To obtain volumetric measurements, the hepatic hemangioma is outlined (A) and the remaining liver tissue is subtracted (B). Standard 3-dimensional volume is then used.
Overall (A) and annual (B) growth were assessed. Initial sizes predicted growth. The changes in linear dimension (R) were 0.617 and 0.594, respectively; P < .001.
Hemangiomas grew in a uniform fashion, with no association with the length of follow-up. The change in linear dimension (R) was 0.00843; P = .92.
Hasan HY, Hinshaw JL, Borman EJ, Gegios A, Leverson G, Winslow ER. Assessing Normal Growth of Hepatic Hemangiomas During Long-term Follow-up. JAMA Surg. 2014;149(12):1266-1271. doi:10.1001/jamasurg.2014.477
Few long-term data describe the natural history of hepatic hemangiomas. Because these lesions are frequently imaged repetitively on studies performed for other indications, health care professionals are commonly confronted with the problem of a growing hemangioma. Because the rate and magnitude of normal growth is not well characterized, it is difficult to recognize lesions growing at an abnormal rate, which may require further evaluation or intervention.
To establish quantitatively the expected growth rate of hepatic hemangiomas and to define a measure of hemangioma growth that could be used clinically to help identify hemangiomas for which growth is more than expected.
Design, Setting, and Participants
Retrospective cohort study at an academic hospital tertiary referral center evaluating the growth rate of hepatic hemangiomas on cross-sectional imaging studies during a 10-year period (1997-2007). The mean (SD) follow-up time was 3.7 (1.9) years. The radiology information system was searched in a 10-year period for hemangioma. Patients with hepatic hemangiomas that were 1 cm or larger as seen on cross-sectional imaging (computed tomography or magnetic resonance imaging), and 1 year or more apart were selected. Images with the longest interval between studies were selected for further review. Each study was rereviewed for diagnostic confirmation and to ensure consistency in measurement technique. Lesions were remeasured in 3 dimensions, and volumes were calculated using 3-dimensional software.
Main Outcomes and Measures
Primary outcomes include the fraction of hepatic hemangiomas that demonstrated growth during long-term follow-up and the annual growth rate of those lesions.
A total of 163 hemangiomas were identified in 123 patients. The mean (SD) initial size was 3.2 (3.1) cm. During follow-up, 39.3% of hemangiomas grew 5% or more in mean linear dimension. The mean (SD) annual linear growth rate was 0.03 (0.21) cm for all lesions and 0.19 (0.23) cm for those that grew 5% or more. By volume, 44.7% of lesions grew 5% or more. The mean (SD) annual volumetric growth rate was 2.8% (21.0%) for all lesions and 17.7% (22.8%) in those that grew 5% or more. The initial size predicted the growth in linear dimension and volume (P < .001). There was no significant change in growth rate over time, indicating uniform growth (R = 0.00843; P = .92).
Conclusions and Relevance
Nearly 40% of hepatic hemangiomas grow over time. Although the overall rate of growth is slow, hemangiomas that exhibit growth do so at a modest rate (2 mm/y in linear dimension and 17.4% per year in volume). Further research is needed to determine how patients with more rapidly growing hemangiomas should be treated.
The detection of asymptomatic benign liver lesions is increasing with more frequent use of abdominal cross-sectional imaging and poses numerous diagnostic and therapeutic challenges.1,2 Hepatic hemangiomas are the most common benign liver lesions, found in 0.7% to 7% of the general population, and are often discovered incidentally on abdominal imaging performed for other indications.1- 7 Traditionally, patients with hemangiomas are advised that these are clinically silent entities that require no further intervention or follow-up.8- 11
Despite this stance, a recent international, multi-institutional study of nearly 1900 patients who received resection of hemangiomas demonstrated that the indication for resection in 31% of patients was hemangioma growth.12 This finding clearly demonstrates that there remain important unresolved issues regarding the natural history of hepatic hemangiomas. Although hemangioma growth has not been well established as an accepted criterion for resection, the lack of data surrounding normal hemangioma growth rates continues to challenge patients and surgeons faced with a finding of a growing hemangioma.
Several previous studies have described the growth of as many as 17% of hepatic hemangiomas over time.7,11,13- 16 However, these studies are limited in that they are relatively small, mostly use ultrasound imaging, and describe the cumulative amount of change in 1 dimension during an undefined period. At present, no normative data are available for clinical use. As a result, imaging studies that reveal any increase in hemangioma size are described quantitatively, without the benefit of comparing an individual patient’s findings with the normal limits for the population.
The aim of this study was therefore 2-fold. First, we aimed to accurately establish the fraction of hemangiomas that grow during long-term follow-up using modern cross-sectional imaging and radiographic techniques. Second, we aimed to define a measure of hemangioma growth that could be used clinically to help identify hemangiomas for which growth is more than expected when compared with the population.
Before study initiation, approval from the University of Wisconsin’s Health Sciences Institutional Review Board was obtained; patient consent was not required. The text of computed tomography or magnetic resonance imaging reports at the University of Wisconsin was electronically searched for the term hemangioma during a 10-year period (1997-2007). Both the body of the report and the impression of computed tomography and magnetic resonance imaging scans were included. Patients included in this study were adults (≥18 years of age) who had at least 1 hepatic hemangioma measuring 1 cm or larger in its greatest dimension on 2 cross-sectional imaging studies obtained 1 year or more apart.
Images were reviewed by 2 experienced, board-certified radiologists (J.L.H. and E.J.B.) for diagnostic confirmation and to ensure that the same lesion was followed up serially. In addition, each lesion was remeasured linearly in 3 dimensions, and volumetric measurements were performed on a dedicated independent postprocessing workstation (GE VolumeShare; GE Medical). The hemangiomas were manually contoured with slice-to-slice interpolation and visual confirmation (Figure 1). In addition, data collected from the medical records included demographic data, comorbidities, indications for imaging, and relevant hepatic outcomes.
To describe hemangioma size and growth, we present changes both in absolute linear dimension (used typically in clinical practice) and relative linear dimension, expressed as the percentage of growth when compared with the initial lesion size. This growth is reported similarly for the volumetric data. Because 3 separate linear dimension measures are used to describe these lesions, we have chosen to report the mean linear dimension (determined by obtaining a mean of the 3 separate dimensions) for consistency.
We defined substantial growth as 5 mm or more in mean linear dimension or 5% or more in mean linear dimension or volume. This arbitrary cutoff is meant to reduce detecting small changes that may be attributed to measurement variability. Because mean measurements for the entire group are not useful clinically when some lesions demonstrated a decrease in size, we report growth rates for the total population and for those demonstrating growth as described above. We also describe growth rates as a function of initial mean linear dimension (1-3 cm, >3-5 cm, and >5 cm).
Continuous variables are presented as the mean (SD). We used t tests and correlation coefficients to examine relationships between patient characteristics and hemangioma growth. Analysis of variance was used to determine whether growth was associated with initial mean linear dimension categories. Fisher exact tests were used to examine relationships between patient characteristics, indicators of 5% hemangioma growth, and initial mean linear dimension categories. All analyses were conducted using the SAS statistical software, version 9.2 (SAS Institute Inc).
A total of 123 patients were identified, with 163 hepatic hemangiomas that met the inclusion criteria. We identified 150 hemangiomas on 2 imaging studies that were 1 year or more apart and were compatible with our 3-dimensional volume-measuring software. A moderate female preponderance was noted at 70 patients (56.9%), and the mean (SD) age at diagnosis was 53.4 (12.6) years (range, 28-81 years). Patients were followed up for a mean (SD) of 3.7 (1.9) years (range, 1.0-10.4 years). Most patients had a single hemangioma (74.8%), with 19.5% of patients having 2 lesions and 5.7% having 3 or 4 lesions. Furthermore, 76.1% of hemangiomas were in the right lobe of the liver (Table 1). Reasons for obtaining the initial image included abdominal pain (29.5%), evaluation of potential malignant neoplasms or follow-up of a known malignant neoplasm (39.3%), abnormal liver function test or known chronic liver disease (5.8%), or a variety of other indications for abdominal imaging (30.3%) (Table 1).
The initial mean (SD) linear dimension of the hemangiomas was 3.2 (3.1) cm (range, 0.9-25.3 cm), and the initial volume was 137.8 (701.4) cm3 (range, 0.7-6983.6 cm3) (Table 1).
Of the 163 hemangiomas followed up, 83 (50.9%) grew by any amount in absolute mean linear dimension, while 74 (45.4%) shrunk and 6 (3.7%) remained unchanged. Similarly, of 150 hemangiomas with volumetric measurements, 72 (48.0%) grew by any amount, while 77 (51.3%) shrunk and 1 (0.7%) remained unchanged. When evaluating lesions that grew by 5 mm or more and 5% or more, 45 (27.6%) of hemangiomas grew by 5 mm or more in 1 dimension, while 32 (19.6%) shrunk by 5 mm or more and the remaining 86 (52.8%) had an absolute change of less than 5 mm. Similarly, 64 (39.3%) hemangiomas grew by 5% or more in mean linear dimension, while 67 (44.7%) grew in volume by 5% or more.
The overall change in mean (SD) linear dimension for all lesions during the follow-up period was 0.13 (0.88) cm, with an annual growth rate of 0.03 (0.21) cm or 0.7% (6.0%) per year. Similarly, the overall change in volume was 79.9 (526.2) cm3, with an annual growth rate of 16.1 (91.0) cm3 or 2.8% (21.0%) per year.
When examining only lesions that grew 5% or more in mean linear dimension (39.3% of all hemangiomas) or volume (44.7% of all hemangiomas with volumetric measurement), growth rates are more meaningful because they are not counterbalanced by the fraction of lesions that regressed in size. Furthermore, this subset of lesions is thought to have exhibited an actual change in size rather than being a result of simple measurement variability. This subset of lesions exhibited overall growth in mean linear dimension by 0.7 (1.1) cm, with an annual growth rate of 0.19 (0.23) cm (Table 2). Similarly, this group of lesions had an overall volume change of 184.3 (777.8) cm3, with an annual growth rate of 37.8 (133.5) cm3 or 17.7% (22.8%) per year.
A strong correlation was demonstrated between initial mean linear dimension and hemangioma growth. The initial mean linear dimension strongly correlated with the overall change in mean linear dimension (R = 0.617; P < .001) and annual change in mean linear dimension (R = 0.594; P < .001) (Figure 2). Similarly, the initial mean linear dimension correlated strongly with overall growth in volume (R = 0.750; P < .001) and annual growth in volume (R = 0.833; P < .001).
Furthermore, even when examined as the percentage of growth, larger lesions showed significantly more growth with time. Of 112 hemangiomas with an initial mean linear dimension of less than 3 cm, only 46 (41.1%) exhibited 5% or more growth compared with 9 of 16 (56.3%) hemangiomas with an initial mean linear dimension of 5 cm or more. The overall and annual mean (SD) growth rates for lesions with an initial mean linear dimension of 3 cm or less were 0.04 (0.44) cm and 0.0065 (0.013) cm, respectively. For lesions with an initial mean linear dimension of 5 cm or less, the overall and annual growth rates were 1.4 (2.1) cm and 0.37 (0.40) cm. A similar pattern was seen when measured by volumes.
During the study period, hemangiomas appeared to grow in a uniform fashion as a function of time. There was no association between the annual growth rate and the number of years between the initial and final studies (R = 0.00843; P = .92) (Figure 3).
The traditional teaching has been that most hepatic hemangiomas remain constant in size over time.9- 11 However, in this study, we have shown that one-half of hepatic hemangiomas exhibit growth during long-term follow-up. Specifically, 50.9% of hemangiomas exhibited some growth in mean linear dimension, with 39.3% growing 5% or more in mean linear dimension and 44.7% growing 5% or more in volume.
A number of small series have reported some growth in hepatic hemangiomas over time. The reported fraction of hemangiomas that grew has ranged from 1% to 17% in these series.2,7,13,16,17 The reason that we found a higher fraction of growing lesions is likely multifactorial. First, in several previous studies, only symptomatic patients were reimaged, increasing the population for larger hemangiomas. Second, we included all patients with hemangiomas larger than 1 cm who had at least 1 year of follow-up, thereby capturing more points in time. Finally, most prior studies relied almost solely on ultrasound imaging, with which reproducible measurement is much more difficult and user dependent. Furthermore, in some series, larger lesions were excluded owing to difficulty in obtaining accurate measurements, thus eliminating the lesions in which we identified the highest growth rates.2,11,13
To our knowledge, this study is the first to use advanced cross-sectional imaging (computed tomography or magnetic resonance imaging) alone to describe the growth rate of hemangiomas during long-term follow-up. Use of advanced cross-sectional imaging allows not only for more precision in diagnosis but also for a more reliable means of assessment of growth, as the lesions can be measured in a defined plane and compared directly on follow-up studies. As a result, this study is the first to examine change in hemangioma volume, which is the most accurate measure of hemangioma size.
While an understanding of the fraction of hemangiomas that grow is important, knowing the normal rate of growth is necessary for clinical utility. Given that these lesions are frequently detected incidentally and then followed up inadvertently as serial images are obtained for another indication, such as cancer surveillance, it is essential to understand what change in hemangioma size is considered to be outside the normal range.14,18,19 Our study defines the predicted annual growth rate of all hemangiomas to be 0.34 mm. For lesions that have growth of 5% or more in mean linear dimension, the annual growth rate is 0.19 cm. Furthermore, we found that the initial mean linear dimension predicts the overall and annual growth rate. Hemangiomas less than 3 cm in mean linear dimension have an annual growth rate of only 0.0065 cm compared with 0.37 cm for hemangiomas 5 cm or more in initial size.
Although there are no convincing data to guide clinical treatment for patients with growing hemangiomas, it is clear that hemangioma growth currently leads surgeons to consider resection. This outcome is well demonstrated by a recent multicenter study of 1899 patients who received resection of hepatic hemangiomas. In this multicenter international study,12 the second most common indication for resection, accounting for 31% of all procedures, was hemangioma growth. Therefore, defining expected growth rates should help guide radiographic interpretation of serial imaging in patients with hepatic hemangiomas. Although this study cannot address the clinical significance of growth outside the range expected for a population of patients with hemangiomas, it provides boundaries that can be used to define outliers. Further long-term prospective evaluation of patients who present with growth beyond what is expected is needed to determine how this growth should affect clinical decision making. This information has the potential to decrease the number of operative resections performed because of hemangioma growth.
Although this is one of the largest studies to report the long-term radiographic follow-up of hepatic hemangiomas, it also has several limitations. First, the data were collected from an imaging database retrospectively, thereby limiting the amount of associated clinical information available. Second, although the follow-up period was long compared with other articles, it is still relatively short when compared with the life of the patient. Perhaps when followed up for many decades, both the fraction of hemangiomas that grow and the rate of growth would differ from those of this study. Finally, we examined only the first and last imaging study for each patient to allow for maximal change in size. It is possible that there were greater fluctuations in size on scans done in the interval that were missed with this strategy.
The natural history of hepatic hemangiomas includes growth up to 2 mm/y for nearly half of these lesions. A normal range for overall and annual growth rates should prove helpful in the clinical setting for deciding which lesions are outliers. Further research is required to determine the clinical significance of the growth of hemangiomas for patients with growth rates outside the range defined by this study.
Accepted for Publication: February 21, 2014.
Corresponding Author: Hani Y. Hasan, MD, Department of Surgery, Medical College of Wisconsin, 9220 W Wisconsin Ave, Milwaukee, WI 53226 (email@example.com).
Published Online: October 15, 2014. doi:10.1001/jamasurg.2014.477.
Author Contributions: Dr Hasan had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hasan, Winslow.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hasan, Gegios, Winslow.
Critical revision of the manuscript for important intellectual content: Hasan, Hinshaw, Borman, Leverson.
Statistical analysis: Hasan, Gegios, Leverson, Winslow.
Administrative, technical, or material support: Hasan, Winslow.
Study supervision: Hasan, Hinshaw, Winslow.
Conflict of Interest Disclosures: None reported.
Previous Presentation: The results of this study were presented at the Americas Hepato-Pancreato-Biliary Association annual meeting; February 23, 2013; Miami, Florida.