Magnetic Resonance Imaging as a Predictor of Outcome in Congenital Lymphatic Malformations of the Neck

Abstract

Objectives: 

To identify factors on magnetic resonance imaging (MRI) that predict the outcomes of children with congenital lymphatic malformations (LMs) of the neck.

Design: 

Retrospective cohort study.

Setting: 

The Royal Children’s Hospital, Melbourne, a quaternary pediatric referral hospital.

Patients: 

All children referred to the vascular anomalies service with LM identified at birth or in the first month of life were born between 1991 and 2015.

Outcome measurements: 

Pattern of involvement on MRI imaging, need for intervention for respiratory difficulty, documented episodes of infection, and aesthetic outcomes.

Results: 

MRI images were available for 27 of 30 patients with LM of the neck present at or soon after birth. Of these, 10 (37%) met the primary outcome of undergoing any intervention whose intention was stated to be for airway compromise, which occurred between 7 days and 4 years of age. The intervention was more likely when MRI showed parapharyngeal involvement (P < .001), although not all cases with parapharyngeal disease required respiratory intervention. Risk increases if other regions of the neck were also involved. Overall, 2 broad patterns of disease were identified, a “posterior pattern” which was typically macrocystic and unilateral, and an “anterior pattern” which was microcystic and bilateral and more likely to have airway compromise. Thirteen patients (48%) had documented episodes of infection.

Conclusion: 

Patterns of disease on MRI are identified which are associated with a greater risk of requiring intervention for airway obstruction. Children with congenital LM of the neck are at high risk of infective episodes.

Introduction

The neck is the most common site for lymphatic malformation (LM)^1–3 and in around one-third of cases, the lesion is clinically evident before or at the time of birth.^4,5 Congenital LMs tend to have more severe diseases than those which present later, with more complications, in particular respiratory obstruction.² Respiratory compromise is not always evident in the perinatal period, and may occur at any age due to an episode of swelling in the lesion such as from hemorrhage or infection. Other complications of congenital LM, including long-term aesthetic compromise may also be significant.

Predicting which children with cervical LM are at risk of developing respiratory problems, and therefore require close monitoring, is not always straightforward. Stratification of cases to identify those at most risk would be of use for guiding monitoring. Furthermore, as cases of LM are increasingly diagnosed in utero, there is a need to identify those who may require respiratory support or intervention at delivery and a better understanding of postnatal magnetic resonance imaging (MRI) is potentially informative.

This study reviews a cohort of children born with congenitally evident LM of the neck, which are here defined as those identified before birth, on delivery, or within the first month of life. The primary aim is to identify which features on MRI imaging are associated with subsequent surgical intervention intended to correct respiratory compromise.

Methods

Children with a diagnosis of LM of the neck were identified from a prospectively maintained database of the vascular anomalies service of The Royal Children’s Hospital, Melbourne with permission of the institute Human Research Ethics Committee (Reference: 35170A). The database covers patients treated from 1999 and was sampled in 2018 and reviewed again in 2020 so that all study patients completed at least 5 years of follow up. Congenital lesions were defined as those clinically evident before, at, or within 28 days of birth. Patients with a significant venous component to their lesion were excluded, as were patients for whom 5 years of follow-up data were not available.

Diagnosis of LM was confirmed by review of clinical details, imaging results, and histology if available. Basic demographic, clinical, and treatment data were recorded. Clinical charts were reviewed to obtain clinical information, in particular, whether airway compromise was clinically suspected and whether any intervention for airway compromise was performed. Any intervention, surgical, interventional radiological, or medical, including intubation for treatment of the LM within the first 5 years of life was recorded. The medical record was interrogated for the indication for treatment as recorded by the treating clinicians, identifying specifically those where the stated indication was respiratory compromise.

Episodes of infection were defined as: positive bacterial culture; local symptoms with fever and either raised C reactive protein or raised white cell count; or as determined by the treating team following a review of evidence (such as where initial treatment of an infective episode was performed elsewhere and relevant tests were not performed).

Magnetic resonance imaging

The earliest available, pretreatment MRI was reviewed for each patient. T2 fat-suppressed transverse, sagittal and coronal images were reviewed by the treating clinicians and by 2 radiologists. Extent of involvement was assessed by dividing the neck into 7 anatomical regions on each side: (1) tongue, including tongue base (T); (2) submental triangle (Sm∆); (3) anterior triangle superior to omohyoid muscle (A∆S); (4) anterior triangle inferior to omohyoid (A∆I); (5) parapharyngeal space, deep to carotid artery (PP); (6) posterior triangle (P∆); (7) cheek and/or parotid gland (Ch). Boundaries between regions were defined according to the anatomical convention. Each region was recorded as involved if diseased tissue could be identified within its anatomical boundaries. The assessment was performed independently by 2 radiologists, and by 1 surgeon. Assessments were compared, and then a final decision on involvement of each region in each case was reached by consensus.

Statistical analysis

To estimate whether respiratory intervention increased with number of anatomical regions involved, numbers of involved regions were categorized into groups (1–3, 4–7, 8–11, and 12–14) and a χ² test for Trend was performed. For all other categorical data, Fisher’s exact test was used. Reliability of MRI assessments was evaluated using the Kappa measure of agreement, with levels above 0.8 being considered strong. A P-value <.05 was considered significant. Statistical testing was performed using the “r” statistical package.

Results

Clinical data

In a database of 590 low-flow malformations, 97 patients were identified as having LMs in the neck (16%). Of these 30 (31%) met the inclusion criteria of lymphatic lesions presenting within 28 days of birth and 27 had MRI scans available for analysis. Four patients with congenital LM did not meet the criteria: 1 patient who moved interstate at age 2 and was therefore not available for review, and 3 other patients did not present for follow-up after 18 months.

Twenty-two of the 27 patients underwent 1 or more procedural interventions up to the age of 5 years: 8 had sclerotherapy alone, 8 had surgery alone, and 6 had both surgery and sclerotherapy. For those having sclerotherapy, the number of procedures ranged from 1 to 6 with a mean of 2.9. Of the 14 who underwent surgery, most had a single operation, 2 patients had 2 procedures and 1 had 4. The initial surgical procedure in most cases was a debulking of visible disease in the neck. Six surgical patients required perioperative blood transfusion. Major complications of surgery included Horner’s syndrome in 2 patients and documented weakness of the marginal mandibular branch of the facial nerve in 2 further patients.

Confirmed or suspected airway obstruction was the indication recorded for intervention in 10 patients (30%) which included surgical debulking, tracheostomy, and injection sclerotherapy. The age at which the first intervention was required ranged from 7 days to 4 years. Six patients underwent a treatment in the first 8 weeks of life where the recorded indication was respiratory obstruction, of which 5 underwent debulking surgery of the neck. One patient underwent tracheostomy at another hospital after being intubated from birth, without debulking of the LM which was performed some years later. This was 1 of 3 patients who underwent tracheostomy, 2, including that patient, performed emergently to secure a safe airway and 1 had an elective tracheostomy in anticipation of multiple sclerotherapy treatments. The 2 emergent tracheostomies remained in place for 18 years, 3 months and 11 years, 9 months. The elective tracheostomy was removed at 18 months of age. Of the 4 patients who required intervention where the indication was respiratory obstruction after 8 weeks of age, 3 underwent debulking surgery and 1 was treated primarily with sclerotherapy.

Thirteen patients in the study group (48%) had 1 or more documented episodes of infection. In 4 cases this occurred without prior intervention in the first 3 months of life. Three episodes of infection occurred in the postoperative period. The overall infection rate in the broader cohort of low-flow malformations from which this study group is derived, of which 35% are LM, has previously been reported as 4.5%.⁶

Pattern of disease on MRI

Pretreatment MRI images showed that 16 cases had unilateral disease and 11 were bilateral. There was a good correlation between the independent assessments of the 2 independent radiologists with Kappa values above 0.8 for most regions. The regions with the poorest correlation were the anterior triangle inferior to omohyoid (Kappa = 0.587), followed by parotid (Kappa = 0.715). After consensus agreement, the side of the neck with the most regions involved was found to be the left in 12 cases, the right in 10 cases, and in 5 there were an equal number of regions involved on each side.

To visualize the patterns of involvement, the side with the most involved regions was designated the primary side and cases were ranked according to the number of regions involved on the ipsilateral side from most to least. Patterns of involvement are presented graphically in Figure 1, along with whether the intervention was performed for respiratory compromise and whether infection occurred. Two broad patterns of disease can be seen. Sixteen cases (cases 3, 12–18, and 20–27) show what can be called a “posterior” pattern, with involvement of the posterior triangle, extension into the anterior triangle in most cases, but no contralateral disease (Figure 2B–D). Nine cases (cases 4 to 11 and case 19) meanwhile, which do not involve the posterior triangle, can be considered “anterior” pattern cases (Figure 2E and F). All but 2 of these are bilateral. Cases 1 and 2 demonstrate extensive, bilateral disease of both posterior and anterior triangles. Although not quantified, posterior pattern cases tended to be macrocystic, and anterior pattern cases were more microcystic.

Figure 1.: Patterns of involvement of the neck by region. The presence of any malformation within the anatomical region is indicated by red color. The side with the most anatomical regions involved has been designated the ipsilateral side and cases have been ranked by number of involved regions from most to least. Yellow color under “Airway” indicates that intervention was required for airway compromise and green under “Infection” indicates a documented episode of infection. Anatomical regions are as follows: T – tongue; SmΔ – Submental triangle; AΔS – Anterior triangle superior to omohyoid; AΔI – Anterior triangle inferior to omohyoid; PP – Parapharyngeal space, deep to carotid sheath; PΔ – Posterior triangle; Ch – Cheek/Parotid gland.

Figure 2.: Representative MRI images. A, Case 1: bilateral, posterior, and anterior involvement. B, Case 14: posterior pattern case without respiratory obstruction. C, Case 16: posterior pattern case who developed respiratory symptoms due to extension of disease posterior to carotid sheath. D, Case 15: posterior case imaged during an episode of infection. E, Case 4: Anterior pattern of disease, bilateral, mixed micro- and macrocystic. F, Case 11: Anterior pattern disease, involving parapharyngeal space but without respiratory obstruction. All images are T2 weighted, fat-suppressed, taken on a Siemens Aera 1.5T or a Siemens Prisma 1.5T machine. MRI indicates magnetic resonance imaging.

From Figure 1 it can be seen that involvement of the parapharygeal space and adjacent area is associated with respiratory obstruction. Ten of 15 children with parapharyngeal disease underwent intervention, while none of the 12 patients without parapharyngeal did so (P < .0005). All 10 children who developed respiratory compromise had involvement of the paraparapharyngeal area and both the superior and inferior anterior triangle on at least 1 side. Not all cases with parapharyngeal disease, however, had respiratory problems, including 2 with bilateral involvement. The risk of respiratory compromise requiring intervention increased with the number of anatomical regions involved (P < .005). The risk of infection was also closely related to extent of the disease.

Cases with a posterior pattern of disease were less likely to require intervention for respiratory compromise than those with anterior disease (P = .01). Only 3 of 16 posterior pattern cases required intervention, of which only one was in the first year of life. In contrast, 6 of 9 anterior pattern cases required intervention. This included all 5 anterior cases with bilateral tongue involvement, 4 of whom intervention occurred in the first 6 weeks of life and the fifth at 3 months. Surprisingly, 1 of the 2 cases with extensive posterior and anterior disease did not require intervention (Figure 2A), although this case did undergo a sleep study which was reported as showing “extremely mild” obstructive sleep apnea. The 3 cases of posterior pattern disease where respiratory intervention was required were specifically reviewed. Case 13 developed an acute infection in the lesion within days of birth leading to intubation for respiratory obstruction and subsequent surgery. Case 12 underwent surgery at the age of 14 months at parental request, and although respiratory obstruction was given as a reason for surgery, this was not well documented. Case 16 was notable for having a posterior triangle lesion with relatively little involvement of other parts of the neck, but MRI showed macrocystic disease with an extension behind the carotid sheath into the parapharyngeal space, displacing the carotid sheath anteriorly (Figure 2C). The child presented via the respiratory medicine unit at age 2 with symptoms of obstructive sleep apnea and underwent 3 sclerotherapy treatments from the age of 2 years 4 months with resolution of symptoms.

Discussion

LMs of the neck can cause significant long-term morbidity, often requiring multiple admissions to hospital through childhood.⁷ The one-third of LMs evident at birth are expected to be the most extensive lesions and therefore to have the most complications. This study confirms the high rate of complications in these patients and that the rate increases with more extensive disease. These children also have a much higher than expected incidence of infection.

Various staging systems have been proposed for LM of the neck. The most used is that of de Serres et al published in 1995 which graded LMs into 5 stages: unilateral disease above the hyoid, below the hyoid, or both, and bilateral disease either above or both above and below the hyoid (a potential sixth category of bilateral disease below the hyoid was not included because no cases fitted this pattern).⁸ The classification was clinical (MRI was not routinely available at this time) and no justification was offered beyond convention for separating above and below the hyoid bone. They reported that bilateral disease (stages IV and V) had the highest complication rate a finding which has since been confirmed.^9,10 Although the de Serres classification is the most widely used to report outcomes of neck LM, it suffers from the limitation that in an uncommon disease like LM, dividing cases into 5 (or 6) categories usually leaves too few cases in some categories for statistical analysis. For this reason, stages IV and V are sometimes combined, as are I to III, making the comparison effectively one between unilateral and bilateral disease.

In this study, rather than defining categories and asking how many cases fit into each, we have approached the subdivision of cases of congenital LM naively, dividing the neck into recognized anatomical segments and observing whether any natural pattern emerged. By plotting regional involvement in Figure 1, we observed that most cases fell naturally into 2 broad categories, which generally correspond with unilateral and bilateral disease. This supports the practice of consolidating the de Serres classification into just these 2 types, especially since even with good quality MRI imaging, dividing the anterior neck into superior and inferior segments may not be reliable.

In recent years, it has become clear that LMs are caused by somatic genetic mutations, mostly gain of function mutations in the gene PIK3CA.⁸ Because such a mutation must necessarily arise in a single cell, the anatomical pattern of disease should depend on the anatomical location and the stage in development at which the mutation arose. We have observed (Figure 1) that congenital unilateral cases nearly always involve the posterior triangle, where bilateral cases rarely do, even though more segments are typically involved. We suggest that this can be explained if unilateral cases most often follow from mutations arising in the posterior triangle, with disease spreading forward, and bilateral disease from mutations arising in the anterior triangle. Cases with combined bilateral anterior and posterior involvement are likely to be those where the mutation has arisen at the earliest stages of development. Although LMs are well known to cross tissue planes, the fasciae of the neck, such as the prevertebral fascia, do seem to offer some barrier to spread. Some posterior lesions appear to abut the carotid sheath without spreading into the anterior triangle. They can, however, sometimes progress in a plane posterior to the sheath into the retropharyngeal space (Figure 2B and C) where despite being separated from the pharynx by the alar fascia¹¹ they may cause respiratory difficulty The preponderance of macrocystic disease in the posterior lesions may be due to lesser resistance to cyst enlargement from the loose connective tissue. In contrast to the multiple fascial layers separating compartments of the anterior triangle, there is no defined fibrous fascial layer in the posterior triangle.^12,13

The proportion of patients requiring intervention for respiratory difficulty in this study is much higher in anterior/bilateral disease posterior/unilateral cases, consistent with previous reports of higher respiratory complications in bilateral disease.^8–10 The mechanism of respiratory obstruction, however, in children with LM of the neck is not completely understood. The presence of disease in the parapharyngeal space is strongly associated with need for respiratory intervention, however, not all patients with parapharyngeal disease, even those with extensive disease necessarily developed respiratory difficulty. Also, the presence of disease distant from the pharynx increased the likelihood of respiratory compromise suggesting a mechanism beyond direct mass effect. A significant limitation of this study is that in a retrospective study with at least 5 years of follow up, most patients did not receive modalities of airway assessment that are now considered routine, particularly polysomnography and flexible endoscopy. This prevents comment on the presence of luminal disease, which has itself been proposed as a method of classification of neck LM.¹⁴ The infiltrative nature of anterior disease suggests that luminal involvement is likely to be more common in such cases, but this will require further study. Also, nearly all patients in this study were treated before the introduction of sirolimus, so whether this or the newer agents such as PI3Kinase inhibitors will have an impact on which children require respiratory intervention remains unknown.

The high rate of infection in children with congenital LM of the neck seen in this study was unexpected. The reasons for this are not clear, but proximity to the mouth and pharynx may provide a portal of entry. The segment of the neck most often involved in cases who developed infection was the superior anterior triangle, suggesting that the floor of the mouth may be a site where bacteria gain entry. The first 3 months of life, when the immune system is not well developed, appears to be a time of increased risk for infection.

Conclusion

For children born with LM, those with involvement of the parapharyngeal area on MR are most likely to require intervention for respiratory difficulty, and their risk increases with the extent of regional involvement. Malformations which predominantly involve the posterior triangle are most often unilateral and less likely to develop respiratory problems.