Keywords contrast-enhanced intraoperative ultrasound (CE-IOUS) - liver surgery - liver cancer
- colorectal liver metastases (CRLM)
Introduction/Background
Surgical resection, following precise lesion localization, remains the most promising
cure for patients with primary liver cancer or colorectal liver metastases (CRLM).
Systemic chemotherapy, when used in conjunction, leads to optimal treatment success.
Despite its limited sensitivity of 85%, many national guidelines still recommend contrast-enhanced
computed tomography (CT) as the primary imaging modality for preoperative diagnostic
evaluation of the liver [1 ]
[2 ]. Magnetic resonance imaging (MRI), especially when using diffusion-weighted sequences
and liver-specific contrast agents, is currently considered the highest resolution
imaging modality with a sensitivity of up to 98% [3 ]. However, its routine use for preoperative staging is hindered by limited availability,
complex application, and associated costs. Detecting liver metastases smaller than
10mm is challenging for CT and MRI, as shown by their remarkably low sensitivity of
47.3% and 60%, respectively [4 ]. Contrast-enhanced intraoperative ultrasound (CE-IOUS) has a primary advantage in
detecting these “occult” lesions [5 ] and is increasingly used in clinical practice [1 ]. Although several clinical studies have highlighted the superior sensitivity and
specificity of CE-IOUS compared to preoperative imaging [4 ]
[6 ], it remains unclear how this additional intraoperative diagnostic step affects the
immediate surgical strategy and long-term outcome. The present prospective observational
study is designed with the primary aim of answering these questions regarding the
clinical relevance of CE-IOUS.
Materials and Methods
Participants and study protocol
This prospective, monocentric, non-interventional study was approved by the institutional
ethics committee (ID: XXX) and registered at clinicaltrials.gov (NCT-XXX). All patients
who underwent elective liver surgery with CE-IOUS within the three-year study period
were included for analysis after informed consent was obtained. CE-IOUS was solely
indicated based on the subjective experience of the responsible hepatobiliary surgeon.
The study team had no influence. The reason for liver surgery was not decisive. Both
benign and malignant primary liver tumors as well as liver metastases were included.
Liver transplantations and the use of only CE-IOUS to confirm the absence of liver
metastases led to exclusion.
Preoperatively, almost all patients received a contrast-enhanced CT scan. Additional
transabdominal CEUS and/or MRI scans were carried out on select patients. This decision
was made by the multidisciplinary tumor conference or by the lead surgeon. The surgical
approach was planned and documented in advance of the operation. All operations in
this study were performed using an open technique. Exploratory laparotomy was followed
by CE-IOUS. This was performed under sterile conditions jointly by an experienced
hepatobiliary surgeon (guidance of the probe) and an experienced radiologist (Degum
III, interpretation of the findings) using a high-end ultrasound platform (Logiq E9;
GE Healthcare, Chicago, IL) equipped with a T-probe (3–9 MHz) to reach all areas of
the liver. For contrast enhancement, 2.4 ml of sulfur hexafluoride microbubbles (SonoVue,
Bracco, Italy) were injected via central venous access, followed by a 10-ml saline
bolus. CE-IOUS was performed systematically over the complete liver and, if necessary,
the contrast bolus was repeated. For dynamic characterization, the examination was
performed for at least 5 min. after the flush. The arterial (15–45 s), portal venous
(45–90 s) and late venous phases (> 2–5 min.) of each detected lesion were recorded
as cine-loops. In addition, each lesion was measured (in cm, 2 axes) and localized
(liver segment) and its status was assessed (benign/malignant). This allowed correlation
of preoperatively known lesions as well as the detection of small additional lesions
and, based on this, the planning of the surgical strategy. An experienced pathologist
examined all resection specimens histopathologically. The results were regarded as
the gold standard for the correlations made in this study. Further technical details
about imaging and histological examination are outlined in the supplementary methods .
The main objective of this study was to test the hypothesis that intraoperative CEUS
alters the immediate surgical strategy in the treatment of liver tumors. Secondary
endpoints included evaluating the sensitivity and specificity of CE-IOUS compared
to preoperative imaging, determining preoperative clinical parameters to predict the
strategic impact of CE-IOUS, and assessing its oncological significance. For the latter,
the recurrence-free survival (RFS) of study patients with HCC, CCC, or colorectal
liver metastases was compared to a control population consisting of patients who underwent
surgery without CE-IOUS. The following scoring system was used to classify the complexity
of surgery 1=left hemihepatectomy; 2=atypical resection/segmentectomy; 3=right hemihepatectomy;
4=complex multi-stage surgery like in situ split or resection with simultaneous ablation;
5=termination of surgery without curative resection.
Surgeon survey
In a multiple-choice survey, experienced hepatobiliary surgeons were asked about their
reasons for indicating a CE-IOUS examination in liver surgery. Survey respondents
were also asked to rate preoperative clinical parameters in predicting a CE-IOUS-related
change in strategy.
Data collection and statistical analyses
All imaging procedures were documented in writing with regard to number, size (in
cm, 2 axes), location within the liver areas (right lateral, right central, left central,
and left lateral), and status classification (benign/malignant) of delineable lesions.
To evaluate the effect of CE-IOUS on the surgical strategy, the preoperatively planned
approach was compared with the final procedure performed.
RFS rates were analyzed using the non-parametric Kaplan-Meier method and a log-rank
test. To develop a prediction model, we used a random forest with either all features
or selected known significant preoperative parameters. The model was trained in leave-one-out
cross-validation and the predictions of each left-out sample were presented. A two-sided
P<0.05 (95% confidence interval) was considered statistically significant. Statistical
analyses were calculated with the following software programs using databases managed
in Microsoft Excel (Office-365): SPSS (v.28), GraphPad Prism (v.9), and R (v.4.0).
Results
Study population
During the three-year study period, 519 elective liver operations were performed in
our hospital, with 206 patients (~40%) being prospectively included due to a CE-IOUS
examination (supplementary Fig. 1a–b ). The patients were evenly distributed throughout the study. The study cohort is
characterized by a broad cross-section of gender, age, underlying disease, and tumor
burden (supplementary Fig . 1c–e ). Male patients predominated with 67% (n=138/206). The mean age was 64 years (range:
10–85). The diagnoses were liver metastases (n=93), benign liver lesions (n=13), hepatocellular
carcinoma (n=60), and intrahepatic cholangiocellular carcinoma (n=34).
Pre- and intraoperative imaging results
Preoperative CT imaging was performed in 97.5% of cases (n=201/206). In the remaining
five cases, consisting of young patients with benign disease, only an MRI scan was
performed to reduce radiation exposure. Additional preoperative MRI (68%; n=140/206)
or CEUS (56%; n=116/206) examination was requested by the lead surgeon and/or tumor
board when deemed necessary for optimal planning. In the study cohort, all surgeons
used CE-IOUS ([Fig. 1 ]a).
Fig. 1 Pre- and intraoperative imaging results. a Circular diagrams showing the percentage of CT imaging (black outer ring) and complementary
imaging modalities in terms of preoperative MRI (1st inner grey ring), preoperative
CEUS (2nd inner grey ring) and CE-IOUS (3rd inner grey ring). b Heatmap showing the number of detected lesions in each imaging method based on color
code (from red = 0 lesions to dark green = 5 lesions). If the examination method was
not carried out, the mean number of the other methods was plotted. c Color-coded visualization of the correct location of liver lesions in the different
imaging methods compared to the histopathological findings. The following liver areas
were defined: right lateral (seg. VI/VII), right central (seg. V/VIII), left central
(seg. IV + I) and left lateral (seg. II/III). Imaging methods not performed are shown
in grey. d Correlation of the assessment as a benign or malignant lesion by the imaging modalities
with histological findings. The shaded column corresponds to the proportion of incorrect
diagnoses. The p-values were calculated using a paired, one-way ANOVA test and corrected
according to Holm-Šídák.
Preoperative CT had the lowest sensitivity in terms of the absolute number of lesions
detected. No liver lesions were detected in 6.0% of CT images (n=12/201), regardless
of their presence on other imaging modalities. However, this false-negative result
of liver tumor clearance occurred in only 0.7% (n=1/140), 3.4% (n=4/116), and 1.0%
(n=2/206) of MRI, CEUS, and CE-IOUS examinations, respectively. Finally, MRI and CE-IOUS
showed higher sensitivity in detecting multiple intrahepatic tumors compared to CT
and preoperative CEUS ([Fig. 1 ]b). The location of lesions is of particular importance for surgical planning. The
basis for this evaluation was the clinically simplified mapping into four liver areas:
right lateral (seg. VI+VII), right central (V+VIII), left lateral (seg. II+III) and
left central (seg. IV). CE-IOUS was clearly superior to all preoperative methods as
shown in [Fig. 1 ]c. CT (7.0%; n=56/804) again revealed the highest number of incorrect results, but
at an equal level to preoperative MRI (5.0%; n=28/564) and CEUS (7.3%; n=34/464).
With regard to distinguishing whether the lesion is benign or malignant according
to the final histopathological outcome, preoperative CT showed the most specificity
errors (9.45%) compared to preoperative MRI, CEUS, and CE-IOUS (5.71% vs. 5.17% vs.
3.88%) ([Fig. 1 ]d).
Impact of CE-IOUS on surgical strategy
Supplementary Fig. 2a summarizes the impact of CE-IOUS on surgical strategy, defined by the following categories:
(1) If only uncertain preoperative assumptions were confirmed, patients were classified
in the “partial” group. (2) If the CE-IOUS result led to a change in the extent of
resection, defined as a shift in the resection line, patients were classified as “yes-minor”.
(3) “Yes-major” included patients for whom the CE-IOUS finding led to a complete change
in surgical approach. A classification was made if the change in strategy was based
solely on the ultrasound findings and not, even partially, on the surgeon’s clinical
findings. Overall, the preoperative plan was carried out as planned in 57% (n=118/206)
of operations. The surgical procedure was modified in 9% (n=18/206) due to intraoperative
surgical assessment and in 34% (n=70/206) due to CE-IOUS findings. Of the latter,
22 cases were classified as “partial”, 13 as “yes-minor” and 35 as “yes-major”.
The characteristics of the “yes-major” subgroup are summarized in detail in supplementary Fig. 2b . This group was 68% male (n=24/35) and had a mean age of 67 years (range: 49–80).
In total, seven patients suffered from cholangiocellular carcinoma (20% of all CCC;
n=7/34), ten from hepatocellular carcinoma (17% of all HCC, n=10/60) and 14 from colorectal
liver metastases (18% of all CRLM, n=14/76). CE-IOUS findings that led to a change
in surgical strategy were located in the left lobe of the liver in 66% of cases (n=23/35)
and in the right lobe in 23% (n=8/35). In the remaining 11% (n=4/35), lesions were
found bilaterally throughout the liver.
The results of CE-IOUS changed the surgical approach in a variety of ways (changing
the extent of resection, combining different techniques, or canceling surgery). In
20% of cases (n=7/35), intraoperative radiofrequency ablation (RFA) was performed
instead of resection alone. Larger resections were necessary in 42.8% of cases (n=15/35),
a reduction of the surgical extent was possible in 5.7% (n=2/35), and the operation
was cancelled in 31.4% (n=11/35). Case no. 34 is worth mentioning. In that case, surgery
was canceled because CE-IOUS classified the lesion as benign, contrary to the results
of preoperative imaging. Intra-operative frozen section analysis of a biopsy confirmed
the ultrasound assessment. In case no. 9, the planned right hemihepatectomy in combination
with RFA was changed to a complex in-situ split approach with primary portal vein
ligation for a second stage, extended hemihepatectomy (ALPPS) after hyperplasia of
the tumor-free lobe.
The correctness of these major strategy changes was determined by the histopathological
results (supplementary Fig. 2c ). CE-IOUS findings were confirmed by pathologists in 71.4% of cases (n=25/35), while
confirmation was not possible in 11.4% (n=4/35) due to missing specimens following
RFA. No vital tumor could be found in 20% of cases (n=7/35), as estimated by CE-IOUS.
Whether this should always be considered a “false positive” remains questionable.
An avital scar can also be interpreted as the result of an excellent response to neoadjuvant
chemotherapy. In one case there was both a confirmed correlation and a false-positive
correlation.
To exemplify the added diagnostic value of CE-IOUS, the case of a 60-year-old patient
(case no. 4, supplementary Fig. 2b ) suffering from metachronous CRLM is presented in [Fig. 2 ]. Preoperative staging using CT and MRI scans revealed extensive metastasis throughout
the right lobe involving segments V, VI, VII, and VIII with infiltration into segment
IVa ([Fig. 2 ]a). Segment I and the left lateral lobe were tumor-free. CE-IOUS identified two additional
tumor lesions in segments II and III. Based on the characteristic contrast dynamics,
these lesions could be classified as typically metastatic ([Fig. 2 ]b–d ). Accordingly, two atypical resections were added to the planned right hemihepatectomy.
Histopathological examination confirmed these findings.
Fig. 2 Exemplary case study of patient with yes-major change in surgical strategy. Illustration
of the 4th case of Figure 3c. A 60-year-old man suffering from colorectal liver metastases.
Preoperatively, lesions only in the right lobe of the liver were detected by CT and
MRI scans a . Two small metastases in segments II and III were diagnosed additionally during CE-IOUS.
The typical contrast behavior of CRLM is characterized in the early arterial phase
with hyperperfusion b , beginning or complete washout in the portal venous c and late phase d . Representative metastases in the right lobe are visible on a CT scan in the coronal
section as hypodense lesions in the portal-venous phase and on an MRI scan as hyperintense
lesions in the contrast-enhanced T1 sequence.
Development of a predictive model to estimate CE-IOUS impact
Given that CE-IOUS changed the surgical procedure in 34% of cases, we next asked whether
this impact could have been predicted by clinical parameters. Unnecessary examinations
could thus be avoided and all patients who may benefit from CE-IOUS could receive
it.
The first step in developing a preoperative algorithm for CE-IOUS indication was a
survey of twelve experienced hepatobiliary surgeons (supplementary Fig. 3a ). According to the surgeons, the most important factors in determining the indication
were the number, location, tumor entity, and classification of the tumors as benign
or malignant, as well as the time elapsed since the last preoperative imaging. The
availability of preoperative MRI and tumor stage were of secondary importance, while
age, gender, and tumor markers were considered unimportant.
To objectify the experience-based indication, a statistical correlation of these parameters
with the potential impact of CE-IOUS was performed (supplementary Fig. 3b–d ). Univariate analysis confirmed the predictive significance of the number of lesions
detected by CT or MRI as well as tumor stage. Considering this, it seemed promising
to develop a decision chain using a random forest algorithm. Unfortunately, this modelling
of a predictive algorithm was unsuccessful using either all parameters (supplementary Fig. 3e ) or only those that showed statistical correlation in the non-parametric calculation
(supplementary Fig. 3f ). The insufficient discriminatory power is reflected in AUC values of 0.573 and 0.575,
respectively. In conclusion, preoperative patient characteristics routinely used by
clinicians do not statistically predict whether and to what extent CE-IOUS will lead
to a change in surgical strategy.
Evaluating the oncological outcome relevance of CE-IOUS
Finally, we analyzed the relevance of the CE-IOUS-related strategy changes with respect
to oncological outcomes. Using a prospective case-control observational study design
(supplementary Fig. 4a ), the RFS of the intervention cohort (IC) was compared with that of a control cohort
(CC) in which liver surgery was performed without CE-IOUS. Patient characteristics
for both cohorts are summarized in supplementary Fig. 4b . They differ significantly with respect to their tumor burden (tumor grading, number
of lesions, affected liver areas) and in the complexity of the operations performed.
Interestingly, despite the expected poorer prognosis, the RSF was identical in both
cohorts (log rank: p=0.672; supplementary Fig. 4c ).
Subgroup analyses by tumor type again showed a significant prognostic relevance of
CE-IOUS. Equivalent or even slightly superior RFS was demonstrated in IC patients
with HCC (supplementary Fig. 4d/g ) or CRLM (supplementary Fig. 4f/i ), even in significantly advanced disease. However, IC patients suffering from CCC
(supplementary Fig. 4e/h ) relapsed significantly earlier and more often (p=0.03). This can be explained by
the significantly higher UICC stages, which were mostly associated with lymphogenic
metastases.
Discussion
The present study was able to demonstrate a significant impact of CE-IOUS using a
high-end ultrasound platform with T-probe for sterile intraoperative use in liver
surgery. The examinations were performed jointly by a liver surgeon and an experienced
sonographer (DEGUM III) in a multidisciplinary approach, ensuring the availability
of the whole range of skills. Changes in surgical strategy occurred in more than 30%
of cases despite using multimodality imaging preoperatively. Contrary to these results,
the implementation of CE-IOUS as an additional intraoperative diagnostic tool is rarely
recommended in international guidelines [7 ]
[8 ]. The German guideline for colorectal carcinoma even explicitly rejects its use because
of the unjustifiable expense [2 ]
[9 ].
In line with the published literature, the superior sensitivity and specificity of
CE-IOUS over preoperative imaging has been confirmed with respect to the detection
of liver lesions, especially compared to the exclusive use of a CT scan [5, 10, 6,
11]. Only MRI with liver-specific contrast agent has comparable detection rates. The
meta-analysis by Chen et al. is also consistent with these findings. In the ten included
studies with patients suffering from CRLM, the pooled sensitivity and specificity
of CE-IOUS were 0.96 (95% CI: 0.95–0.97) and 0.75 (95% CI: 0.70–0.80), respectively
[10 ]. Apart from improving the proportion of complete resections, the intraoperative
visualization of the spatial relationship between lesions and blood vessels or biliary
structures improves postoperative outcome by reducing intraoperative blood loss and
postoperative bile leak complications [5 ]
[11 ]. A limiting factor for the objective interpretation of our study results is the
surgeons’ knowledge of the preoperative imaging results when performing CE-IOUS. Despite
the outstanding specificity of CE-IOUS, the results did not correlate with the histological
results in 3.88% of cases (n=8/206). Even though CE-IOUS classified these cases as
malignant, no vital tumor was detectable pathologically. As discrimination of a benign
from a malignant tumor on CEUS is mainly based on irregular vascularization with hyperenhancement
in the arterial phase and washout in the portal venous and late phases, the differentiation
between benign atypical adenomas, partially thrombosed hemangiomas, regeneration nodules,
and HCC is particularly difficult [12 ].
Although several studies have shown the ability of CE-IOUS to detect lesions unknown
preoperatively [13 ]
[14 ], the resulting implications for the surgical approach during liver surgery have
usually only been a secondary endpoint. For example, Torzilli et al. [9 ] and Arita et al. [15 ] described a CE-IOUS-related change in strategy for CRLM patients in 18% and 10%
of cases, respectively. A German [16 ] as well as a Chinese [17 ] research group demonstrated an impact in 27 (54%) and 9 (18%) patients in each of
the 50 patients suffering from HCC. The diagnostic benefits of CE-IOUS must be weighed
against the additional time, economic effort, and examination risks. The latter are
very low, with an incidence of 0.020% for any side effects and 0.007% for serious
side effects.
To the best of our knowledge, the present study is the first evaluating the oncological
impact of CE-IOUS on RFS in patients with HCC, CCC, or CRLM. Impressively, despite
a significantly higher tumor burden in the intervention cohort, a comparable oncological
outcome was observed in both groups. In the subgroup analysis for patients undergoing
liver resection for HCC, an even higher RFS was observed in the IC. This difference
becomes apparent after twelve months and can probably be explained by small HCC satellite
nodules that escape preoperative imaging and thus lead to a shift in the resection
plane. Another possible explanation is the additional discovery of sub-centimetric
HCC foci, especially in fibrotic livers, which are hardly delineated by CT and difficult
to delineate by MRI. For patients with CCC, no oncological advantage could be demonstrated
for the use of CE-IOUS. This is understandable considering the increased risk of recurrence
associated with locoregional lymph node metastases. Therefore, complete resection
of preoperatively undetected intrahepatic lesions cannot extend RFS [18 ]. An oncological advantage could be achieved in patients suffering from CRLM using
CE-IOUS-guided, more complex surgical techniques. An almost identical RFS of the IC
could be observed even with a significantly higher tumor burden. Localizing CRLM,
especially after neoadjuvant therapy, is particularly challenging, as lesions that
“disappear” on imaging are not equivalent to an oncological cure at that site [19 ]. Intraoperative re-detection is an outstanding value of CE-IOUS for optimizing the
surgical approach. In very complex cases, fusion techniques based on cross-sectional
imaging prior to neoadjuvant therapy and CE-IOUS may help to relocate even sub-centimetric
residual metastases. A prospective clinical study published in 2021 underlines the
added value of CE-IOUS, especially for lesions smaller than 5 mm, even if both MRI
with liver-specific contrast agent and multiphasic CT were performed preoperatively
[6 ].
Conclusion
Complete resection remains the best curative therapeutic option for liver-derived
tumors and colorectal liver metastases. Despite advances in preoperative imaging,
the detection of small lesions remains a challenge, particularly in fibrotic or chemotherapy-treated
livers. It is precisely in these situations that the strength of CE-IOUS becomes apparent.
Its impact on surgical strategy, combined with a complex surgical approach, improves
RFS even in patients with high tumor burden.
