

This retrospective study collected data from patients treated in our centre over a period of nearly ten years, for liver metastases or HCC with SBRT delivered in more than 4 fractions due to OAR proximity. Despite a reduction in the total prescribed dose, which resulted in lower BED10 and EQD2 values, LC remains satisfactory.

This is the first study to accurately delineate liver recurrence volumes on follow-up imaging. This delineation enabled us to compare recurrence volumes with treatment isodoses and thus better define and describe LR.

Our study showed a LC rate of 87.5% for the treated lesions, based on the number of recurrences sharing more than 80% of their volume with the V80% isodose. This LC rate decreases to 71.9% when considering recurrences sharing more than 20% volume with the V80% isodose.

Previous studies have assessed lesion response on follow-up imaging but without volume delineation. Our results are in accordance with these studies. Kwon et al. found an initial lesion response followed by an in-field progression at 2.5 years of 28.6% after SBRT treatment in 3 fractions [14], while Joo et al. found 32% in-field recurrence at 3 years following delivery in 3 to 4 fractions [15]. The slightly lower LC in our study can be explained by the lower BED₁₀ and EQD2 in our study due to the delivery of treatments in more than 4 fractions.

However, the definition of in-field recurrence was not the same in our study as in the two earlier studies. Indeed, Kwon et al. defined in-field response as “target lesion response within the irradiated field”, while Joo et al. defined local failure as a “new nodular or irregular enhancing lesion in the previous tumour site", excluding new out-of-field liver metastases.

Our study found a LC rate of 64% based on 36% of LR overlapping with the initial volumes and isodose treatment, regardless of the percentage of shared volume.

One meta-analysis published in 2019 indicated a 2-year LC rate of 83% for both HCC and liver metastases when treated with SBRT delivered in 3 or 4 fractions, BED₁₀ > 93 Gy and EQD2 > 78 Gy [22]. Another literature review on HCC treated with SBRT delivered in 4 to 6 fractions reported a LC rate ranging from 73 to 98% at 3 years, with BED₁₀ between 51 and 180 Gy and EQD2 between 42 and 150 Gy [23]. For liver metastases LC rate ranged between 67 and 95% in prospective studies and from 58 to 100% in retrospective studies, with the most frequently reported BED₁₀ at 112.5 Gy [24].

Our LC rate is slightly lower than that observed in these trials, which can be attributed to the dose reduction necessary to comply with the dose constraints for the OAR in our patient population. We reported BED₁₀ values ranging from 52.5 to 100 Gy, with a most commonly prescribed dose of 59 Gy, and EQD2 values ranging from 43.75 to 93.33 Gy, with most commonly prescribed dose of 49.58 Gy.

Also, the definition of LC in published papers is not always clear or precise: are we considering in-field progression or marginal progression? What is the definition of "in-field" progression: a lesion in the same liver segment or sharing volume with the initial treatment volumes? This may have influenced the interpretation of the results, and comparisons between different studies should be made cautiously.

As described in the literature [3, 8,9,10], we observed that lower BED₁₀, for HCC or liver metastases, was associated with lower LC and recurrence occurring within the initial treatment isodose volume.

While some studies have reported a potential association between tumour volume and LC at various cut-off points [3, 14] we found no such association in our study, possibly due to the majority of lesions studied being smaller than 5 cm.

According to the departmental protocol, all patients treated during the recruitment period were scheduled to receive standard treatment delivered in 3 fractions. Only when this failed to comply with the dosimetric constraints due to OAR, the delivery in 4 or more fractions was then planned to reach compliance and maintain PTV coverage as high as possible.

Even at the lower prescription dose, we found some non-compliance with these constraints due to extreme proximity between the target lesion and the OAR, especially in the duodenum or stomach. Fortunately, these non-compliances were minimal and resulted in no significant toxicity compared to published data.

There are limitations to this work. First this study remains a retrospective cohort with a small number of patients and a heterogeneous population including HCC and liver metastasis regardless of the initial histology.

Moreover, image fusion between planning CT and MRI or CT of recurrence was not always easy, with changes in liver parenchyma anatomy due to cirrhosis, intercurrent local treatment or different respiratory phase. The volumes delineated on CT scans are also less suitable than those delineated on MRI. To increase the validity of our data, image fusion and recurrence volumes were reviewed by two radiation oncologists and all recurrences were identified by a radiologist. If more than one recurrence was found on the same follow-up image, we chose to delineate the closest recurrence to account for the worst-case scenario.

Three patients were treated without fiducial markers using Spine tracking prior to 2015, due to medical contraindications or tumour locations deemed close enough to the vertebrae. These patients did not benefit from PTV margin adjustments in contrast to the other patients treated with fiducial markers. However, these patients did not appear to have a consequentially higher recurrence rate than the others: two patients showed no recurrence during their follow-up, while one patient experienced a marginal recurrence at the 80% isodose.

Given the heterogeneity of the dose delivered in SBRT, it is reasonable to hypothesise that maintaining a higher ablative dose within the PTV may, even with lower coverage, allow satisfactory and in particular safe treatment of the target lesion, thanks to a steep dose gradient sparing the OAR. In a secondary analysis of the SABR-COMET trial, Van Oirschot et al. [13] suggested that PTV coverage was not associated with improved OS or LC. These findings were confirmed in a recent secondary analysis of the SABR-5 trial [25] in oligometastatic patients with prioritisation of OAR in a larger cohort.

Further prospective data are now needed to determine whether prioritising a high ablative BED₁₀/EQD2 is more beneficial than ensuring maximum coverage of the PTV at all costs. This is essential to ensure satisfactory control of the target lesion while sparing the nearby OAR.