Safe marginal resection of atypical neurofibromas in neurofibromatosis type 1

Abstract

OBJECTIVE

Patients with neurofibromatosis type 1 (NF1) are predisposed to visceral neurofibromas, some of which can progress to premalignant atypical neurofibromas (ANFs) and malignant peripheral nerve sheath tumors (MPNSTs). Though subtotal resection of ANF may prevent malignant transformation and thus deaths with no neural complications, local recurrences require reoperation. The aim of this study was to assess the surgical morbidity associated with marginal resection of targeted ANF nodules identified via preoperative serial volumetric MRI and ¹⁸F-FDG-PET imaging.

METHODS

The authors analyzed clinical outcomes of 16 NF resections of 21 tumors in 11 NF1 patients treated at the NIH Clinical Center between 2008 and 2018. Preoperative volumetric growth rates and ¹⁸F-FDG-PET SUV_Max (maximum standardized uptake value within the tumor) of the target lesions and any electromyographic or nerve conduction velocity abnormalities of the parent nerves were measured and assessed in tandem with postoperative complications, histopathological classification of the resected tumors, and surgical margins through Dunnett’s multiple comparisons test and t-test. The surgical approach for safe marginal resection of ANF was also described.

RESULTS

Eleven consecutive NF1 patients (4 male, 7 female; median age 18.5 years) underwent 16 surgical procedures for marginal resections of 21 tumors. Preoperatively, 13 of the 14 (93%) sets of serial MRI studies and 10 of the 11 (91%) ¹⁸F-FDG-PET scans showed rapid growth (≥ 20% increase in volume per year) and avidity (SUV_Max ≥ 3.5) of the identified tumor, respectively (median tumor size 48.7 cm³; median growth rate 92% per year; median SUV_Max 6.45). Most surgeries (n = 14, 88%) resulted in no persistent postoperative parent nerve–related complications, and to date, none of the resected tumors have recurred. The median length of postoperative follow-up has been 2.45 years (range 0.00–10.39 years). Histopathological analysis confirmed significantly greater SUV_Max among the ANFs (6.51 ± 0.83, p = 0.0042) and low-grade MPNSTs (13.8, p = 0.0001) than in benign neurofibromas (1.9).

CONCLUSIONS

This report evaluates the utility of serial imaging (MRI and ¹⁸F-FDG-PET SUV_Max) to successfully detect ANF and demonstrates that safe, fascicle-sparing gross-total, extracapsular resection of ANF is possible with the use of intraoperative nerve stimulation and microdissection of nerve fascicles.

Neurofibromatosis type 1 (NF1) is a tumor predisposition syndrome associated with benign nodular and plexiform neurofibromas. Inactivating autosomal dominant mutation of the NF1 gene located at position q11.2 of chromosome 17 results in defective encoding of the neurofibromin protein, a negative regulator of the Ras/Raf/MAPK pathway; unregulated activation of this pathway, likely in progenitor Schwann cells, results in the development of neurofibromas (Fig. 1).^24,29,43 In individuals with NF1, malignant peripheral nerve sheath tumors (MPNSTs) typically arise from preexisting plexiform or nodular neurofibromas.^14,15,27,42 Over their lifetimes, 8% to 15.8% of NF1 individuals develop MPNST with dismal 5-year survival rates (35%–50%).^13,40

FIG. 1.

Coronal and axial noncontrast STIR MR images of representative neurofibromas included in this study. A and B: Isolated nodular neurofibroma of left neck. (Arrowheads indicate tumor.) Unknown parent nerve, likely arising in close proximity to the vagus nerve in the neck. C and D: Nodular neurofibroma in the left arm arising from the median nerve. E and F: Plexiform neurofibroma of right thigh with a growing nodule within the region highlighted by white arrowheads.

Unlike schwannomas, which grow eccentrically to the parent nerve, neurofibromas are thought to grow concentrically and intimately involve the nerve fascicles. Therefore, it has been traditionally thought that resection of neurofibromas is often accompanied by deficits of the involved nerve function.^7,11 Atypical neurofibromas (ANFs) are considered premalignant lesions¹ to MPNSTs and may arise from isolated nodular neurofibromas or individual nodules within plexiform neurofibromas. Once an ANF has de-differentiated into a high-grade MPNST (hgMPNST), resection with wide negative margins is the only potentially curative treatment.^12,14,37

Subtotal resection of ANFs with positive margins prior to de-differentiation into MPNSTs has been shown to prevent nerve damage and may prevent malignant transformation; however, local recurrences of some of the tumors in these cases have required reoperation.² Pioneering peripheral nerve surgeon Dr. David Kline has previously observed that it is possible to perform total resection of symptomatic tumors in NF1 patients without serious motor nerve deficits of the major nerves.^22,23 A recent consensus conference on NF1-related ANF/MPNST management has recommended that resection of ANF, if feasible without major morbidity, should be considered to prevent malignant transformation and that in these cases, in contrast to cases of hgMPNST, wide negative margins should not be attempted.³⁵

Imaging surveillance via serial volumetric MRI and ¹⁸F-FDG-PET has allowed for early detection of characteristics suggestive of ANF or MPNST, but reliable detection of ANF has been traditionally difficult, and the use of imaging surveillance to define timing and strategy for surgical intervention has not been evaluated.^{3,5,17,18,21,25,30,31} Furthermore, the surgical morbidity associated with marginal resection utilizing the capsular planes of ANF in NF1 patients has not yet been elucidated. In this paper, we aim to achieve both.

We first assessed the growth rates and avidity of neurofibromas via serial volumetric MRI and ¹⁸F-FDG-PET, respectively, to detect potentially atypical transformation.^{17,18, 21, 28, 33,41} Based on these imaging studies, we aimed for nerve fascicle–sparing gross-total resections of the targeted neurofibroma nodules using intraoperative nerve stimulation in patients most likely to have ANF; the lesion type was confirmed postoperatively via histopathological analysis and inking. Next we assessed patients in the postsurgical period for any transient or persistent complications.^20,36 Then we retrospectively compared the histological stratifications of the resected tumors with preoperative imaging patterns to assess the utility of preoperative MRI and ¹⁸F-FDG-PET imaging in predicting ANF. Last, we have continued long-term follow-up of the patients by the same imaging modalities to assess tumor recurrence. We postulate that ANF nodules, like schwannomas, may grow eccentrically, allowing for their nerve fascicle–sparing marginal resection.

Methods

Patient Cohort

All patients in this study were previously enrolled at the National Cancer Institute (NCI) in the prospective NF1 natural history study (clinicaltrials.gov identifier NCT00924196; last amendment approval on December 3, 2018), which was approved by the NCI institutional review board and for which patients or their guardians provided informed consent to participate. The only additional inclusion criterion of our study was the necessity for marginal resection. We retrospectively analyzed the clinical outcomes in 11 consecutive NF1 patients following 16 neurofibroma resections of 21 tumors at the National Institutes of Health (NIH) Clinical Center between 2008 and 2018. Three of these patients were previously included in a descriptive study of clinical features of ANF (patients 1, 2, and 4), and the tumors of 2 of these 3 patients (patients 1 and 2) were also examined in a genetic study of ANF development.^18,34 We evaluated patient characteristics including age at time of surgery, sex, neurofibroma locations, reasons for patient referral, and presenting signs and symptoms (tumor firmness/tenderness and Tinel’s sign). Reasons for referral for neurosurgery included increased MRI volumetric growth rates and ¹⁸F-FDG-PET uptake of the tumors, failure of biological therapy, and pain.^9,44 All patients who demonstrated significant clinical indicators, such as pain, increased growth rate on MRI, and/or increased ¹⁸F-FDG-PET avidity, were referred for surgery and underwent resection of their neurofibromas in this study (Supplementary Fig. 1).

Imaging Parameters and Analysis

Using serial volumetric MRI and PET imaging, we calculated, respectively, the yearly growth rates and ¹⁸F-FDG uptakes of the targeted neurofibroma nodules. Tumor size was assessed by semi-automated volume segmentation using MEDx software.³⁹ Mean yearly neurofibroma growth rates were derived from the slopes of linear regressions between tumor size and duration of the patients’ preoperative screenings (Fig. 2). Following a previously set definition of tumor progression, we defined increased neurofibroma growth rate as a ≥ 20% increase in tumor volume per year.¹⁰ PET imaging was performed to assess metabolic avidity in neurofibromas with increased growth rates on MRI. The duration of patient preoperative screening, duration of postoperative follow-up, SUV_Max (maximum standardized uptake value within the tumor) of the targeted neurofibroma nodules, and recurrence to date of the targeted tumors since resection were measured. As no definitive SUV_Max cut-off exists to differentiate between benign neurofibroma and MPNST, we adopted a preliminary SUV_Max cut-off of 3.5, as used by other studies, to ascribe PET avidity.^8,38,41

FIG. 2.

Preoperative MRI scans and intraoperative photographs obtained in a case (case 4B) of a right brachial plexus neurofibroma showing growth suggestive of atypical transformation. A: Sagittal STIR image from initial MRI scan demonstrating a plexiform neurofibroma with characteristic small nodularity of the right brachial plexus. B: Follow-up sagittal STIR image obtained 2.3 years later revealing growth of 1 dominant neurofibroma nodule within the plexiform mass. C–H: Intraoperative photographs. The patient is supine, with the head toward the left side of the panel, and the right shoulder toward the bottom right corner. An initial skin incision and dissection through the platysma are performed (C). The parent nerve is identified and isolated, and its junction with the tumor (asterisk) is identified (D). Nerve fascicles traversing the tumor surface are identified visually (E) and mapped with bipolar stimulation (F). The nerve fascicles are incised along their main axis and split to reveal tumor capsule (asterisk) (G). The neurofibroma (asterisk) is circumferentially dissected from the parent nerve while sparing parent nerve fascicles (black arrows) (H).

Electromyography and Nerve Conduction Studies

We performed preoperative electromyographic (EMG) and nerve conduction velocity (NCV) studies to assess the extent of potential neurogenic changes in the parent nerves of neurofibromas, as previously reported.¹⁶ Sensory and motor nerve amplitude and conduction velocity as well as motor nerve distal and F-wave latencies were recorded and evaluated. We compared patient values with the NIH laboratory normal values and ranked the patients’ values accordingly (Supplementary Tables 1 and 2). An abnormal sign recorded from any branch of the parent nerve was reported as an abnormal nerve finding; studies with at least 1 abnormal motor or sensory finding were recognized as abnormal.

Marginal Resection

The senior author (P.C.) resected all tumors with intraoperative monitoring using a standardized resection protocol (Fig. 2; Video 1). The parent nerve was identified and isolated proximally and distally when possible. The tumor surface was mapped with bipolar stimulation (low-stimulation setting at 0.5 mA) and intraoperative EMG response monitoring. Under an operating microscope, the regions with minimal EMG stimulation or absent nerve fascicles were identified and sharply incised along the long axis of nerve fascicles to initiate capsular dissection. Circumferential capsular dissection with careful, periodic neuromonitoring was performed to remove the tumors en bloc with intact capsules. We then measured, inked, and sectioned the resected tumors and stained the sections with H & E.

Histopathological and Outcome Stratifications

Using the guidelines from the NIH consensus overview, we stratified the tumors as follows: 1) neurofibroma, 2) neurofibroma with nuclear atypia alone (atypical neurofibroma, ANF), 3) cellular neurofibroma, 4) atypical neurofibromatous neoplasm of uncertain biological potential (ANNUBP), or 5) low- or high-grade MPNST (lgMPNST or hgMPNST) (Fig. 3).³² We also classified the neurofibromas by their margins of resection and by their patterns of growth, either nodular or plexiform. We histologically ascribed marginal resections to sections with full inking around the perimeter of the tumor with visualization of the tumor capsule. If multiple tumors were isolated during a single resection, we imputed the histopathological classification of the most atypical tumor to the resection. Last, we evaluated the postoperative complications and assessed whether they were transient or persistent and whether they arose from injury to the neurofibroma parent nerves.

FIG. 3.

Representative histopathological images of H & E–stained sections of tumors included in the current study. A: Benign neurofibroma with a multinodular growth pattern within a plexiform neurofibroma. B: Atypical neurofibroma demonstrating nuclear atypia. C: Atypical neurofibromatous tumor with uncertain biological potential contains a mitotic figure (red oval) in addition to nuclear atypia. D: Low-grade MPNST showing nuclear atypia, increased cellularity, and mitotic activity (red ovals). E and F: Interpretation of inking. Smooth inked margins suggested a marginal resection (E), while a ragged margin with ink particles suggested a positive margin (F). Original magnification (E and F) ×200.

To assess for postoperative tumor recurrence, our clinical pathway entails yearly physical examinations and MRI for all patients; for the patients also in the clinical trial for the biological therapy selumetinib, we perform physical examinations and MRI every 4 months for the first 2 years and every 6 months thereafter. If an MRI scan indicates rapid growth (≥ 20% per year), then ¹⁸F-FDG-PET imaging is also performed to test for tumor avidity. If the ¹⁸F-FDG-PET imaging also suggests transformation into ANF or MPNST, we refer the patient for neurosurgical consultation for resection. We have not set termination dates for patient follow-up. If a patient develops pain before the yearly follow-up appointments at the NIH Clinical Center, the patient is referred to his or her NF1 home care team.

Statistical Analysis

We analyzed the differences in yearly growth rates and ¹⁸F-FDG-PET uptake between the benign neurofibroma stratum and the ANF, ANNUBP, and MPNST strata using Dunnett’s multiple comparisons test. Further, we used t-tests to compare the differences in yearly growth rates and ¹⁸F-FDG-PET uptake between individual histological strata.

Results

Patient Characteristics

We included 11 consecutive patients (7 female and 4 male) and 16 surgeries in this study (Table 1); the patients’ median age at the time of surgery was 18.5 years. In total, 21 tumors were resected; multiple neurofibroma nodules were resected in 3 surgeries. Five patients each underwent surgeries in 2 temporally distinct sessions. Tumor locations included the neck/brachial plexus (n = 7), chest/back (n = 2), abdomen/pelvis (n = 3), and upper (n = 1) and lower (n = 3) extremities. The tumor nodules were located along the parent nerve or within a plexiform neurofibroma. Three patients (27%) had histories (n = 1) or family histories (n = 2) of hgMPNST, and 4 patients (36%) experienced pain at rest in the regions of their neurofibroma. The targeted neurofibroma nodules of 3 patients (27%) were nonresponsive to biological therapies (selumetinib [n = 2] or rapamycin [n = 1]). None of the patients displayed Tinel’s sign.

TABLE 1.

Patient and tumor-level data

Patient & Surgery ID*	Age (yrs) at Surgery, Sex	Resection Location	Reason for Surgery Referral; Symptoms
1A	26, F	Rt calf	History of MPNST, rapid size increase, PET avidity
1B	29, F	Lt gluteal	History of MPNST, rapid size increase, PET avidity
2A	13, M	Rt thigh	Family history of MPNST, rapid size increase, PET avidity
2B	16, M	Rt brachial plexus	Family history of MPNST, rapid size increase
3	22, M	Rt superficial thoracic	Exponential growth, chemotherapeutic agent AZD6244 was ineffective
4A	15, F	Rt neck	Pain, rapid size increase, PET avidity
4B	16, F	Rt supraclavicular fossa	Rapid size increase, PET avidity; pain at rest
5	18, F	Lumbar paraspinal (L2–3), posterior rib cage (T10)	Rapid size increase, PET avidity, chemotherapeutic agent rapamycin was ineffective
6	16, F	Lt retropharyngeal space	Family history of MPNST, rapid size increase
7	19, F	Rt C6 nerve root	Pain, rapid size increase, PET avidity, SUV increase; pain at rest
8A	41, M	Rt thigh	Size increase, biopsy-proven cellular atypia; tenderness
8B	41, M	Lt arm	History of ANF, size increase
9	20, F	Rt brachial plexus	New distinct nodule from preexisting plexiform neurofibroma
10	17, F	Lt posterior neck	Focal discomfort, AZD6244 (35 mg 2× day) was ineffective
11A	11, M	Lt gluteal	Discomfort w/ pressure on tumor; tenderness
11B	20, M	Rt gluteal	Rapid size increase, PET avidity; tenderness

ANF = neurofibroma with nuclear atypia; MPNST = malignant peripheral nerve sheath tumor; SUV = standardized uptake value.

^*In the alphanumeric patient identifiers, different numbers each signify individual patients, while the letters represent separate surgical interventions.

Imaging and Follow-Up Data

Prior to the resective surgeries, we performed serial volumetric MRI (n = 14, 88%) and ¹⁸F-FDG-PET imaging (n = 11, 69%) (Table 2). Thirteen (93%) of the 14 sets of serial preoperative MRI scans showed rapid growth (≥ 20% increase in volume per year) of the neurofibroma (median size 48.7 cm³, median growth rate 92% per year). Ten (91%) of the 11 ¹⁸F-FDG-PET scans indicated increased PET avidity (median SUV_Max 6.45). Two patients (patients 5 and 7) underwent needle biopsies of their targeted neurofibroma nodules to confirm ANF prior to their respective surgeries. The median duration of follow-up before surgery was 3.95 years, and the median duration of follow-up after surgery, as of this writing, is 2.45 years (range 0.00–10.39 years). Via postoperative imaging, we have found that to date none (0%) of the resected tumors have recurred. Oncological follow-up has not varied based on margin positivity.

TABLE 2.

Preoperative serial volumetric MRI-derived yearly growth rates and ¹⁸F-FDG-PET SUV_Max

Patient & Surgery ID	Duration of Screening Before Surgery (yrs)	SUVMax	Size (cm3)	Growth Rate (% vol change per yr)	FU After Surgery (yrs)	Recurrence
1A	6.3	6.58	45.8	111	4.42	No
1B	9.5	5.46	69.2	226	1.39	No
2A	1.2	6.1	34.4	78	4.94	No
2B	3.4	7.9	54.2	304	2.56	No
3	3	—	49.4	110	3.58	No
4A	3.8	13.8	149	124	2.91	No
4B	4.1	6.84	47.9	103	2.63	No
5	8.1	6.45	36	81	2.34	No
6	6.2	—	8.9	34	2.66	No
7	1.8	8.02	30.6	71	1.09	No
8A	0.3	6.3	318	—	1.38	No
8B	1	1.9	96.2	17	0.00	No
9	11.6	—	19.1	—	0.94	No
10	12.7	—	14.2	31	0.91	No
11A	3.8	—	61.8	355	10.39	No
11B	8.8	6.2	512	106	1.89	No

FU = follow-up; SUV_Max = maximum SUV within the tumor.

Electromyography and Nerve Conduction Abnormalities

We performed preoperative EMG and NCV studies before 8 (50%) of the surgeries (in 6 patients, 3 female and 3 male) (Supplementary Table 3). Preoperative studies were not feasible for 4 surgeries (25%). The findings of 3 (38%) of the studies were normal, and 5 (62%) were abnormal. Of the 5 abnormal motor studies, amplitude was decreased or not recordable in 2 (40%). Distal latencies and F-wave latencies were abnormal in 3 studies (60%) but were not within the demyelinating range. There was mildly decreased conduction velocity in all 5 studies (100%). Slowing of the conduction velocity across the site of tumor was observed in 1 study (in patient 8 before his second surgery [surgery B]). Otherwise, there was no evidence of a demyelinating neuropathy in any of the motor studies. Of the 5 sensory studies with abnormal findings, conduction velocity was decreased or not recordable in 4 (80%), and amplitude was decreased or not recorded in all 5 (100%). These findings were isolated to the nerve within the distribution of the tumor and may reflect tumor burden. Parent nerves of the resected neurofibromas included the posterior cord of the brachial plexus (n = 1), a sympathetic plexus (n = 1), the C6 (n = 1) and L2 (n = 1) nerve roots, and the sciatic (n = 4), deep peroneal (n = 1), femoral (n = 1), radial (n = 1), recurrent laryngeal (n = 1), spinal accessory (n = 1), and median (n = 1) nerves. The remaining parent nerves (n = 2) were unidentifiable superficial nerves.

Postoperative Outcomes

Gross-total, extracapsular resections of the targeted neurofibroma nodules were recorded by the surgeon (P.C.) in the notes for all 16 (100%) surgeries (Table 3). Ten (63%) surgeries resulted in no complications related to the parent nerves of the resected neurofibroma nodules. Six surgeries (37%) resulted in parent nerve–related complications, including muscle weakness (n = 2), numbness (n = 2), recurrent laryngeal nerve palsy (n = 1), and mild Horner’s syndrome (n = 1). Muscle weakness and numbness (n = 4) were transient and resolved prior to subsequent follow-up appointments. The recurrent laryngeal nerve palsy and mild Horner’s syndrome (n = 2) were persistent. Two surgeries (13%) resulted in complications that were not related to parent nerves; both complications were hematomas, which were immediately evacuated. Overall, most surgeries (n = 14, 88%) resulted in no persistent parent nerve–related complications; none resulted in additional persistent complications or postoperative infections. The median length of postoperative stay in the hospital was 3 days (discharge on postoperative day 3). For surgeries with postoperative complications (n = 6), the median length of postoperative stay was 4 days (transient = 3.5 days; persistent = 4 days).

TABLE 3.

Postoperative outcomes and tumor histopathological classification

Patient & Surgery ID	Nerve-Related Complications	Other Complications	No. of NFs Resected	Histopathology	NF Type	Surgical Margins
1A	No	No	1	ANF	Nodular	Pos/unreliable
1B	No	No	1	ANF	Plexiform	Neg
2A	Small area of numbness (transient)	No	1	ANF	Nodular	Neg
2B	Slight weakness of biceps (transient)	No	1	ANF	Nodular	Neg
3	No	No	1	ANF	Nodular	Neg
4A	Recurrent laryngeal nerve palsy (persistent)	No	1	lgMPNST	Nodular	Neg
4B	No	No	1	ANF	Plexiform	Neg
5	No	No	3	NF	Nodular	Neg
				ANNUBP	Nodular	Neg
				ANF	Nodular	Neg
6	Horner’s syndrome (persistent)	No	1	ANF	Plexiform	Neg
7	No	No	2	NF	Plexiform	Pos/unreliable
				ANNUBP	Plexiform	Neg
8A	Feet direction change/weakness (transient)	No	3	ANF	Nodular	Pos/unreliable
				ANF	Plexiform	Neg
				ANNUBP	Plexiform	Neg
8B	Decreased light touch sensation (transient)	No	1	NF	Nodular	Pos/unreliable
9	No	No	1	ANF	Nodular	Pos/unreliable
10	No	No	1	ANF	Nodular	Neg
11A	No	No	1	ANF	Nodular	Neg
11B	No	No	1	ANF	Plexiform	Neg

ANNUBP = atypical neurofibromatous neoplasm of uncertain biological potential; neg = negative; pos = positive.

Histopathological Analysis

Upon histopathological examination of the 21 resected tumors, 16 (76%) were confirmed to have marginal resections with circumferential inking visible on histopathological preparations. In 5 (24%), unreliable or positive margins were detected. During surgeries for 2 (40%) of the neurofibromas with unreliable or positive margins, at least 1 other neurofibroma was also simultaneously resected with marginal resection. Thirteen (62%) of the resected neurofibromas were solitary nodular tumors, whereas 8 nodules (38%) resided within larger, plexiform neurofibromas. Three tumors (14%) were benign neurofibromas with no atypical characteristics, 14 (67%) were ANFs, 3 (14%) were ANNUBPs, and 1 (5%) was an lgMPNST. No hgMPNSTs were discovered in this cohort. There were 2 ANNUBP subtypes: NF with nuclear atypia and additional hypercellularity (n = 2), and neurofibroma with nuclear atypia and additional high mitotic activity (n = 1). Eighteen (86%) of the 21 resected tumors demonstrated histological nuclear atypia or additional atypical characteristics, whereas 3 (14%) were benign neurofibromas (false positives). None of the neurofibromas with unreliable or positive margins were ANNUBPs or lgMPNSTs.

Imaging Correlates of Histopathological Diagnoses

The neurofibroma, ANF, ANNUBP, and lgMPNST groups had mean ¹⁸F-FDG-PET SUV_Max values of 1.9, 6.51 ± 0.83, 6.92 ± 0.95, and 13.8 and mean yearly growth rates on MRI of 17%, 139.2% ± 115.5%, 76% ± 7.1%, and 124%, respectively. The SUV_Max was significantly lower in the neurofibroma group than in the ANF (p = 0.0042), ANNUBP (p = 0.0038), or lgMPNST (p = 0.0001) group (Fig. 4 left; Dunnett’s multiple comparisons test). However, the difference in SUV_Max between the ANF and ANNUBP groups was not significant (unpaired t-test, p = 0.5238).The differences in yearly growth rates between the neurofibroma group and ANF (p = 0.6000), ANNUBP (p = 0.9452), or lgMPNST (p = 0.8313) group were also not significant (Fig. 4 right).

FIG. 4.

Relationship between ordinal histopathological neurofibroma groups and imaging characteristics. The maximum tumor uptake (SUV_Max) of ¹⁸F-FDG on PET imaging is significantly increased in tumors recognized as ANF, ANNUBP, or lgMPNST when compared with benign neurofibromas (left). No such between-group differences were found in tumor growth rates measured on MR images (right). NF = neurofibroma; ANF = neurofibroma with nuclear atypia; ANNUBP = atypical neurofibromatous neoplasm of uncertain biological potential; lgMPNST = low-grade MPNST. **p < 0.01; ****p < 0.0001.

Discussion

In the absence of effective medical therapies for ANF and MPNST, resection of ANFs in patients with NF1 has been recommended as a means of preventing malignant transformation. We selected patients for marginal resections of their neurofibroma nodules based on increased yearly growth rates, clinical symptoms, and increased ¹⁸F-FDG-PET avidity of neurofibroma nodules as reported in previous literature.^{5,17,21,25,30} Marginal resections of neurofibromas arising from peripheral nerves were achieved with minimal complications. While some patients did experience postoperative complications, these were mostly transient and resolved prior to subsequent follow-up.

Surgical Complications

The 2 persistent postoperative complications—recurrent laryngeal nerve palsy in patient 4 after her first surgery (surgery A) and mild Horner’s syndrome in patient 6—resulted from lesions to the recurrent laryngeal nerve and sympathetic plexus in the cervical region, respectively. Following histopathological examination of the neurofibroma arising from the recurrent laryngeal nerve, we found that this tumor met the histological criteria for a focal lgMPNST. Low-grade MPNSTs have been shown to have higher recurrence rates than ANF; thus, we believe the benefits of gross-total resection significantly outweighed the potential for neural injury.^2,18 Unilateral true vocal cord palsy in this patient was treated with laryngoplasty and thyroplasty. In the same patient, marginal resection of a spinal accessory nerve ANF was achieved in a second surgery (patient 4, surgery B) with no transient or persistent complications. These outcomes warrant further exploration into the feasibility of marginal resection of suspected ANF in the different cranial nerves and sympathetic plexuses of the cervical region. A previous literature review of cervical vagus nerve neurofibroma marginal resections validates the difficulty in avoiding postsurgical vocal cord paralysis and hoarseness, especially in cases of MPNST.⁴ The use of ultrasonography in tandem with MRI may improve diagnostic visualization of these rare neurofibromas of the vagus nerve and other cervical masses.²⁶

Growth Patterns of Atypical Neurofibroma Nodules

Identification of eccentric patterns of growth of the targeted ANF nodules within the parent nerves or plexiform neurofibromas was clinically relevant to our surgical planning and approach. Unlike schwannomas, which result solely from the excessive proliferation of Schwann cells in peripheral nerve sheaths, neurofibromas have been shown to arise from less-differentiated progenitor Schwann cells and include fibroblasts, perineural cells, and mast cells.^24,45 While both neurofibromas and schwannomas are benign, the developmental differences between the 2 tumor types have significant clinical implications for the approaches used for their respective resections and the resulting surgical margins.⁷ Schwannomas form fibrous capsules and expand eccentrically from the underlying nerve fascicles, allowing for removal of the tumors via gross-total resection with minimal morbidity.¹⁹ Neurofibromas, on the other hand, have been shown to encompass the underlying nerve fascicles in an intimate fashion, making gross-total resection more difficult around the capsules. Thus, subtotal resection of suspected ANFs has been used as a standard surgical approach to prevent damage to the nerve fascicles and avoid potential complications.² Although subtotal resection may not lead to any metastases or deaths, it does allow for local recurrences, which in many cases require reoperation.

Here, we demonstrate that the relatively rapid growth rates of the ANF nodules, likely from clonal expansion, resulted in eccentric, schwannoma-like tumor development (Fig. 5). Especially in ANFs arising from plexiform arising from plexiform neurofibroma masses, these eccentric growth patterns facilitated distinction of the ANF nodules from the remaining plexiform masses. This ultimately allowed us to attempt fascicle-sparing, circumferential dissection around the capsules of the targeted ANF nodules in tandem with intraoperative nerve stimulation to achieve marginal resection with minimal collateral peripheral nerve damage.³⁶ We further confirmed this with histopathological assessment of tumor margins, which revealed that circumferential, inked margins were detected in most cases. We verified upon postoperative imaging surveillance that none of these ANFs have recurred to date. These findings emphasize the value of inking neurofibroma nodules removed during surgery and proper sectioning of the lesion. In this study, none of the neurofibromas with unreliable or positive margins were ANNUBPs or lgMPNSTs.

FIG. 5.

A schematic representation of neurofibromas in NF1 patients. Neurofibromas are known to engulf many nerve axons within nerve fascicles with continued growth (left). NF1 patients typically have many neurofibroma nodules within adjacent nerve fascicles. We hypothesize that in NF1, occasional neurofibroma nodules may become dominant (asterisk) with rapid growth due to acquisition of cellular atypia (right). Such rapid growth may displace quiescent neurofibroma nodules and nerve fascicles to the margin on this nodule. This pattern may allow safe marginal dissection of the dominant nodule during surgery. Artists: Ethan Tyler, Erina He, and Alan Hoofring. Medical Arts, Office of Research Services, National Institutes of Health.

Imaging and Histopathology

Our preoperative surgical indicators, which included the increased yearly growth rates and ¹⁸F-FDG-PET uptake, were successful in predicting atypical or additional atypical characteristics of the targeted neurofibroma nodules in most cases. We found significant differences in SUV_Max between our benign neurofibroma stratum and the ANF, ANNUBP, and lgMPNST strata (Fig. 4 left). Given our minimal false-positive detection of benign neurofibromas but also lack of hgMPNST detection as determined by histopathological analysis, we confirm that preoperative imaging surveillance, especially through identification of increased ¹⁸F-FDG-PET SUV_Max, may serve as an excellent indicator for early detection of atypical transformation. Currently, no optimal cut-offs for SUV_Max exist to distinguish the neurofibroma, ANF, ANNUBP, and lgMPNST histopathological strata. Previous studies have suggested cut-offs, mostly between SUV_Max 3.0 and 4.0, to only differentiate between benign neurofibromas and MPNSTs.^8,38,41

We did not find a significant difference in SUV_Max between the ANF and ANNUBP strata, and due to our specific focus on ANF and therefore limited sample size, we did not capture many benign neurofibromas or lgMPNSTs in this study. However, our SUV_Max distribution for the ANF suggests that the SUV_Max cut-off points for ANF are likely near SUV_Max 5.46 and 7.9. Given that the preoperative tumor growth rates on MRI were highly variable among the ANF, our findings suggest that ¹⁸F-FDG-PET imaging may be more reliable in distinguishing histopathological grading. Genetic analysis of neurofibromas in NF1 patients has suggested that whereas loss of a single CDKN2A/B copy in completely NF1-inactived cells may result in ANF development, the loss of both CDKN2A/B copies may lead to the formation of an ANNUBP.^1,6,34 Further genetic studies will be required to assess a potential correlation between PET avidity and successive CDKN2A/B gene deletions in NF1 patients.

Conclusions

Reliable detection of ANF has traditionally been difficult, and the use of imaging surveillance to define timing and strategy for surgical intervention has not been previously evaluated. Here, we report on the use of preoperative imaging from MRI and ¹⁸F-FDG-PET SUV_Max to successfully detect and resect ANF with minimal morbidity. Our study demonstrates that safe, fascicle-sparing marginal resection of ANF can be achieved using intraoperative nerve stimulation and microdissection of nerve fascicles. Lastly, we are continuing to monitor our patients to assess for tumor recurrence.

Supplementary Material

Supple video

VIDEO 1. Resection procedure for right brachial plexus neurofibroma shown in Fig. 2. The patient is supine, with head toward the left side of the screen, and the right shoulder toward the bottom right corner. An initial skin incision and dissection through the platysma is performed. The parent nerve is identified and isolated, and its junction with tumor identified. Nerve fascicles traversing over the tumor surface are then identified visually and mapped with bipolar stimulation. Next, the nerve fascicles are incised along their main axis, and split to reveal the tumor capsule. Lastly, the neurofibroma is circumferentially dissected from the parent nerve while sparing parent nerve fascicles. Surgical procedure performed by Dr. Chittiboina at the NIH Clinical Center. Click here to view.

ABBREVIATIONS

ANF

atypical neurofibroma (neurofibroma with nuclear atypia)

ANNUBP

atypical neurofibromatous neoplasm of uncertain biological potential

EMG

electromyographic

hgMPNST

high-grade MPNST

lgMPNST

low-grade MPNST

MPNST

malignant peripheral nerve sheath tumor

NCI

National Cancer Institute

NCV

nerve conduction velocity

NF1

neurofibromatosis type 1

NIH

National Institutes of Health

SUV_Max

maximum SUV within the tumor