Positron Emission Tomography–Computed Tomography in Paraneoplastic Neurologic Disorders: Systematic Analysis and Review

Objective  To evaluate the cancer detection rate of whole-body positron emission tomography–computed tomography (PET-CT) in a paraneoplastic neurologic context.

Design  Retrospective medical record review.

Setting  Mayo Clinic, Rochester, Minnesota.

Patients  Fifty-six consecutive patients with clinically suspected paraneoplastic neurologic disorders who underwent PET-CT after negative standard evaluations, including CT.

Main Outcome Measure  Rate of cancer detection.

Results  Abnormalities suggestive of cancer were detected using PET-CT in 22 patients (39%); 10 patients (18%) had cancer confirmed histologically. Cancers detected (limited stage in 9 of 10 patients and extratruncal in 4) were as follows: 2 thyroid papillary cell carcinomas, 3 solitary lymph nodes with unknown primary (2 adenocarcinomas and 1 small cell carcinoma), 1 tonsil squamous cell carcinoma, 3 lung carcinomas (1 adenocarcinoma, 1 small cell, and 1 squamous cell), and 1 colon adenocarcinoma. Detection of a well-characterized neuronal nuclear or cytoplasmic paraneoplastic autoantibody was associated with a successful PET-CT–directed cancer search (P < .001). Detection of limited-stage cancer facilitated early initiation of oncologic treatments and immunotherapy; cancer remission was reported in 7 patients, and sustained improvements in neurologic symptoms were reported in 5 (median follow-up, 11 months; range, 2-48 months). Combined data from 2 previous studies using conventional PET alone (123 patients) revealed that 28% of patients had a PET abnormality suggestive of cancer and that 12% had a cancer diagnosis.

Conclusion  In a paraneoplastic neurologic context, PET-CT improves the detection of cancers when other screening test results are negative, particularly in the setting of seropositivity for a neuronal nuclear or cytoplasmic autoantibody marker of cancer.Published online January 11, 2010 (doi:10.1001/archneurol.2009.336).

In the assessment of patients with suspected paraneoplastic neurologic disorders, routine noninvasive oncologic evaluations may be unrevealing. These standard evaluations include physical examination; computed tomography (CT) of the chest, abdomen, and pelvis; mammography in women; and testicular ultrasonography and prostate-specific antigen testing in men. Cancers detected in a paraneoplastic context are frequently small and restricted to 1 anatomical site, and they may be identified only at autopsy.¹ The detection of well-characterized neuronal nuclear or cytoplasmic paraneoplastic autoantibodies by means of specialized serologic testing refines this search and permits the detection of cancer in 68% of seropositive patients.² Commonly recognized examples include collapsin response-mediator protein 5 (CRMP-5) IgG related to small cell lung carcinoma and thymoma,³ antineuronal nuclear autoantibody type 1 (ANNA-1, anti-Hu) related to small cell lung carcinoma,^4,5 antiglial/neuronal nuclear autoantibody type 1 related to small cell lung carcinoma,⁶Purkinje cell cytoplasmic autoantibody type 1 (PCA-1, anti-Yo) related to ovarian or breast carcinoma in women,⁷ PCA-2 related to small cell lung carcinoma,⁸ and amphiphysin autoantibody related to small cell lung carcinoma and breast carcinoma in women.^9,10 Both ANNA-2 (anti-Ri)¹¹ and ANNA-3¹² (both are related to small cell lung carcinoma; ANNA-2 is also related to breast carcinoma) and PCA-Tr (related to Hodgkin disease)¹³ are less common. In the setting of seropositivity for one of those antibodies, routine evaluations for cancer may be unrevealing, leading to consideration of exploratory endoscopy or surgery.

Magnetic resonance imaging (MRI) is useful for evaluating the extent (staging) of a known cancer, including breast¹⁴ and rectal¹⁵ adenocarcinomas, and for distinguishing thymoma from thymic carcinoma.¹⁶ In patients with autoantibodies predictive of small cell lung carcinoma, MRI may be useful where CT findings have been equivocal, but MRI is generally not used as a primary oncologic screening tool.

Positron emission tomography (PET) of the whole body (usually orbits to thighs) seems to be useful in localizing otherwise occult cancers. The use of PET permits the detection of radiolabeled fludeoxyglucose (FDG) preferentially taken up by highly metabolically active cancers.¹⁷ In selected patients known to be seropositive for a paraneoplastic autoantibody, PET has been reported to have greater sensitivity for cancer than does CT.^18,19 A sensitivity of 90% for PET was reported in 10 patients with ANNA-1, PCA-1, or PCA-Tr, whereas CT had a sensitivity of just 30%.¹⁹ In that series, sequential CT and PET had a sensitivity of 100% for cancer.¹⁹ Also, PET helps direct the cancer search in unselected patients with suspected paraneoplastic neurologic disorders where CT has been unrevealing.^20,21The accuracy of PET can be further enhanced by co-registering metabolic PET abnormalities with anatomical CT abnormalities (PET-CT).^22,23 To our knowledge, PET-CT has not been systematically evaluated in patients with suspected paraneoplastic neurologic disorders.

Factors that might predict the detection of cancer by means of PET in a paraneoplastic context have not been determined. Herein, we describe the neurologic, radiologic, laboratory, and oncologic findings of Mayo Clinic patients in whom a paraneoplastic neurologic disorder was suspected by a staff neurologist and who underwent whole-body PET-CT to detect a primary cancer because standard oncologic evaluations were unrevealing (2005-2008). We also present a review of the literature pertaining to the use of PET in the oncologic evaluation of patients with paraneoplastic neurologic disorders.

This study was approved by the institutional review board of the Mayo Clinic, Rochester, Minnesota. We identified patients with a neurologic syndrome, suspected to be paraneoplastic, who had undergone comprehensive but unrevealing evaluations for cancer before undergoing PET-CT by using the diagnostic term paraneoplastic to search retrospectively in the Mayo Clinic Rochester medical records linkage system (January 1, 2005, to December 31, 2008). By cross-referencing this list with the Mayo Clinic Department of Radiology's nuclear medicine imaging database, we identified 112 patients for whom whole-body PET-CT was requested to search for cancer. We eliminated from consideration 56 patients who underwent PET-CT for cancer staging. For the included 56 patients, we abstracted from the Mayo Clinic medical record demographic, clinical (neurologic and oncologic), laboratory (serologic and spinal fluid analysis), and radiologic data. All included patients had a neurologic syndrome suspected to have a paraneoplastic etiology by a staff neurologist. Reasons for suspicion of a paraneoplastic neurologic disorder by the physician ordering the PET-CT included 1 or more of the following: neurologic symptom onset was acute or subacute, a history of cancer, a history of smoking, a paraneoplastic antibody detected in serum or cerebrospinal fluid (CSF), and inflammatory spinal fluid.

Sera for all 56 patients and CSF for 28 patients were evaluated in the Mayo Clinic Neuroimmunology Laboratory using standardized immunofluorescence criteria for IgG neuronal nuclear and cytoplasmic IgG markers of paraneoplastic neurologic autoimmunity (ANNA-1, ANNA-2, ANNA-3, amphiphysin antibody, PCA-1, PCA-2, PCA-Tr, CRMP-5 IgG, and antiglial/neuronal nuclear autoantibody type 1). Rarer neural-specific antibodies identified by means of immunofluorescence but as yet unclassified for antigen specificity or cancer association²⁴ are also included in these results. Patient sera were tested additionally by using radioimmunoprecipitation assays for neuronal voltage-gated cation channel antibodies (calcium channel [P/Q-type and N-type] and potassium channel [VGKC]), muscle and neuronal ganglionic (α3) nicotinic acetylcholine receptor (AChR) antibodies, and glutamic acid decarboxylase 65-isoform antibodies; enzyme-linked immunosorbent assay for striational antibodies; and recombinant Western blot for CRMP-5 IgG.

Associations between PET-CT–directed cancer diagnosis and potential predictors, such as age, sex, smoking status, history of cancer, neurologic presentation, and CSF and serologic findings, were assessed using the Fisher exact test for categorical variables and the Wilcoxon rank sum test for continuous variables owing to smaller sample sizes (JMP 7.0 software; SAS Institute Inc, Cary, North Carolina). All the tests were 2-sided, and P < .05 was considered statistically significant.

We searched PubMed using the terms positron emission tomography and paraneoplastic. We included all studies (excluding single case reports) that assessed PET in the oncologic evaluation of suspected paraneoplastic neurologic disorders in which diagnostic information was documented for all patients with PET abnormalities (eg, benign pathologic findings, cancer, no biopsy performed).

Of 56 included patients, 50% were men. Median age at neurologic symptom onset was 61 years (age range, 22-80 years). Neurologic manifestations were multifocal in 21 patients (38%). The neuraxis levels affected, in descending frequency, were as follows: cerebral cortex, 36%; cerebellum, 33%; peripheral nerve, 25%; spinal cord, 22%; brainstem, 18%; nerve root, 14%; basal ganglia, 11%; autonomic nervous system, 7%; cranial nerve, 6%; anterior horn cell, 5%; and muscle, 5%.

Twenty-two patients (39%) were current smokers. Ten patients (18%) had a history of cancer (3 breast, 2 testicular, and 1 each uterine, myeloma, lymphoma, lung, and prostate) that preceded the neurologic presentation by a median of 9 years (range, 2-33 years), and all were in cancer remission clinically and radiologically at neurologic presentation. Ultimately, PET-CT revealed recurrence of a known cancer in 2 of these patients, 1 with prostatic adenocarcinoma and 1 with squamous cell carcinoma of the lung.

Antibodies with known paraneoplastic significance were detected in the serum or CSF of 39 patients (70%). One or more neural autoantibodies were detected in 38 of 56 patients (68%) using the standard Mayo Clinic serologic evaluation. In addition, Ma1 antibody was identified serologically in 1 patient tested (Athena Diagnostics, Worcester, Massachusetts). The neuronal nuclear or cytoplasmic paraneoplastic autoantibodies detected by means of immunofluorescence in 13 patients (23%) were ANNA-1, 6 patients; PCA-1, 1; PCA-2, 1; CRMP-5 IgG, 1; amphiphysin antibody, 1; and unclassified neuronal nuclear or cytoplasmic antibodies, 3. All except the unclassified antibodies were confirmed by means of Western blot analysis. Seven of these 13 antibodies were identified in serum only, 5 were identified in serum and CSF (ANNA-1, 4; and CRMP-5 IgG, 1), and 1 was identified in CSF only (PCA-1). Seven of these 13 seropositive patients (54%) were ultimately found to have cancer via PET-CT.

One or more additional neuronal or muscle autoantibodies were detected in 22 patients (39%), in descending frequency: α3 AChR, 8; voltage-gated calcium channel, 7 (P/Q-type, 4; and N-type, 3); striational, 6; VGKC, 2; muscle AChR, 1; and Ma1, 1. Also, CRMP-5 IgG was identified using Western blot alone in 7 patients in which findings from immunofluorescence were negative. Three of these 26 patients (12%) were ultimately found to have cancer via PET-CT; both patients seropositive for VGKC antibody had a cancer detected.

Results of standard CSF analyses were available for 43 patients. The CSF was determined to have 1 or more inflammatory markers (elevated leukocyte count, supernumerary oligoclonal bands, and an elevated IgG index, consistent with a paraneoplastic neurologic disorder) in 16 patients (37%).

Before PET-CT, included patients underwent a median of 3 (range, 1-6) radiologic or endoscopic investigations at Mayo Clinic to look for a primary site of cancer. Investigations included chest CT (54 patients; 52 had negative and 2 had indeterminate findings), abdomen and pelvis CT (49 patients), mammography (15 patients), upper gastrointestinal endoscopy (14 patients), colonoscopy or colonography (14 patients), abdominal ultrasound (6 patients), neck ultrasonography (3 patients), and transvaginal ultrasonography of the pelvis (2 patients).

Whole-body PET-CT was suggestive of cancer in 22 patients (39%) (Table 1); 20 of those patients underwent targeted evaluations (tissue biopsy in 19 and laryngoscopy in 1).

Cancer was confirmed histologically in 10 patients (18%), representing half of all patients with a PET finding suggestive of cancer (Figure). All 10 patients were also seropositive for a paraneoplastic autoantibody (Table 2): 7 had neuronal nuclear or cytoplasmic antibodies identified by means of immunofluorescence and 3 had ion channel antibodies (VGKC, 2; and α3 ganglionic AChR, 1).

Nine of the 10 identified cancers were limited in stage (Table 2). These cancers included 3 lung carcinomas (1 adenocarcinoma, 1 small cell carcinoma, and 1 squamous cell carcinoma), 3 lymph node carcinoma metastases (2 adenocarcinomas and 1 small cell carcinoma), 2 thyroid carcinomas (both papillary), 1 colon adenocarcinoma, and 1 palatine tonsil squamous cell carcinoma. Of the 3 patients in whom carcinoma was found in lymph nodes, 1 had metastatic prostatic adenocarcinoma, 1 had breast adenocarcinoma discovered after axillary lymph node biopsy, and 1 had small cell carcinoma with no primary identified (despite surveillance for 4 more years). Patients 5 and 7 had indeterminate pulmonary abnormalities that were suggestive of cancer (>1 cm in size) detected by means of CT. These abnormalities were highly suggestive of cancer using the PET-CT criteria and were subsequently confirmed histologically to be carcinoma.

No cancer was detected in 12 patients (55%) who had PET-CT abnormalities suggestive of cancer (Table 1). Two had premalignant lesions: a Hurthle cell adenoma of the thyroid (patient 1) and a tubulovillous adenoma of the colon (patient 11). Biopsy of the paratracheal node revealed noncaseating granulomas suspected to be sarcoidosis (patient 19). Biopsy results were negative in 5 patients, and other PET-directed evaluations without biopsy were negative in 2. No further evaluations were undertaken in 2 patients, 1 of whom died soon after PET-CT.

Of the 10 patients who had a histologically confirmed cancer, 7 received cancer-directed therapy and immunotherapy, 2 received cancer-directed therapy alone, and 1 received only immunotherapy (Table 2). Seven of these 10 patients had remission from cancer in the posttreatment surveillance period (median, 11 months; range, 2-48 months). Improvements in neurologic symptoms were reported by their physicians for 5 patients (attributed to combined cancer-directed therapy and immunotherapy in 4 patients and to cancer-directed therapy alone in 1). Sustained stabilization of neurologic symptoms after immunotherapy was documented in another 3 patients.

The median time from neurologic symptom onset to PET-CT was not different for patients in whom cancer was found (8.5 months; range, 3-12 months) and in whom cancer was not found (9 months; range, 1-168 months). The association of a neuronal nuclear or cytoplasmic paraneoplastic autoantibody detected by means of immunofluorescence and a successful cancer search using PET-CT was significant (P < .001); 7 of 10 patients with cancer (70%) and 6 of 44 patients without cancer (14%) were seropositive. No other statistically significant associations with cancer detection by PET-CT were identified.

We identified 4 other studies^18-21 that evaluated the utility of PET for cancer diagnosis in patients with paraneoplastic neurologic disorders who met the inclusion criteria of the present study (Table 3). Detection of a paraneoplastic autoantibody was an inclusion criterion in 2 of the studies,^18,19 and clinical suspicion of a paraneoplastic neurologic disorder was an inclusion criterion in 2 studies.^20,21 The design of both of the latter studies was similar to that of the present study. The combined data for 123 patients from those 2 studies^20,21 revealed a PET abnormality suggestive of cancer in 34 patients (28%), and a cancer diagnosis was confirmed in 15 of these patients (12%). Of 16 patients in whom 1 or more paraneoplastic antibodies were found, 6 (38%) had a PET-directed cancer diagnosis.^20,21 The present study is the only study for which PET-CT data are available.

The use of PET-CT increased the diagnostic yield for cancer by 18% in patients with suspected paraneoplastic neurologic disorders for whom results of standard oncologic tests were negative. Data from other reported studies^20,21 using similar methods revealed that 15 of 123 patients (12%) with a suspected paraneoplastic neurologic disorder had a PET-directed cancer diagnosis. The difference in diagnostic yield for cancer between the present study and previously reported studies may be explained by the larger number of patients available for follow-up in the present study or possibly by the enhanced sensitivity of PET-CT over PET alone.

Detection of a well-characterized neuronal nuclear or cytoplasmic antibody in paraneoplastic serologic or CSF evaluation was strongly associated with a PET-directed cancer diagnosis; 56% of seropositive patients had cancer (3 had ANNA-1; 2 had unclassified neuronal nuclear or cytoplasmic antibodies; 1 had CRMP-5 IgG, and 1 had amphiphysin antibody). Cation channel autoantibodies do not have as strong an association with cancer but are instructive in some patients. Cancer was found in both patients in whom VGKC antibodies were detected (1 thyroid papillary and 1 colon adenocarcinoma). Tan et al²⁵ recently reported a 33% frequency of cancer detection in patients with VGKC antibodies, with notable inclusion of colon and thyroid cancers. It is instructive that patient 3 had a cancer associated with ganglionic AChR antibody seropositivity (papillary carcinoma of thyroid).²⁶

The cancer screening investigation most commonly acquired before PET-CT was CT of the chest, abdomen, and pelvis. However, 4 of the 10 detected cancers using PET-CT were outside the anatomical scope of CT of the chest, abdomen, and pelvis (thyroid, 2; cervical lymph node, 1; and palatine tonsil, 1). The other 6 detected cancers were too small to be detected by appropriate regional CT (lung, 4; axillary lymph node, 1; and colon, 1). Clearly, CT alone is not sufficient to exclude cancer in cases with a high index of suspicion for cancer. Lucchinetti et al⁵ reported that initial screening with CT is unrevealing in 60% of patients with ANNA-1, but cancer is detected in 90% of those patients when evaluation is more extensive or is repeated. In the present study, PET-CT enabled the detection and precise anatomical localization of metabolic abnormalities suggesting cancer in 22 patients, for whom further evaluations led to cancer diagnoses in 10 (45%).

In patient 7, PET-CT was critical in determining the oncologic significance of a subpleural pulmonary nodule initially detected using CT. Veronesi et al²⁷ reported that PET-CT had 88% sensitivity and 93% specificity for cancer in patients with progressive pulmonary nodules exceeding 8 mm. In patient 9, who was seropositive for ANNA-1, recurrent metastatic prostatic adenocarcinoma was detected, although a coexisting occult small cell carcinoma was suspected before biopsy. A previous diagnosis of prostatic adenocarcinoma is obtained in 39% of patients with small cell carcinoma of the prostate, which is a rare entity.²⁸

Follow-up of neurologic symptoms and cancers in the 10 patients herein was short (median, 11 months). Oncologic treatment initiated by the detection of cancer at an early stage using PET-CT was associated with remission in 7 of the 10 patients. Nine of 10 patients improved or stabilized neurologically after oncologic treatment or immunotherapy. Neurologic symptoms resolved in both patients with VGKC antibodies after surgical resection of a thyroid cancer (patient 4) or a colon cancer (plus immunotherapy in patient 13).

We recognize several limitations of this study: the retrospective design, the relatively small sample size, the exclusion of patients who had negative findings on standard evaluations but did not undergo PET-CT at Mayo Clinic, and selection bias for patients with detectable neural autoantibodies. Some patients may have had PET-CT after leaving Mayo Clinic, and others may have been denied the test on financial grounds. Also, patients were systematically tested only for autoantibodies routinely available in the Mayo Clinic paraneoplastic evaluation, which does not include testing for Ma/Ta and N-methyl-D-aspartate receptor autoantibodies. Because seropositivity for a paraneoplastic autoantibody and clinical suspicion for a paraneoplastic neurologic disorder (and, hence, ordering a PET-CT) are inextricably linked, not testing for 1 or more known autoantibodies inevitably leads to ascertainment bias. The inclusion of 2 patients with indeterminate findings on chest CT, both of whom were later determined to have cancer after PET-CT and biopsy, contributed to the 18% diagnostic yield.

To our knowledge, PET-CT has not been evaluated as a primary oncologic screening modality in a paraneoplastic context. The latter approach has potential advantages and limitations. It is acknowledged that PET is more sensitive than is CT for many neoplasms, including cancers of the head and neck,²⁹ lung,³⁰ breast,³¹ pancreas,³² bile duct,³² stomach,³² colon,³³ and uterus.³⁴ However, endoscopy remains the optimal screening test for esophageal, stomach, and colon cancers³²; for pancreatic carcinomas, PET and endoscopic ultrasonography are equally sensitive.³⁵ Other imaging modalities are superior to PET for detecting primary prostatic carcinoma (MRI and transrectal ultrasonography)³⁶ and testicular cancers (CT and ultrasonography).³⁷ Alone, PET is suboptimal for the detection of bladder and kidney cancers owing to high levels of physiologic FDG uptake by those organs.³⁸ Testes and ovaries may also have physiologically elevated FDG uptake that may hinder the interpretation of PET.^34,37 The PET-CT has poor resolution for primary hepatocellular carcinomas smaller than 5 cm in diameter³⁹ because these neoplasms have low or no FDG uptake.³² Increased FDG uptake seen on PET is not specific for cancer in all cases, as observed in 10 of 20 patients further evaluated after PET-CT in this series; other causes of increased FDG uptake by tissues include premalignant lesions and inflammatory and infectious disorders.⁴⁰

Recognizing the limitations of PET-CT, we favor this modality for initial oncologic evaluation of patients in whom a paraneoplastic neurologic disorder is strongly suspected. Serum autoantibody profiles and familial risk factors for cancer help guide tests for cancers for which there is poor resolution on PET-CT (eg, pelvic ultrasound for ovarian teratomas). Elimination of whole-body imaging with CT alone before further imaging with PET-CT could reduce radiation exposure and the total financial burden of testing.

Correspondence: Sean J. Pittock, MD, Department of Neurology, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (pittock.sean@mayo.edu).

Accepted for Publication: September 14, 2009.

Published Online: January 11, 2010 (doi:10.1001/archneurol.2009.336).

Author Contributions: All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: McKeon and Pittock. Acquisition of data: McKeon, Apiwattanakul, Lachance, Lennon, Boeve, Britton, and Drubach. Analysis and interpretation of data: McKeon, Lennon, Mandrekar, Mullan, Mokri, and Pittock. Drafting of the manuscript: McKeon. Critical revision of the manuscript for important intellectual content: Apiwattanakul, Lachance, Lennon, Mandrekar, Boeve, Mullan, Mokri, Britton, Drubach, and Pittock. Statistical analysis: Mandrekar. Administrative, technical, and material support: Lennon. Study supervision: Pittock.

Financial Disclosure: Dr Lennon stands to receive royalties for commercial assays to detect aquaporin-4–specific autoantibodies. The intellectual property is licensed to a commercial entity for the development of a simple, antigen-specific assay, to be made available worldwide for patient care. The test will not be exclusive to the Mayo Clinic. Until now, Dr Lennon has received less than $10 000 in royalties. None of the authors receive royalties from the sale of antibody testing performed in the Mayo Clinical Neuroimmunology Laboratory; however, Mayo Collaborative Services Inc does receive revenue for conducting these tests.

Previous Presentation: This study was presented in part at the American Academy of Neurology Annual Meeting; April 29, 2009; Seattle, Washington.