OBJECTIVES. Large facial infantile hemangiomas have higher rates of complications than small localized hemangiomas, more often require treatment, and can be associated with neurological, ophthalmologic, and cardiac anomalies (PHACE syndrome). The anatomic patterns of these hemangiomas are often referred to as “segmental” despite a lack of precise anatomic definitions. Our study aims to define “segmental” hemangiomas based on clinically observed patterns. Our secondary goal is to relate the observed patterns to currently accepted developmental patterns to gain insight into hemangioma pathogenesis and craniofacial development.
METHODS. Photographic data were extracted from a large cohort of patients with infantile hemangiomas. We mapped 294 hemangiomas and recorded common morphologic patterns. Anatomic descriptions of the most common patterns were described and compared with accepted concepts of craniofacial development.
RESULTS. Four primary segments were identified (Seg1–Seg4). Seg2 and Seg3 correspond with the previously recognized maxillary and mandibular prominences. Seg1 and Seg4 differ from standard human embryology texts. The frontotemporal segment, Seg1, encompasses the lateral forehead, anterior temporal scalp, and lateral frontal scalp. The segment Seg4, encompassing the medial frontal scalp, nasal bridge, nasal tip, ala, and philtrum, is substantially narrower on the forehead than the previously described frontonasal prominence.
CONCLUSIONS. The patterns provide new clues regarding facial development. The observed patterns resemble previously described facial developmental units on the lower face but are distinctly different on the upper face. The patterns suggest that neural crest derivatives may play a role in the development of facial hemangiomas. Finally, these patterns (Seg1–Seg4) help standardize the nomenclature of facial segmental hemangiomas to analyze more effectively hemangioma risks and behavior.
infantile hemangiomas are extremely common. They have unique growth characteristics, typically being absent or barely noticeable at birth, followed by a period of rapid growth and then subsequent involution. The majority of affected infants have an uncomplicated clinical course, but a significant subset have function-threatening complications or permanent and disfiguring scars that require medical or surgical interventions.
Three recent observations regarding infantile hemangiomas spurred our interest in studying their anatomic patterns. First, there is a clear association between hemangiomas, particularly large facial hemangiomas, and certain structural birth defects. This association, known as PHACE syndrome (Online Mendelian Inheritance in Man No. 606519), includes a group of abnormalities primarily affecting the central nervous system, cerebrovasculature, aorta, heart, eye, and sternum. Many of these birth defects originate between 6 and 8 weeks of gestation, which is surprising, because hemangiomas are absent at birth. Second, recent reports have divided hemangiomas into “localized” (arising from a central focus) and “segmental” (covering an anatomic territory) subtypes, noting a higher risk of complications and associated structural anomalies in segmental hemangiomas when compared with localized hemangiomas.1 Third, Waner et al2 demonstrated that facial hemangiomas are not random in their distribution and suggested that segmental facial hemangiomas correspond with known developmental units such as the mandibular and maxillary prominences.
We hypothesized that a systematic analysis of photographic data of patients with facial hemangiomas could enable us to determine whether segmental hemangiomas correspond to known developmental units,3 follow other purported subdivisions associated with embryogenesis such as dermatomes or the lines of Blaschko,4 or represent the functioning of additional unspecified mechanisms that pattern the craniofacial complex. Just as the recognition that a facial port-wine stain in the “V1” dermatome distribution confers a recognized risk of Sturge-Weber syndrome,5 clarification of segmental hemangioma patterns might aid in identifying those infants at greatest risk for certain complications and/or structural anomalies. Moreover, studying patterns of facial hemangiomas may not only provide insight into the underlying pathogenesis but also contribute more generally to our understanding of craniofacial development.
We analyzed photographs collected as part of a large prospective cohort study of infantile hemangiomas. Patients were eligible if they had a facial hemangioma, were aged 12 years or younger, and were examined between September 2002 and October 2003. Lesions were classified by morphologic subtype as segmental, localized, indeterminate, or multifocal using criteria in a manual created for this study. “Segmental hemangiomas” were defined as those hemangiomas or clusters of hemangiomas with a configuration corresponding to a recognizable and/or significant portion of a developmental segment or involving a broad anatomic territory of skin. “Localized hemangiomas” were defined as those hemangiomas that seemed to grow from a single focal point or were localized to an area without any apparent linear or metameric configuration. “Indeterminate hemangiomas” were those that were not readily classified as either localized or segmental, and “multifocal hemangiomas” referred to those found on patients with ≥10 cutaneous lesions. A total of 294 hemangiomas were observed as being either segmental (115) or indeterminate (179). Seventy two of the segmental hemangiomas and 93 of the indeterminate hemangiomas were located on the face and were used for this analysis. Each facial hemangioma was mapped onto a facial template.
A total of 294 hemangiomas (115 segmental and 179 indeterminate) were reviewed. Of these, 34 segmental and 68 indeterminate hemangiomas were excluded, because they were nonfacial. Nine segmental and 18 indeterminate hemangiomas were excluded because of poor image quality that precluded accurate assessment of anatomic boundaries. Therefore, a total of 165 facial hemangiomas (72 segmental and 93 indeterminate) were analyzed.
Initial image analysis revealed that the hemangiomas could be segregated into 8 categories based on location (Fig 1): (1) temporal ocular; (2) medial maxillary; (3) lateral mandiblular including the entire lower lip; (4) central mandibular; (5) nasal tip; (6) frontonasal; (7) lateral maxillary; and (8) retro-orbital/periorbital (Fig 1A). However, larger hemangiomas often encompassed >1 anatomic category, implying that certain contiguous anatomic categories could be combined. This led to the identification of 4 primary segments: Seg1–Seg4 (Fig 1B). Twenty patients had hemangiomas that encompassed >1 segment.
Seg1: 14 Patients
The frontotemporal segment, Seg1, encompassed the lateral forehead, anterior temporal scalp, and lateral frontal scalp. Frequently, all or part of the upper eyelid was involved as well. The glabella and central forehead were consistently spared (Fig 2). This segment does not correspond to any classically described embryonic “metamere” or facial prominence.
Seg2: 39 Patients
Hemangiomas in the Seg2 territory respected the nasomesial sulcus medially and extended to the lateral cheek but did not involve the philtrum or preauricular area (Fig 3). Involvement of upper lip lesions was always unilateral and never crossed the midline. This segment approximates the maxillary prominence as described in medical embryology texts.3
Seg3: 44 Patients
The mandibular segment, Seg3, corresponded to the preauricular region, mandible, chin, and lower cutaneous and vermillion of the lower lip (Fig 4). Bilateral Seg3 involvement was noted in several patients. This segment approximates the mandibular prominence as described in medical embryology texts.3
Seg4: 32 Patients
The medial frontal scalp, nasal bridge, nasal tip, ala, and philtrum comprised the frontonasal segment, Seg4 (Fig 5). Seg4 corresponds with embryological descriptions of the frontonasal prominence, with 1 notable exception: in contrast to the depiction of this prominence extending superiorly and laterally from the glabella to form a rounded territory encompassing a majority of the forehead, Seg4 formed a V-shaped inverted triangle on the central forehead, extending downward onto the nasal bridge, with sharp linear borders diagonally traversing the lateral forehead.3
Periorbital Hemangiomas: 16 Patients
Segmental hemangiomas involving the periorbital area did not consistently segregate into Seg1, Seg2, Seg3, or Seg4 patterns. Lesions adjacent to the medial canthus were often present alone or in combination with Seg1 or Seg2 lesions, suggesting that the periorbital area may be an area of overlap or may be contiguous with either Seg1 or Seg2.
Although they respected segment boundaries, hemangiomas did not always encompass an entire segment. For example, some patients had a hemangioma located on the upper eyelid with only a small extension toward the lateral forehead, suggesting an incomplete segmental pattern.
The classification of infantile hemangiomas into segmental and localized forms arose from the observation that hemangiomas are clinically diverse. Some are small and localized, whereas others display a linear and/or geographic distribution, covering larger anatomic regions, which we refer to as “segmental.” These segments do not correspond with facial dermatomes or lines of Blaschko, which are mosaic expressions of neuroectodermal migration.4 Waner et al2 proposed that the patterns corresponded with embryological facial prominences.
Four segmental patterns were observed in this study: Seg1 (frontotemporal segment), Seg2 (maxillary segment), Seg3 (mandibular segment), and Seg4 (frontonasal segment). Some patterns have marked similarities to facial prominences described in embryology textbooks but with critical differences, particularly regarding lesions of the upper half of the face. Diagrams of the facial prominences usually depict the frontonasal prominence as extending to the lateral forehead.3 In contrast, our results showed that the frontonasal prominence was limited to an area on the medial forehead and scalp, whereas another developmental unit, termed Seg4, included the lateral forehead and temple.
Hemangiomas in the periorbital area occurred in isolation or contiguously with hemangiomas of the Seg1 or Seg2 distribution. Large plaques in the Seg1 distribution often extended onto the lateral half of the upper eyelid and occasionally extend across the entire width of the upper eyelid. Some hemangiomas in the Seg2 distribution extended onto the lower eyelid. The inconsistent association of periorbital hemangiomas with either Seg1 or Seg2 segments suggests an area of overlap between Seg1 and Seg2. This may reflect the independent development of the orbit, with its ventral emergence in the face. Conceptually, a variable region for segmental hemangiomas may be analogous to dermatomal variation seen in the distribution of V1 and V2 port-wine stains.5
The patterns of hemangiomas on the upper and lower lip reflect accepted models of central facial development. The frontonasal prominence forms the philtrum and separates the maxillary processes. The mandibular prominences fuse at the midline. Hemangiomas involving the upper lip were either limited to the philtrum or unilateral, respecting the lateral border of the philtrum. They did not extend from one commisure to the other as was observed with many lower lip hemangiomas. In contrast, segmental hemangiomas of the upper face did not resemble the shape of a single frontonasal metamere but suggested that there is 1 central (Seg4) and 2 lateral developmental segments (Seg1).
Some hemangiomas involved >1 segment, whereas others neither encompassed an entire segment nor arose from a single focus. Such lesions have been referred to as “indeterminate.”1 The borders of indeterminate hemangiomas respected the developmental segment boundaries we have observed, supporting the concept of “incomplete segmental” hemangiomas. Our findings address a clinically significant subset of all infantile hemangiomas. Approximately 50% of hemangiomas occur on the head and neck, and up to 20% of these are segmental. Segmental hemangiomas on the trunk and extremities may also resemble developmental units, but these were not addressed in this study.
Although specific cellular and molecular mechanisms underlying infantile facial hemangiomas remain obscure, the observed segmental patterns suggest that in each case some common developmental process has gone awry. In particular, the nonrandom distribution of these aberrant tissues indicates that on the cellular level they are responsive to the same morphogenetic mechanisms that establish boundaries among individual facial primordia. There are 5 distinct embryonic primordia in the developing face. The paired mandibular primordia give rise to the lower jaw, lips, and chin. The paired maxillary primordia produce the upper jaw and lip (excluding the philtrum), the secondary palate, and the cheeks. The frontonasal primordium generates the middle and upper face from the forehead to the philtrum of the lip, as well as the primary palate. The frontonasal primordium can be additionally subdivided into lateral and medial domains that, respectively, give rise to the sides and bridge of the nose. Each facial primordium is populated by mesenchymal cells derived predominantly from the cranial neural crest and is enclosed by an epithelium that originates from nonneural ectoderm.
Although these primordia share similarities in organization and constitution, the molecular mechanisms regulating their patterned outgrowth seem to be distinct.6–8 This may be attributable, in part, to the origins of neural crest cells that populate each primordium, as well as to regional differences in overlying ectoderm.9,10 For example, neural crest cells that occupy the frontonasal primordium come from a more rostral level of the neural tube than those that reside in the maxillary or mandibular primordia. As a result, each subpopulation of neural crest may have a unique molecular identity, hold different potential, and be regionally compartmentalized. Ultimately, this would affect the patterned growth and subsequent integration of the facial primordia and all of their constituents including skeletal, muscular, and vascular components. For some yet-to-be-determined reason, segmental hemangiomas seem to perpetuate embryonic boundaries that, in all other regards, have seemingly disappeared.
In addition to initial embryonic boundaries that exist among the various facial primordia, another important interface exists within the craniofacial complex that likely has a mechanistic bearing on delimitations of segmental hemangiomas. Experiments using tissue transplants between quail and chick embryos have revealed a boundary in the head that stretches from the anterior mesencephalic-prosencephalic junction to the laryngotracheal diverticulum and separates dorsal mesenchyme of mesodermal origin and ventral mesenchyme derived from the neural crest.11 All of the skeletal and connective craniofacial tissues are generated by these mesenchymal populations, and they maintain their respective mesodermal or neural crest histories.
Extrapolating these avian fate map data to humans, we hypothesize that the juxtaposition of neural crest and mesodermally derived bones along the lateral margins of the skull (ie, lateral forehead and temples) can explain the discrete domains that we observe for hemangiomas in the Seg1 developmental unit. Such a conclusion is substantiated by transgenic experiments in mice, which have provided equivalent fate maps using indelibly labeled neural crest cells.12 Analysis of these mice confirms that the mammalian interface between neural crest and mesoderm runs through the temporal region, which might explain why segmental facial hemangiomas respect this boundary given their association with neural crest mesenchyme.
The observation that infantile hemangiomas exhibit growth patterns that respect embryonic segments is intriguing, because angioblasts that form blood vessels do not obey segment-specific boundaries.13,14 Rather, angioblasts arising within cephalic mesoderm migrate in all directions and contribute to blood vessels throughout the face. Thus, restriction of segmental hemangiomas to the derivatives of embryonic facial primordia does not seem to be the result of segmental identity imposed a priori on the endothelial cells. Segmental restriction more likely results from environmental influences that govern formation of the microvasculature of the skin. Proliferation of endothelial cells and growth of hemangiomas may occur in response to, or be constrained by, signals derived from cell types that exhibit segmental or regional identity, such as cranial neural crest cells. In this scenario, segmental restriction would be a secondary consequence of interactions between endothelial and neural crest cells.
A direct relationship between endothelial cells and neural crest cells exists. Pericytes, the connective tissue support cells of blood vessels, are derived from cranial neural crest cells,15 and interactions between pericytes and endothelial cells influence maintenance, maturation, and remodeling of blood vessels.16 Hence, the idea that formation of a hemangioma results from a primary defect in a segmentally or regionally restricted tissue, such as the neural crest, deserves additional exploration. In particular, a better understanding of cellular and molecular mechanisms that produce embryonic boundaries among facial primordia may explain the reasons for the observed correlation between segmental patterns of infantile facial hemangiomas and various facial regions.
Identifying hemangioma patterns should help standardize nomenclature for future studies that correlate risks and behaviors to specific sites and patterns. Preliminary evidence suggests that hemangiomas in the Seg1 distribution have a higher risk for associated brain anomalies of PHACE syndrome, whereas those involving Seg3 have a greater risk of cardiac defects.17 Future studies are needed to more effectively compare clinical outcomes in these groups of patients. The novel observation that hemangioma patterns along the dorsolateral margins of the face correspond precisely to boundaries present between mesenchyme derived from the cranial neural crest and that of mesodermal origin also raises fundamental questions about the origins of hemangiomas, which have typically been assumed to be because of a defect in the endothelial cells from which neoplastic blood vessels arise. These vessels have recently been recognized to have a unique vascular phenotype similar to placental microvasculature.18 The patterns observed suggest that other tissue components could be the real culprits and permit aberrant growth of these blood vessels after birth. We believe that a better understanding of the causes of segmental infantile hemangiomas can be achieved by focusing on molecular and cellular processes that regulate facial development. Identifying segmental patterns may also reveal novel mechanisms underlying normal and abnormal craniofacial patterning.
This work was supported by the Dermatology Foundation, which supported Dr Haggstrom's pediatric dermatology fellowship and granted her a clinical investigator award to fund this and other studies of infantile hemangiomas. The American Skin Association generously provided funding for data analysis.
We gratefully acknowledge the hard work and assistance of the Hemangioma Investigator Group (Beth A. Drolet, Eulalia Baselga, Sarah Chamlin, Nancy B. Esterly, Ilona J. Frieden, Maria Garzon, Anita N. Haggstrom, Kimberly A. Horii, Anne Lucky, Anthony J. Mancini, Denise Metry, Brandon Newell, and Amy Nopper), whose members were responsible for patient recruitment, data collection, and invaluable guidance. We also thank Sarajo Frieden for illustrating Fig 1.
- Accepted July 13, 2005.
- Address correspondence to Anita N. Haggstrom, MD, 5007 Benton Ave, Bethesda, MD 20814. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Larsen, WJ. Human Embryology. New York, NY: Churchill Livingstone; 1993:321;360–371
- ↵Enjolras O, Riche MC, Merland JJ. Facial port-wine stains and Sturge-Weber syndrome. Pediatrics.1985; 76: 48–51
- ↵Wedden SE, Ralphs JR, Tickle C. Pattern formation in the facial primordia. Development.1988; 103(suppl): 31–40
- ↵Couly G, Le Douarin NM. Head morphogenesis in embryonic avian chimeras: evidence for a segmental pattern in the ectoderm corresponding to the neuromeres. Development.1990; 108: 543–58
- ↵Hu D, Marcucio R, Helms JA. A zone of frontonasal ectoderm regulates patterning and growth in the face. Development.2003; 130: 1749–1758
- ↵Noden DM. Origins and patterning of avian outflow tract endocardium. Development.1991; 111: 867–876
- Copyright © 2006 by the American Academy of Pediatrics