MRI Diagnosis of Clival Cancer and Sixth Nerve Palsy

Imaging diagnosis of clival cancer may be difficult, in part because of normal variation in marrow signal with aging. Identifying whether clival cancer has damaged the sixth cranial nerve is a further challenge because minimal clival abnormalities could impinge on the nerve, which travels very close to the clivus.

Original Contribution Section Editors: Clare Fraser, MD Susan Mollan, MD MRI Diagnosis of Clival Cancer and Sixth Nerve Palsy Ahmad Halawa, MD, John Kim, MD, Eric Liao, MD, Jonathan D. Trobe, MD C Background: Imaging diagnosis of clival cancer may be difficult, in part because of normal variation in marrow signal with aging. Identifying whether clival cancer has damaged the sixth cranial nerve is a further challenge because minimal clival abnormalities could impinge on the nerve, which travels very close to the clivus. Methods: Two neuroradiologists, who were unaware of previous imaging and clinical diagnoses, reviewed MRI studies of 25 patients with cancer but no clival involvement and no sixth nerve palsy, 24 patients with clival cancer but without sixth nerve palsy, and 31 patients with clival cancer and sixth nerve palsy. The radiologists were tasked with determining whether there was clival cancer, whether there was a sixth nerve palsy and its laterality, and with indicating the pulse sequences used to make those determinations. Results: Both neuroradiologists correctly identified all 25 cases with a normal clivus. In about half of those cases, they depended on finding a homogeneously bright marrow signal; in the remaining cases, they excluded cancer by determining that the clivus was not expanded and that there were no focal signal abnormalities. Both neuroradiologists correctly identified clival cancer in 54 (98%) of the 55 cases with and without sixth nerve palsy. In doing so, they relied mostly on clival expansion but also on focal signal abnormalities. Both neuroradiologists were at least 80% correct in identifying a sixth nerve palsy, but they often incorrectly identified a palsy in patients who did not have one. When there was a one-sided signal abnormality or the clivus was expanded in one direction, both neuroradiologists were accurate in identifying the side of the sixth nerve palsy. Conclusion: Current MRI pulse sequences allow accurate differentiation of a normal from a cancerous clivus. When the marrow signal is not homogeneously bright in adults, cancer can be diagnosed on the basis of clival expansion or focal signal abnormalities. MRI is less accurate in predicting the presence of a sixth nerve palsy. However, the side of a unilateral palsy can be predicted when the clivus is clearly expanded in one direction or there is a focal signal abnormality on one side. ancer of the clivus is caused by primary osseous or chondral lesions, such as chordomas or chondrosarcomas, local spread of oropharyngeal masses, and metastasis of solid or hematopoietic tumors (1–4). Because the sixth nerve traverses the tight space of the Dorello canal adjacent to the clivus, it is readily compromised by a tumor, such that sixth nerve palsy may be a presenting sign (1–14). When cancerous infiltration of the clivus is limited to the marrow, as is especially common with metastases, it may produce only subtle imaging abnormalities. Radiologists depend on finding loss of the uniform precontrast T1 shortening bright signal associated with normal age-related fatty replacement of hematopoietic marrow (15). However, marrow replacement proceeds at a variable pace with normal aging (16). In 1 report, a precontrast T1 sagittal signal was uniformly bright in nearly all patients by age 24 (17), but in a subsequent study (18), such uniform brightness was present in only one-third of the cohort. When there is nonuniform speckled, heterogeneous clival marrow signal intensity in adults, exclusion of cancer depends on finding a comparable signal in the calvarial and cervical vertebral marrow and lack of clival expansion or focal T1 enhancement, T2 brightness, or restricted diffusion (19). The accuracy of MRI diagnosis of clival cancer and whether it can predict the presence of a sixth nerve palsy have not been widely documented. The goals of this study were to discover whether neuroradiologists can reliably identify a cancerous clivus from MRI studies alone, the pulse sequences they use in that task, and whether they can predict the presence of a sixth nerve palsy and its location. Journal of Neuro-Ophthalmology 2023;43:126–130 doi: 10.1097/WNO.0000000000001655 © 2022 by North American Neuro-Ophthalmology Society METHODS Department of Ophthalmology and Visual Sciences (AH, JDT), (Kellogg Eye Center), Ann Arbor, Michigan; Department of Radiology (JK, EL) (Neuroradiology), Ann Arbor, Michigan; and Department of Neurology (JDT), University of Michigan, Ann Arbor, Michigan. The authors report no conflicts of interest. Address correspondence to Jonathan D. Trobe, MD, 1000 Wall Street, Ann Arbor, MI 48105 [e-mail] 126 We used the Michigan Medicine electronic medical record search engine called EMERSE (20) to identify 3 patient groups: Group 1: 25 patients with a diagnosis of cancer, but whose brain imaging had been reviewed by at least 2 neuroradiologists as being free of clival cancer, and examined by at least 1 ophthalmologist to determine that they do not have a sixth nerve palsy; Group 2: 24 patients with a diagnosis of clival cancer as determined by at least 2 Halawa et al: J Neuro-Ophthalmol 2023; 43: 126-130 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution neuroradiologists and examined by at least 1 ophthalmologist to determine that they do not have a sixth nerve palsy; Group 3: 31 patients with a diagnosis of clival cancer as determined by at least 2 neuroradiologists and examined by at least 1 ophthalmologist to determine that they have a sixth nerve palsy attributable to the cancer. The 2 clinician authors (AH and JDT) selected patients who met those criteria and had undergone brain MRI studies at Michigan Medicine within 6 months of the clinical diagnosis. Two Michigan Medicine faculty neuroradiologists (JK and EL) were tasked with reviewing deidentified MRI studies of all patients. They had been told that some cases had a normal clivus, that some cases had a cancerous clivus, and that a subset of patients with a cancerous clivus also had a sixth nerve palsy. They were not told how many patients were in each group. They were instructed to determine whether cancer was present in the clivus, to note the imaging features that factored into that determination, to indicate whether there was a sixth nerve palsy, and if so, to indicate the side of the palsy. The clinical criteria used to designate a sixth nerve palsy were an abduction deficit together with an incomitant esotropia after excluding other causes of these features. RESULTS Patient Demographics and Causes of the Lesions Patients ranged in age from 1 to 81 years (median 56) in Group 1, from 17 to 78 years (median 56) in Group 2, and from 7 to 93 years (median 61) in Group 3. Men were 7 of 25 in Group 1, 11 of 24 in Group 2, and 17 of 31 in Group 3. The causes of clival cancer in Group 2 (no sixth nerve palsy) were metastases 12 (50%), chordoma 7 (29%), nasopharyngeal carcinoma 3 (13%), cholesteatoma 1 (4%), and oral squamous cell carcinoma 1 (4%). The metastatic tumors were from breast 4, multiple myeloma 3, lung 2, prostate 1, synovium 1, and pancreas 1. The causes of clival cancer in Group 3 (sixth nerve palsy) were metastases 22 (71%), chordoma 5 (17%), chondrosarcoma 2 (6%), sphenoid sinus carcinoma 1 (3%), and fibroelastic tissue 1 (3%). The metastatic tumors were from breast 5, multiple myeloma 4, prostate 4, liver 2, gastrointestinal tract 2, leukemia 1, lymphoma 1, lung 1, neuroendocrine 1, and rhabdomyosarcoma 1. The sixth nerve palsy was always unilateral (11 on the right side and 20 on the left side). Accuracy in Identifying a Normal Clivus Both reviewing neuroradiologists correctly indicated that all 25 cases in Group 1 had a normal clivus (Table 1). In the 24 cases involving adults, neuroradiologist 1 made that determination by finding a homogeneously bright clival marrow on precontrast T1 imaging in 13 (54%) cases, and neuroradiologist 2 made that determination by finding a homogeneously bright clival marrow on precontrast T1 imaging in 10 (42%) cases (Fig. 1A). In the remaining adult cases, both neuroradiologists found the precontrast T1 signal to be heterogeneous so that they based their diagnosis of a cancer-free clivus on similar signal heterogeneity in calvarial and cervical vertebral marrow and on the absence of focal postcontrast T1 enhancement, T2 fat-suppressed bright signal, and restricted diffusion on DWI (Fig. 1B). In the single case involving a 10-year-old child, both neuroradiologists determined that a uniformly dark marrow signal was sufficient to exclude cancer. In the single case involving a 1-year-old, the precontrast T1 signal was heterogeneous and considered noncancerous because it was similar to adjacent bones. Accuracy in Identifying a Cancerous Clivus The reviewing neuroradiologists were each correct in identifying cancer in 54 (98%) of 55 cases (Table 2). Neuroradiologist 1 failed to diagnose a single case of metastatic lung cancer, acknowledging after joint review that the marrow was heterogeneous but that focal signal abnormalities had been overlooked. Neuroradiologist 2 failed to diagnose a single case of metastatic breast cancer, acknowledging after joint review that a homogeneously dark marrow had been overlooked. In the cases with heterogeneous marrow signal, both neuroradiologists largely used the finding of clival expansion or a focal clival lesion to make a diagnosis of cancer (Fig. 2). The precontrast T1 sequence was most valuable, followed by the postcontrast T1 fat-saturated sequence, and then the T2 fat-saturated sequence. Diffusion-weighted imaging was found to contribute the least, likely due to geometric TABLE 1. MRI features used to correctly identify a normal clivus in 25 cases MRI Features Homogeneously bright marrow signal Homogeneously dark marrow signal* Heterogeneous marrow signal but normal postcontrast T1, fat-suppressed T2, and diffusion-weighted imaging Cases With Those Features Identified by Neuroradiologist 1 Cases With Those Features Identified by Neuroradiologist 2 13 1 11 10 1 14 *Child aged 10 years. Halawa et al: J Neuro-Ophthalmol 2023; 43: 126-130 127 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution FIG. 1. A. Complete fatty marrow conversion in normal aging, reflected in homogeneously bright precontrast T1 signal in sagittal (top left) and axial (top right) planes (arrow). B. Incomplete fatty marrow conversion in normal aging, reflected in heterogeneous precontrast T1 signal in sagittal (top left) and axial (top middle) planes (arrow). Cancer was excluded because there was no abnormal signal on postcontrast T1 fat-saturated (top right), T2 fat-saturated (bottom left), and diffusionweighted (bottom right) sequences. distortion of the diffusion sequence at the skull base. There was no significant difference in the distribution of the focal signal abnormalities in patients with and without sixth nerve palsy. In the single case of a 7-year-old, the signal was heterogeneous and metastatic rhabdomyosarcoma was identified by clival expansion. Accuracy in Identifying Sixth Nerve Palsy in Patients with Clival Cancer Among the 31 patients who had a sixth nerve palsy (Group 3), neuroradiologist 1 correctly diagnosed a sixth nerve palsy in 25 (81%) and neuroradiologist 2 correctly diagnosed a sixth nerve palsy in 30 (97%) (Fig. 3A). However, they often over diagnosed sixth nerve palsy in the Group 2 patients, none of whom had such a palsy. Thus, neuroradiologist 1 incorrectly diagnosed such a palsy in 18 (75%) of 24 cases (25% specificity) and neuroradiologist 2 incorrectly diagnosed such a palsy in 23 (96%) of 24 cases (4% specificity). Both neuroradiologists were more accurate in identifying the side of a sixth nerve palsy. Thus, neuroradiologist 1 correctly predicted the side of the palsy in 17 (55%) of 31 cases, based on identifying relatively unilateral clival expansion in 13 cases and on identifying a unilateral clival TABLE 2. Predominant MRI features used to correctly identify a cancerous clivus in 54 cases Predominant MRI Features* Clival expansion Focal clival signal abnormalities Diffuse clival signal abnormalities on multiple pulse sequences without clival expansion or focal signal abnormalities Homogenously hypointense precontrast T1 marrow signal Cases With Those Features Identified by Neuroradiologist 1 Cases With Those Features Identified by Neuroradiologist 2 36 (67%) 11 (20%) 4 (7%) 26 (48%) 20 (35%) 3 (6%) 4 (7%) 6 (11%) *In some cases, there was more than 1 predominant MRI feature. 128 Halawa et al: J Neuro-Ophthalmol 2023; 43: 126-130 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution FIG. 2. Cancerous clivus. Heterogeneous precontrast T1 signal with outlined left hypointense T1 signal (left), which demonstrates corresponding focal enhancement on postcontrast T1 fat-saturated (center) and hyperintense T2 signal (right). lesion in 4 cases. In the 14 cases in which neuroradiologist 1 incorrectly predicted the side of the palsy, he believed that the palsy would be on both sides because the clivus was expanded in both directions (Fig. 3B). Neuroradiologist 2 correctly predicted the side of the palsy in 22 (70%) cases, based on finding relatively unilateral clival expansion in 11 cases and on finding a focal unilateral clival lesion in 11 cases. In the 9 incorrectly diagnosed cases, he believed that the palsies would be bilateral, given that there was bilateral clival expansion or bilateral focal clival lesions. The principal implication of these results is that current MRI sequences provide a reliable way to separate a normal from a cancerous clivus. Radiologists can expect to dismiss cancer in about 50% of cases on the basis of a homogeneously bright marrow in adults and a homogeneously dark marrow in children. In the remaining 50% of cases, they will encounter a heterogeneous marrow signal and must then rely principally on focal T1 hypointensity, the absence of clival expansion or enhancement on T1 fat-suppressed CONCLUSIONS The 2 reviewing neuroradiologists, who were masked to the presence of clival cancer and sixth nerve palsy, were 100% accurate in identifying the 25 normal clivus cases. In about half of those cases, their determination was based on identifying a homogeneously bright precontrast T1 signal in the 24 adults and a homogeneously dark precontrast T1 signal in the 1 child. In the remaining half of cases, the marrow signal was heterogeneous so that the diagnosis of a normal clivus was based on finding normal postcontrast T1, fat-suppressed T2, and diffusion-weighted imaging (Table 1). Both neuroradiologists were 98% accurate in identifying a cancerous clivus. In doing so, they relied chiefly on finding clival expansion and to a lesser extent on finding focal signal abnormalities (Table 2). Both neuroradiologists were much less accurate in predicting the presence of a sixth nerve palsy. Although they identified most of the sixth nerve palsies, they also incorrectly identified palsies in many patients who did not have such palsies (specificity 4% and 25%). When they found imaging indications of clival cancer, they evidently assumed that a sixth nerve palsy would also be present. On the other hand, the 2 neuroradiologists were reasonably accurate in predicting the side of a sixth nerve palsy (55% and 70%). That determination was based on finding asymmetric lateral expansion of the clivus or a unilateral focal lesion. When they were wrong about the side of the palsy—which was always unilateral—it was in diagnosing bilateral palsies in patients whose clival expansion or focal lesions were bilateral. Halawa et al: J Neuro-Ophthalmol 2023; 43: 126-130 FIG. 3. Cancerous clivus with left sixth nerve palsy correctly predicted (top panels). Outlined discrete left hypointense pre-contrast clival T1 signal (top left) displays heterogenous enhancement on post-contrast T1 (top right), T2 hyperintensity (bottom left), and restricted diffusion (bottom right). Both reviewing neuroradiologists correctly predicted the side of the palsy. Cancerous clivus with left sixth nerve palsy incorrectly predicted (bottom panels). Pre-contrast T1 sequence (bottom left) displays outlined diffuse low T1 signal with clival expansion on both sides. Post-contrast T1 fat-saturated imaging (bottom middle) shows associated enhancement and T2 fat-saturated imaging (bottom right) shows T2 hyperintensity. The radiologists’ errors were attributed to the bilateral clival expansion. attributed to the bilateral clival expansion. 129 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution imaging, focal fat-suppressed T2 hyperintensity, and to a much lesser extent on abnormal restricted diffusion. As reliable as MRI may be in the identification of clival cancer, it is less helpful in predicting whether a sixth nerve palsy is present. The reviewing neuroradiologists in this study hugely overcalled the presence of a sixth nerve palsy, assuming that clival expansion or focal signal abnormalities would always cause such a palsy. On the other hand, they were quite accurate in predicting the side of the palsy when the expansion and focal signal abnormalities were clearly directed toward one side. We acknowledge that our study conditions do not entirely mimic real-world radiologic practice. For example, the reviewing neuroradiologists were told that clival cancer was present in some cases, so they could direct attention at the critical diagnostic features, including the precontrast T1 sagittal pulse sequence, which might be less scrutinized when there is no suspicion of cancer. Another limitation of this study is that the “gold-standard” diagnoses were based on previous review by nonstudy radiologists, which could be disputed. Finally, our cohort included only 3 children younger than 10 years. However, the distribution of cancer types is comparable to that of reported series (1). The radiologists had access to the full range of standard pulse sequences. Thus, we believe our results can be generalized to conclude that, with proper deployment of multiple pulse sequences and an awareness of cancer somewhere in the body, radiologists should be able to recognize clival cancer at any patient age. The variable development of yellow marrow with normal aging should not be a limiting factor. However, imaging is a poor predictor of whether clival cancer has caused a sixth nerve palsy, presumably because the sixth nerve passes through a tight space (Dorello canal) adjacent to the clivus such that a minimal expansion could produce a palsy. Predicting the side of a sixth nerve palsy is also challenging unless the clivus is clearly expanded in one direction or the focal signal abnormality is located on one side. STATEMENT OF AUTHORSHIP Conception and design: A. Halawa, J. Kim, E. Liao, J. D. Trobe; Acquisition of data: A. Halawa, J. Kim, E. Liao, J. D. Trobe; Analysis and interpretation of data: A. Halawa, J. Kim, E. Liao, J. D. Trobe; Drafting the manuscript: J. D. Trobe; Revising the manuscript for intellectual content: A. Halawa, J. Kim, E. Liao, J. D. Trobe; Final approval of the completed manuscript: A. Halawa, J. Kim, E. Liao, J. D. Trobe. REFERENCES 1. Jozsa F, Das JM. Metastatic lesions of the clivus: a systematic review. World Neurosurg. ;158:190–204. 2. Neelakantan A, Rana AK. Benign and malignant diseases of the clivus. Clin Radiol. 2014;69:1295–1303. 130 3. Dekker SE, Wasman J, Yoo KK, Alonso F, Tarr RW, Bambakidis NC, Rodriguez K. Clival metastasis of a duodenal adenocarcinoma: a case report and literature review. World Neurosurg. 2017;100:62–68. 4. Pallini R, Sabatino G, Doglietto F, Lauretti L, Fernandez E, Maira G. Clivus metastases: report of seven patients and literature review. Acta Neurochir (Wien). 2009;151:291–296; discussion 296. 5. Amouzgarhashemi F, Vakilha M, Sardari M. An unusual metastatic breast cancer presentation: report of a case. Iran J Radiat Res. 2005;3:43–45. 6. Chandrashekhar MN, Kishore K, Lakshmaiah V, Nagesha CK, Kumar BN. Clivus metastasis presenting as isolated abducens nerve palsy – a case report. J Evol Med Dent Sci. 2013;43:8383–8385. 7. Turner JL, Sweeney P, Hardy R. Ewing’s tumor metastatic to the clivus, simulating meningitis: case report. Neurosurgery 1980;7:619–620. 8. Fumino M, Matsuura H, Hayashi N, Arima K, Yanagawa M, Kawamura J. A case of renal cell carcinoma with metastasis in clivus presenting as diplopia. Hinyokika Kiyo 1998;44:319– 321. 9. Ulubaş B, Ozcan C, Acka G, Aydn O, Saritaş E. Clivus metastasis of squamous cell carcinoma: a rare location. J Clin Neurosci. 2005;12:97–98. 10. Escarda A, Vaquer P, Bonet L, Miralbés S, Gómez C, Obrador A. Clivus metastasis from hepatocarcinoma associated with transarterial hepatic chemoembolization. Gastroenterol Hepatol. 2006;29:401–404. 11. Malloy KA. Prostate cancer metastasis to clivus causing cranial nerve VI palsy. Optometry 2007;78:55–62. 12. Marchese-Ragona R, Maria Ferraro S, Marioni G, Staffieri C, Manara R, Restivo DA, Staffieri A. Abducent nerve paralysis: first clinical sign of clivus metastasis from tonsillar carcinoma. Acta Otolaryngol. 2008;128:713–716. 13. Kolias AG, Derham C, Mankad K, Hasegawa H, O’Kane R, Ismail A, Phillips NI. Multiple cranial neuropathy as the initial presentation of metastatic prostate adenocarcinoma: case report and review of literature. Acta Neurochir (Wien) 2010;152:1251–1255. 14. Kapoor A, Beniwal V, Beniwal S, Mathur H, Kumar HS. Isolated clival metastasis as the cause of abducens nerve palsy in a patient of breast carcinoma: a rare case report. Indian J Ophthalmol. 2015;63:354–357. 15. Loevner LA, Tobey JD, Yousem DM, Sonners AI, Hsu WC. MR imaging characteristics of cranial bone marrow in adult patients with underlying systemic disorders compared with healthy control subjects. AJNR Am J Neuroradiol 2002;23:248–254. 16. Bayramoglu A, Aydingöz U, Hayran M, Oztürk H, Cumhur M. Comparison of qualitative and quantitative analyses of agerelated changes in clivus bone marrow on MR imaging. Clin Anat. 2003;16:304–308. 17. Okada Y, Aoki S, Barkovich AJ, Nishimura K, Norman D, Kjos BO, Brasch RC. Cranial bone marrow in children: assessment of normal development with MR imaging. Radiology 1989;171:161–164. 18. Kimura F, Kim KS, Friedman H, Russell EJ, Breit R. MR imaging of the normal and abnormal clivus. AJNR Am J Neuroradiol 1990;11:1015–1021. 19. Moon WJ, Lee MH, Chung EC. Diffusion-weighted imaging with sensitivity encoding (SENSE) for cranial bone marrow metastases: comparison with T1-weighted images. Korean J Radiol. 2007;8:185–191. 20. Hanauer DA. EMERSE: the electronic medical record search engine. AMIA Annu Sump Proc. 2006:941. Halawa et al: J Neuro-Ophthalmol 2023; 43: 126-130 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited.