SCIENTIFIC ARTICLE

Whole-body MRI in neurofibromatosis: incidental findingsand prevalence of scoliosis

Jacob L. Jaremko & Peter J. MacMahon &

Martin Torriani & Vanessa L. Merker &

Victor F. Mautner & Scott R. Plotkin &

Miriam A. Bredella

Received: 2 October 2011 /Revised: 12 November 2011 /Accepted: 13 November 2011 /Published online: 7 December 2011# ISS 2011

AbstractObjective To demonstrate incidental findings and scoliosison whole-body MRI (WBMRI) in patients with neurofibro-matosis type 1 and 2 (NF1 & NF2, respectively), andschwannomatosis.Materials and methods Institutional review board approvaland written informed consent were obtained for this prospec-tive HIPAA-compliant study. A total of 247 subjects (141withNF1, 55 with NF2, 51 with schwannomatosis; 132 women(53.5%); mean age, 41 years, range, 18–97 years) underwentWBMRI using coronal STIR (TR/TE: 4190/111 ms, TI:150 ms) and T1-weighted images (TR/TE: 454/10 ms), 10-mm slice thickness, imaging time ~40 min. Images werereviewed for the presence of incidental findings, outside ofnerve sheath tumors. The presence of scoliosis was recordedand curve morphology was assessed and quantified.Results Incidental findings other than scoliosis were recordedin 104/247 (42%) patients, most often affecting the musculo-skeletal system (65/247 patients, 26%). We found 16/247

(6.5%) significant incidental findings likely to affect clinicalmanagement, including avascular necrosis of bone in eightpatients (five with NF2), eight insufficiency fractures, andfour non-neurogenic neoplasms (Hodgkin's lymphoma, lip-osarcoma, dermoid cyst, large uterine myoma requiring exci-sion). Scoliosis was seen in 50/247 patients (20%), including8/55 with NF2 (15%) and 11/51 with schwannomatosis(22%).Conclusions Incidental findings in the neurofibromatosesfrequently involve the skeleton. Given the relatively highincidence of unsuspected osteonecrosis and stress fractures,close attention to the skeleton on WBMRI is advised. Inaddition, knowledge of common incidental findings canhelp clinicians prepare patients who undergo WBMRI forpotential unexpected findings.

Keywords Whole-bodyMRI (WBMRI) .

Neurofibromatosis . Incidental findings

Introduction

Recent advances in imaging technology and clinical under-standing of phakomatoses have led to increased interest inwhole-body scans of neurofibromatosis types 1 and 2 (NF1and NF2, respectively) and schwannomatosis to determinetumor burden, evaluate the response to novel therapies, andto facilitate clinical follow-up [1–3]. Incidental findings arecommon in whole-body MRI (WBMRI) [4]. Understandingwhich incidental findings are expected can aid the clinicianin counseling patients about to undergo WBMRI, and canassist the interpreting radiologist in knowing which findingsto specifically assess for. Here, we sought to determine thetype and frequency of incidental findings in patients withNF1, NF2, and schwannomatosis. We hypothesized that (1)

Scott R. Plotkin and Miriam A. Bredella were equal senior authors.

J. L. Jaremko : P. J. MacMahon :M. Torriani :M. A. Bredella (*)Musculoskeletal Imaging and Intervention,Department of Radiology, Massachusetts General Hospitaland Harvard Medical School,Yawkey 6E, 55 Fruit Street,Boston, MA 02114, USAe-mail: mbredella@partners.org

V. L. Merker : S. R. PlotkinDepartment of Neurology and Cancer Center,Massachusetts General Hospital,Boston, MA 02114, USA

V. F. MautnerDepartment of Neurology,University Medical Center Hamburg-Eppendorf,Hamburg, Germany

Skeletal Radiol (2012) 41:917–923DOI 10.1007/s00256-011-1333-x

some of these findings are more common than in the generalpopulation, especially if linked in some way to the diseaseprocesses of the neurofibromatoses and common therapies;(2) certain incidental findings are sufficiently common and/or important to be routinely checked by radiologists reviewingWBMRI of neurofibromatoses patients; and (3) scoliosis isfrequent and of curve morphology different from idiopathicscoliosis.

Materials and methods

Research plan

As part of a prospective trial on tumor burden and distributionin neurofibromatoses performed in collaboration between theDepartments of Neurology and Radiology at two institutions,we reviewed a series of prospectively collected WBMRI andthe electronic medical records on patients with neurofibroma-tosis type 1 (NF1) or 2 (NF2), and schwannomatosis to detectand categorize incidental findings and to evaluate for andquantify the degree of scoliosis.

Subjects

This prospective study was performed with the approval of theinstitutional review board of both institutions, and was com-pliant with the Health Insurance Portability and AccountabilityAct. We included consecutive patients at least 18 years of agewho participated in a clinical trial between September 2007and November 2010 with a clinical diagnosis of NF1, NF2, orschwannomatosis based on National Institutes of Health crite-ria [5, 6]. Patients were excluded if theywere unable to providewritten consent in English for the study, could not undergoMRimaging while awake because of a medical or psychologicalcondition or metallic implant, or were pregnant or breast-feeding at the time of presentation. With these criteria, wehad 247 subjects who underwent whole-body MRI (141patients with NF1, 55 patients with NF2, and 51 patients withschwannomatosis); 132 were female (53.5%); and mean agewas 41 years (range, 18–97 years).

Whole-body MRI

WBMRI was performed in supine position in a 1.5-TMR unit(Siemens Avanto, Siemens Medical Systems, Malvern, PA)with an integrated full-body coil. We used two coronalsequences: T1-weighted turbo spin-echo (repetition time450 ms, echo time 10 ms, echo train length 3) and short-tauinversion recovery turbo spin echo (STIR; repetition time4,190 ms, echo time 111 ms, inversion time 150 ms, echotrain length 25). The slice thickness was 10mm, and the field ofview was 500 mm (320×240 matrix) for each of five imaging

stations with overlap of at least 40 mm between adjacentstations. To eliminate the need for whole-body localizers, wepositioned the initial imaging station at the patient’s chin andthe initial coronal section at the level of the tabletop. Totalimaging time was approximately 40 min (STIR 15 min, T125 min). Images from the five acquisitions were saved in theDigital Imaging and Communications in Medicine (DICOM)format and fused into a single whole-body series using softwareavailable on the MR workstation (Siemens Syngo, version MRB13 4VB13A; Siemens Medical Solutions).

Image analysis

All images were reviewed in consensus by a board certifiedsenior musculoskeletal radiologist (MAB, 13 years of experi-ence and 7 years experience in reading WBMRI) and one oftwo junior board certified radiologists (JLJ, PJM, each 6 yearsof experience) to record incidental findings and the presenceand extent of scoliosis. We excluded findings that were directlyattributable to the underlying disease (nerve sheath tumors,post-treatment changes related to surgery or radiation of thesetumors). For each case, incidental findings were recorded, aswell as whether each finding was a “significant” finding likelyto change patient management. We also assessed our imagingarchive and electronic medical record to seek previous descrip-tions of each incidental finding, and presence of any follow-upimaging studies or changes in clinical management that oc-curred as a result of that incidental finding.

Regarding scoliosis, we recorded the presence and extentof supine lateral spinal curvature based on curve location(thoracic, thoracolumbar, or lumbar), curve direction (convexleft or right) and curve magnitude based on measurement ofthe Cobb angle between perpendiculars to endplates at theupper and lower vertebrae forming each spinal curve. As mildspinal curves can be due to patient positioning and are unlikelyto progress clinically, we followed usual practice [7] by defin-ing significant scoliosis as curves with Cobb angle greater than10 degrees.

Statistical analysis

Statistical analysis was performed using Microsoft Excel(Microsoft, Inc., Redmond, WA, USA, v 10.6) and SPSS(SPSS Inc., Chicago, IL, USA, v. 9). Basic descriptive statis-tics are reported as mean ± standard deviation. Differencesbetween proportions using binary variables were tested usingFisher's exact test, at level of significance p<0.05, one-tailed.

Results

All WBMRIs were of diagnostic quality. There was at leastone incidental finding other than scoliosis in 104/247 (42%)

918 Skeletal Radiol (2012) 41:917–923

patients, of which 16/247 (6.5%) were significant findingslikely to affect clinical management (Table 1). Thirteenincidental findings (6.2% of patients) prompted follow-upimaging or intervention. The most frequent significant inci-dental finding was avascular necrosis (AVN) (eight patients,3.2%; mean age 43 years, range 25–70 years; Fig. 1). AVNwas seen in 5/55 (9.3%) of NF2 patients, which was signif-icantly more frequent compared to other diagnoses in thisstudy (p<0.05). Three of the NF2 patients received multiplecourses of corticosteroids in the past for hearing loss; noneof the patients had prior radiation at the site of AVN. Othersignificant incidental findings included previously unsus-pected insufficiency fractures (4/247, 1.6%; two in patientswith NF1 and two in schwannomatosis; Fig. 2), and fournon-neurogenic tumors (Fig. 3). A wide range of otherincidental findings was present (Fig. 4, Table 1). Therewas no evidence of diffuse marrow signal intensity abnor-malities or marrow infiltration.

Scoliosis with supine Cobb angle greater than 10° was seenin 56 curves in 50/247 patients (20%). Most curves weremoderate (mean 24°, median 21°), with ten curves greater

than 30° (4% of patients). Of the 56 curves, 24 (43%) wereright-convex thoracic, 12 were left thoracic (21%; 8/12 werein patients with NF1; Fig. 5), 12 left lumbar (21%), six rightlumbar (11%), and two left thoracolumbar (4%). When com-paring patients with each diagnosis, there was no significantdifference in the prevalence of scoliosis (31/141 with NF1,22%; 8/55 with NF2, 15%; 11/51 with schwannomatosis,22%; p>0.05); curve magnitudes (mean Cobb angle 23.8°with NF1, 23.8° with NF2, 26° with schwannomatosis) orproportion of left-convex thoracic scoliosis (p>0.05). By age,the proportion of patients with scoliosis was fairly constant at12–20% within each decade from 20–70 years (97% of ourpatients), while 86% patients over age 70 years had scoliosis.Women made up 54% of our study group and had 65% of thescoliotic curves.

Discussion

Our study demonstrated that incidental findings are com-mon in patients with neurofibromatosis using WBMRI

Table 1 Incidental findingson whole-body MRI inneurofibromatoses

Finding * (clinically significant) N (%)

Musculoskeletal 65 (26.3%)

* Avascular necrosis 8 (3.2%)

* Insufficiency fractures 4 (1.6%)

Enchondroma 6 (2.4%)

Nonossifying fibroma 3 (1.2%)

Fibrous dysplasia 3 (1.2%)

Bone cyst 1 (0.4%)

Intraosseous hemangioma 11 (4.5%)

Lipoma 4 (1.6%)

Effusion or periarticular fluid collection 16 (6.5%)

Rotator cuff tear 2 (0.8%)

Acetabular labral tear 1 (0.4%)

Iliopsoas bursitis 1 (0.4%)

Miscellaneous (osteitis pubis, intramuscularhemangioma, non-specific marrow edema)

5 (2.0%)

Tumor * (other than neurogenic) Hodgkin’s lymphoma, liposarcoma,ovarian dermoid, uterine myoma, requiring resection

4 (1.6%)

Visceral cysts 21 (8.5%)

Hepatic cysts/hemangioma 9 (3.6%)

Renal cyst(s) (>3 cm) 6 (2.4%)

Adnexal cyst(s), 1 requiring resection 3 (1.2%)

Other cysts (breast, thyroid) 3 (1.2%)

Structural variant 8 (3.2%)

Renal abnormality (atrophy, pelvic kidney) 3 (1.2%)

Urogenital abnormality (bladder diverticulum, arcuate uterus) 3 (1.2%)

Other (Richter’s hernia, apical bulla) 2 (0.8%)

Paranasal sinus opacification(complete or air/fluid level)

6 (2.4%)

TOTAL 104/247 (42%)

Skeletal Radiol (2012) 41:917–923 919

(42% total, 6.5% clinically significant). Incidental findingswere slightly less common in a prior WBMRI study ofhealthy young adult subjects (29% total, 3.4% highly clinical-ly significant) and increased significantly with age and in-creasing body mass index [4]. Our study also demonstrated ahigher frequency of incidental findings related to bone thanseen in other reports, whether of whole-body MRI [4] orseparate body regions [8, 9]. This might be due to the natureof neurofibromatosis. There were insufficiency fractures in1.6% of NF1 patients. A potential mechanism could be thegeneralized mesodermal abnormality in this condition, whichcan lead to osteoporosis [10, 11]. Fibrous dysplasia, one of the‘distinctive bone lesions’ included in the clinical diagnosticcriteria for NF1 [12], was less frequent (1.2%) in our studycompared to another series in pediatric patients with NF1(4%) [13]. Focal bone lesions such as nonossifying fibromasor enchondromas (total 4%) were less common in our studythan in a pediatric series in NF1 (38%) [14]. This might be dueto spontaneous “healing/filling in” of these lesions during

adolescence. The rate of enchondromas (2.4%) was similarto the 2.9% seen in healthy adults having MRI for internalderangement of the knee [15]. In a study of incidental findingsin healthy adults undergoing WBMRI [4], the only bonylesion identified was one solitary bone cyst, although the onlywhole-body images obtained in that study were axial T1-weighted, limiting sensitivity.

In our study, 9.1% of patients with NF2 were found to haveAVN. Three of these patients received prior corticosteroidtherapy for sudden hearing loss, suggesting that patients trea-ted with these medications should be monitored for AVN.

In our study there were incidental renal and hepatic T2-intense lesions in 2.4% and 3.6% of patients, respectively. In areported study of incidental findings in WBMRI, a higher rateof renal cysts (14%) and similar rate of hepatic T2-hyperintenselesions (3.4%) were found [4]. In addition to 10-mm T1-weighted axial images that study used axial T2 GRE imagingof the upper abdomen, which was likely more sensitive to smallrenal cysts than our coronal 10-mm T1 and STIR sequences.

Fig. 1 A 32-year-old man withneurofibromatosis-2 (NF2).Portion of STIR (a) andT1-weighted (b) coronalwhole-body MRI showingbilateral curvilinear signalabnormalities in femoral heads(arrows), typical for avascularnecrosis

Fig. 2 Insufficiency fracture on whole-body MRI in a 48-year-oldwoman with neurofibromatosis-1 (NF1). Portion of coronal STIR (a)and T1-weighted (b) whole-body MRI showing T2 hyperintensitysurrounding an incomplete basicervical fracture line at the lateral

aspect of left femoral neck (arrows). This patient had fallen 2 weeksprior to the scan, but was ambulatory. After orthopedic consultation,she was treated conservatively. Subsequent dual-energy X-ray absorp-tiometry (DEXA) showed her to be osteoporotic

920 Skeletal Radiol (2012) 41:917–923

Allowing for this, the underlying rates of visceral cysts in ourpatients are likely similar to healthy adults.

The rate of scoliosis (20%) in our patients matched the 10–20% range observed elsewhere in neurofibromatosis [16] andis much higher than the 1–2% seen in the general population[7]. Nearly one-quarter (22%) of our scoliotic curves wereleft-convex thoracic, a pattern rarely seen in idiopathic scoli-osis but frequent in NF1, especially in dystrophic-type curves,which have a higher risk of progression and poor prognosis[17, 18]. Whole-spine MRI has been shown to identify previ-ously unsuspected cases of this higher-risk dystrophic curva-ture [19]. The 10-mm-thick coronal slices with large field ofview in this study did not permit enough bony detail toconfidently demonstrate or exclude the presence of segmen-tation anomaly or dysplastic spinal elements as the cause ofthe scoliosis. For this reason, we did not directly classify

Fig. 3 Incidental neoplasms discovered at whole-body MRI. a 30-year-old woman with neurofibromatosis-1. Portion of coronal STIR whole-bodyMRI showing heterogeneously T2-hypointense anterior mediastinalmass (arrow). b Follow-up axial fused PET-CT image with intravenouscontrast in the same patient showing hypermetabolic activity in the mass(arrow). The mass was subsequently found to represent Hodgkin’s lym-phoma on biopsy and was treated with chemotherapy. c 28-year-old

woman with neurofibromatosis-1. Portion of coronal T1 whole-bodyMRI showing a large, markedly T1-hyperintense round left pelvic mass(arrow). Subsequent excision confirmed a dermoid cyst. d, e 64-year-oldman with neurofibromatosis-1. Portion of coronal T1-weighted and STIRwhole-bodyMRI showing a small ovoid T2-hyperintense lesion, with T1signal isointense to muscle (arrows), in the left distal anterior thigh. Thiswas resected and found to represent a liposarcoma

Fig. 4 Nonaggressive incidentalbone lesions discovered atwhole-body MRI. A 24-year-oldwoman with neurofibromatosis-1. Portion of coronal T1-weighted whole-body MRIshowing bilateral distal femoralcortical deformity, mild expan-sion, and marrow T1 hypointen-sity (arrows), suggesting largenonossifying fibromas or fibrousdysplasia

Skeletal Radiol (2012) 41:917–923 921

curves as dystrophic or non-dystrophic. In cases of previouslyunknown scoliosis detected on WBMRI, we suggest erect PAand lateral spinal radiographs for further evaluation, withdedicated spinal cross-sectional imaging to follow, if clinicallyappropriate.

NF2 and schwannomatosis are much rarer than NF1 [20,21]. To our knowledge, this study is the first report of theprevalence of scoliosis in adults with these conditions (14%in NF2, 23% in schwannomatosis). The curve types, mag-nitudes, and directions in both conditions resembled thoseseen in NF1, though we had insufficient statistical power totest these distinctions at this study size. It is possible that thescoliosis seen in NF2 and schwannomatosis might be due toa mesodermal abnormality of bone as in NF1, and/or tochronic adaptation to presence of paraspinal tumors.

Our study had several limitations. First, the 10-mm slicethickness of our images limits resolution of small lesions.Second, measurement of scoliosis by Cobb angle on supine

MRI underestimates the degree of scoliosis seen on erectradiographs, due to lack of gravity load on the spine. In astudy where supine and erect Cobb angles were measured atthe same visit, the mean curvature increased from 23–31degrees when standing [22]. Thus, the prevalence and curvemagnitude of scoliosis in our patients, already high onsupine WBMRI, would both be further increased if erectspinal radiographs were performed on all patients. Thisunderscores the point that scoliosis is highly associated withall the neurofibromatoses.

Third, due to the nature of our study as a retrospectivereview of chart and imaging data, we cannot be certainwhether incidental findings such as scoliosis had been iden-tified clinically prior to the scan. In the case of scoliosis, it islikely that many mild cases were clinically occult, whilemost severe cases would have been diagnosed previously.None of the highly significant incidental findings (insuffi-ciency fractures, avascular necrosis, and non-neurogenictumors) had been identified clinically prior to WBMRI.

Fourth, most of our diagnoses of benign bone lesionswere presumptive, since we generally had no subsequentimages or biopsy of these lesions, and rarely hadcorresponding clinical history, although this is unsurprisingsince lesions such as nonossifying fibroma or focal fibrousdysplasia are expected to be asymptomatic.

Finally, we did not have clinical follow-up in all patients,as many of our patients had primary care elsewhere and onlyvisited our site in consultation. It is possible that our ob-served incidental findings resulted in a change of clinicalmanagement in more patients than we could report, espe-cially the significant findings.

Although we detected important incidental findings byWBMRI, we do not suggest use of WBMRI in the neuro-fibromatoses with the primary purpose of detecting thesefindings. Rather, we suggest that WBMRI performed forassessment of tumor burden should be carefully inspectedto detect occult lesions unrelated to neurogenic tumors thatmay influence management, such as fractures, avascularnecrosis, or other malignancies.

In conclusion, in relation to our initial hypotheses wefound that (1) musculoskeletal incidental findings are morecommon in the neurofibromatoses than expected in thegeneral population, and are clearly visible on WBMRI. (2)Insufficiency fractures and AVN should be specificallysought by radiologists. (3) There is a substantial, previouslyunreported incidence of moderate scoliosis in adults withschwannomatosis and NF2. We also again confirmed thewell-known high incidence of left-convex thoracic scoliosiscurves in NF1. Knowledge of these patterns of incidentalfindings can help clinicians prepare patients who undergoWBMRI for potential unexpected findings.

Fig. 5 Scoliosis. Coronal STIR (a) and T1-weighted (b) whole-bodyMRI showing the typical sharp focal left-convex upper thoracic scoli-osis of neurofibromatosis-1, with apex marked by arrows, in a 36-year-old woman

922 Skeletal Radiol (2012) 41:917–923

Conflict of interest The authors declare that they have no conflictsof interest.

Funding support Department of Defense Neurofibromatosis Re-search Program (NF0502020) and the NIH/NINDS (P01NS024279).

References

1. Cai W, Kassarjian A, Bredella MA, et al. Tumor burden in patientswith neurofibromatosis types 1 and 2 and schwannomatosis: deter-mination on whole-body MR images. Radiology. 2009;250:665–73.

2. Mautner VF, Asuagbor FA, Dombi E, et al. Assessment of benigntumor burden by whole-body MRI in patients with neurofibroma-tosis 1. Neuro Oncol. 2008;10:593–8.

3. Van Meerbeeck SF, Verstraete KL, Janssens S, Mortier G. Whole-body MR imaging in neurofibromatosis type 1. Eur J Radiol.2009;69:236–42.

4. Morin SH, Cobbold JF, Lim AK, et al. Incidental findings inhealthy control research subjects using whole-body MRI. Eur JRadiol. 2009;72:529–33.

5. Gutmann DH, Aylsworth A, Carey JC, et al. The diagnostic eval-uation and multidisciplinary management of neurofibromatosis 1and neurofibromatosis 2. JAMA. 1997;278:51–7.

6. MacCollin M, Chiocca EA, Evans DG, et al. Diagnostic criteria forschwannomatosis. Neurology. 2005;64:1838–45.

7. Soucacos PN, Zacharis K, Soultanis K, Gelalis J, Xenakis T, BerisAE. Risk factors for idiopathic scoliosis: review of a 6-year pro-spective study. Orthopedics. 2000;23:833–8.

8. Glockner JF. Incidental findings on renal MR angiography. AJRAm J Roentgenol. 2007;189:693–700.

9. McKenna DA, Laxpati M, Colletti PM. The prevalence of inci-dental findings at cardiac MRI. Open Cardiovasc Med J.2008;2:20–5.

10. Abdel-Wanis ME, Kawahara N. Bone development in neurofibro-matosis 1. Med Hypotheses. 2003;60:459–62.

11. Tucker T, Schnabel C, Hartmann M, et al. Bone health and fracturerate in individuals with neurofibromatosis 1 (NF1). J Med Genet.2009;46:259–65.

12. Jett K, Friedman JM. Clinical and genetic aspects of neurofibro-matosis 1. Genet Med. 2010;12:1–11.

13. Alwan S, Tredwell SJ, Friedman JM. Is osseous dysplasia a pri-mary feature of neurofibromatosis 1 (NF1)? Clin Genet.2005;67:378–90.

14. Mandell GA, Dalinka MK, Coleman BG. Fibrous lesions in thelower extremities in neurofibromatosis. AJR Am J Roentgenol.1979;133:1135–8.

15. Walden MJ, Murphey MD, Vidal JA. Incidental enchondromas ofthe knee. AJR Am J Roentgenol. 2008;190:1611–5.

16. Crawford AH, Herrera-Soto J. Scoliosis associated with neurofi-bromatosis. Orthop Clin North Am. 2007;38:553–62. vii.

17. Winter RB, Moe JH, Bradford DS, Lonstein JE, Pedras CV, WeberAH. Spine deformity in neurofibromatosis. A review of one hun-dred and two patients. J Bone Joint Surg Am. 1979;61:677–94.

18. Sirois 3rd JL, Drennan JC. Dystrophic spinal deformity in neuro-fibromatosis. J Pediatr Orthop. 1990;10:522–6.

19. Ramachandran M, Tsirikos AI, Lee J, Saifuddin A. Whole-spinemagnetic resonance imaging in patients with neurofibromatosis type1 and spinal deformity. J Spinal Disord Tech. 2004;17:483–91.

20. Lu-Emerson C, Plotkin SR. The Neurofibromatoses. Part 1: NF1.Rev Neurol Dis. 2009;6:E47–53.

21. Lu-Emerson C, Plotkin SR. The neurofibromatoses. Part 2: NF2and schwannomatosis. Rev Neurol Dis. 2009;6:E81–6.

22. Wessberg P, Danielson BI, Willen J. Comparison of Cobb angles inidiopathic scoliosis on standing radiographs and supine axiallyloaded MRI. Spine (Phila Pa 1976). 2006;31:3039–44.

Skeletal Radiol (2012) 41:917–923 923