Antibody-positive paraneoplastic neurologic syndromes

Patients and methods.

Between October 1997 and July 2002, we identified 13 consecutive patients with new or worsening neurologic symptoms, who were positive for an antineuronal antibody as marker for paraneoplastic etiology and in whom no tumor was known or a possible tumor recurrence was looked for. All patients were seen personally by one of two authors (M.S. and R.V.). Antineuronal antibodies were determined as described using immunohistochemistry as screening method and Western blot with recombinant HuD and Yo protein for confirmation.16,17⇓ The serum with anti-Tr reactivity was confirmed in the laboratory where anti-Tr was initially characterized.18 A tumor or tumor recurrence was diagnosed in 10 patients and confirmed histologically except in one patient (no. 9, see Results).

In all of these patients, we systematically performed in parallel both spiral CT scans of thorax, abdomen, and pelvis and whole body FDG-PET imaging as a part of routine search for a tumor or tumor recurrence. CT was performed after application of oral and IV contrast medium (single slice thickness: 5 mm). FDG-PET was performed after patients had been fasting for at least 6 to 8 hours with a dedicated scanner (ECAT EXACT HR+, Siemens/CTI, Knoxville, TN). The time between CT and FDG-PET was less than 4 weeks (median 6 days) in all patients. The image reading of the CT and FDG-PET was done strictly independently by experienced radiologists and nuclear medicine physicians. If available, previous examinations were taken into account for the image interpretation (this was true for 5/13 CT scans and for none of the FDG-PET scans).

For this study, the official written radiology/nuclear medicine reports were combined with proof of tumor and classified accordingly as true positive (TP), true negative (TN), false positive (FP), and false negative (FN). If the imaging procedure directly pointed out the tumor and the tumor could be verified, the imaging procedure was judged as TP. A misleading report regarding a tumor (e.g., the imaging showing a suspicious structure/signal leading to surgery, but without histologic confirmation of a tumor and with a negative clinical follow-up) was called FP, and a negative report despite the presence of a tumor was interpreted as FN. Since there is clinical agreement that a paraneoplastic origin of neurologic symptoms is unlikely once a tumor is not discovered within 4 to 5 years,1,3–5⇓⇓⇓ we called a report TN when the tumor was not detected by repeated imaging and clinical follow-up within that period.

For comparison of FDG-PET and CT results Fisher’s exact test was used. In addition, sensitivity (SE), specificity (SP), accuracy of the diagnostic procedures, and prevalence of disease (P) were calculated. The probability of disease given a positive test (PD+) was established according to Phillips et al.,19 as follows:

PD+ = (P * SE)/{(P * SE) + [(1 − P) * (1 − SP)]}

Because the PD+ is influenced by the prevalence of the disease, we additionally calculated the OR (which is not influenced by the prevalence)20 for the probability that in the PNS patients a tumor was detectable either with CT or FDG-PET (if both methods are equally successful and there is no difference in the detection rate, the OR would be 1).20

Results.

Thirteen consecutive patients (table; 9 women, 4 men, aged 59 ± 14 years) with an antibody-positive (anti-Hu: 8, anti-Yo: 4, anti-Tr: 1) paraneoplastic syndrome and the clinical question for a tumor or tumor recurrence were included.

CT was suspicious for a tumor in three patients showing a tumor mass in the lungs in two patients (5 and 9) later being diagnosed as SCLC and adjacent to the kidney in one patient (2) later diagnosed as a recurrence of an adult neuroblastoma of the adrenals. In the other 10 patients, CT did not raise suspicion for a tumor. In Patient 6, a single 1.5-cm slightly enlarged lymph node in the groin was noted. Since slightly enlarged inguinal lymph nodes are a common finding and no other enlarged lymph nodes or a tumor were detected, this finding did not meet the criteria for malignancy in routine image interpretation. In another patient (8), a lymph node of 1.6 cm in the left axilla was noted. As this node had been known for 15 years, it never aroused suspicion of malignancy.

FDG-PET imaging was suspicious for a tumor in 10 patients. Of those, a tumor was correctly diagnosed and confirmed by histology in 9 patients (see the table). In one patient (4), surgical lymph node exploration of the right axilla was performed following the suspicious FDG-PET result, however, no pathologic cells were detected, and no tumor diagnosed at follow-up over 5 years. In retrospect, the pathologic FDG-PET signal came from an inflammatory joint affection of the right shoulder. In Patient 11, 5 years after resection of an ovarian carcinoma a newly arisen cerebellar ataxia in combination with anti-Yo antibodies raised the question of a possible recurrence, but neither gynecologic examination nor CT accounted for a tumor. Only FDG-PET showed a clearly suspicious signal. In Patient 2 with a neuroblastoma recurrence, FDG-PET did not show a clear pathologic result, while CT revealed a left adrenal mass. In retrospect, all tumors that were correctly identified by FDG-PET could be localized on CT images by second look and direct side by side inspection of the images (figure), even if, initially, CT was read as negative.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Comparison of CT and ¹⁸F fluoro-2-deoxy-glucose (FDG)-PET imaging for tumor diagnosis in patients with antineuronal antibody. (A) Patient 6: Corresponding transaxial slices of contrast-enhanced CT and FDG-PET of an anti-Tr-positive patient and suspected paraneoplastic cerebellar degeneration. In the clinical examination a small inguinal indolent lymph node was palpable. The CT of the pelvis revealed a 1.5 cm slightly enlarged lymph node in this location. Thoracic and abdominal CT showed no further enlarged lymph nodes or a tumor. Since slightly enlarged inguinal lymph nodes are a common finding, this finding did not raise suspicion for malignancy. FDG-PET demonstrated a pathologically increased glucose metabolism in this inguinal lymph node, which was correctly interpreted as malignant (histologically confirmed partial infiltration by a mixed type Hodgkin’s lymphoma). No further tumor manifestation was detectable in the body trunk. (B) Patient 11: Corresponding transaxial slices of contrast-enhanced CT and FDG-PET of an anti-Yo-positive patient and suspected paraneoplastic cerebellar degeneration. In this patient a recurrent ovarian carcinoma was suspected since an ovarian carcinoma was removed by surgery in 1995. Repeated CT scans of abdomen and pelvis did not reveal a local recurrence. The mass, which was retrospectively visualized on CT in knowledge of the pathologic FDG-PET finding, initially had been interpreted as bowel (histologically confirmed recurrent ovarian carcinoma). (C) Patient 12: Corresponding transaxial slices of contrast-enhanced CT and FDG-PET of an anti-Yo-positive patient and suspected paraneoplastic cerebellar degeneration. In this patient a recurrent breast cancer was suspected since a breast cancer was removed by surgery in 1990. Mammographs and CT of thorax, abdomen, and pelvis were not suspicious of a tumor. FDG-PET demonstrated multiple areas with pathologically increased glucose metabolism in the pelvis, which were correctly interpreted as malignant (histologically confirmed ovarian carcinoma with peritoneal spreading). The corresponding masses, which could be retrospectively visualized on CT in knowledge of the pathologic FDG-PET findings, initially had been interpreted as bowel.

A new tumor or tumor recurrence was diagnosed in 10/13 patients (prevalence = 77%: 5 SCLC, 2 ovarian cancer, and 1 each neuroblastoma, Hodgkin’s lymphoma, and lymph node metastasis of adenocarcinoma). All tumors were confirmed histologically except in one patient (9). This anti-Hu positive patient demonstrated a massive tumor of the lung on CT and FDG-PET, and was neurologically too ill for biopsy and died shortly after tumor diagnosis; autopsy was not performed. However, as NSE was massively elevated, a SCLC was the most likely diagnosis.

In a long-term follow-up in 3/13 patients (3, 4, 10) no tumor or tumor recurrence was found. In these three patients, an initial worsening of the neurologic symptoms was the reason for the tumor search, and all three patients have remained clinically stable over the subsequent follow-up. In two patients (3, 10), a tumor (ovarian carcinoma) was known before neurologic symptoms started, but no recurrence was identified in a follow-up over more than 4 years (Patient 3: last clinical follow-up 10/03, last FDG-PET 10/01; Patient 10: last clinical and PET follow-up 11/03). In Patient 3 with anti-Hu antibodies associated with—for these antibodies atypical—ovarian cancer, no other tumor was detected. In the third patient (4, anti-Hu positive), no tumor has been identified over a follow-up of 5 years after beginning of the neurologic symptoms. An axillary lymph node dissection following a suspicious FDG-PET showed normal lymphatic tissue. The annual follow-up FDG-PET scans (last PET scan in 9/02) did not reveal pathologic findings. Neurologically, the symptoms have remained stable (last follow-up 1/04). In the early course of the disease, however, the patient had lost 15 kg of weight, which she regained spontaneously; a spontaneous regression of an undetected tumor as described earlier seems possible.21,22⇓ Furthermore, there is clinical agreement that a paraneoplastic origin of neurologic symptoms is unlikely once a tumor or tumor recurrence is not discovered within 4 to 5 years.1,3–5⇓⇓⇓ Therefore, Patients 3 and 10 were judged as TN for both CT and FDG-PET, and Patient 4 was judged as TN for CT and as FP for FDG-PET.

Sensitivity of CT was 30%, specificity 100%, and accuracy 46%. The probability of tumor given a positive result in CT (PD+) was 23%. For FDG-PET, sensitivity was 90% (p < 0.01 vs CT), specificity 67%, accuracy 85%, and PD+ 77%. The OR was 7.5 ([9/4]/[3/10]) comparing FDG-PET and CT. If CT and FDG-PET results were analyzed in combination (positive result if both or one or the other was positive), there was only one false (positive) result in Patient 4 where FDG-PET initially was misleading. In all other 12 patients, the combination of CT and FDG-PET was correct in diagnosing or excluding a tumor or tumor recurrence. Therefore, the sensitivity of the combination CT/FDG-PET was 100%, specificity 67%, accuracy 92%, and the probability of tumor given a positive result in CT/FDG-PET was 84%.

Discussion.

In this comparison between CT and FDG-PET for diagnosis of tumor or tumor recurrence in patients with a variety of antineuronal antibody associated paraneoplastic syndromes, we demonstrate a significantly higher sensitivity and accuracy by FDG-PET compared to CT. The probability of tumors given a positive result (PD+) was threefold higher in FDG-PET vs CT (77% vs 23%; OR = 7.5). But the most reliable results were obtained if both tests were combined (sensitivity = 100%, PD+ = 84%).

Recently, FDG-PET scanning has been advocated for tumor diagnosis in patients with anti-Hu positive paraneoplastic syndromes.13 We expand this experience to patients with other antineuronal antibodies including anti-Yo and anti-Tr. In our series, in patients with these antibodies, FDG-PET was more helpful than CT in 4/4 patients. Here, in all four, the tumor or tumor recurrence was only detected by FDG-PET, and the corresponding CT scans were read as nonsuspicious. In contrast, in anti-Hu patients, this was only the case in half (3/6) the patients. In the other three patients, CT and FDG-PET were equally useful in 2/3, and in one patient with a neuroblastoma recurrence CT was even better than FDG-PET. This experience may be explained by the different tumors associated with different antineuronal antibodies.

It is well known that CT has only little sensitivity for diagnosing small gynecologic tumors, while many gynecologic cancers, such as breast cancer, ovarian cancer, and cervical cancer, are now recognized as good indications for imaging with FDG-PET.23,24⇓ In suspected lung cancer, CT is the procedure of choice for initial staging, because of its high sensitivity for peripherally localized lesions.25,26⇓ However, the FP rate of CT lung scanning may be very high.26–28⇓⇓ In a series of 1,035 current or former smokers a prevalence of 29% of pulmonary nodules was reported in the screening period of 2 years using single-slice spiral CT.26 Another screening trial using the more sensitive multi-slice CT identified uncalcified pulmonary nodules in 1,049 of 1,520 patients at risk for lung cancer (69%) in a period of 3 years.27 However, in both studies 93% and 98% of these nodules were benign, and, therefore, FP for cancer.26,27⇓ Consequently, in patients with risk factors such as heavy smoking, the combination with FDG-PET starts to get advocated, in order to increase specificity of CT.26

In centrally, mediastinally localized lung tumors the sensitivity of CT is substantially lower than in the periphery.25 The only CT finding suggestive of mediastinal lymph node involvement in SCLC is nodal size and deformation. In general, the smaller the nodal diameter used to distinguish normal and abnormal nodes, the higher the sensitivity and the lower the specificity of CT. When using 1 cm as the upper limit of normal, sensitivity and specificity of CT is about 60%.25 However, in anti-Hu associated PNS metastases of SCLC are often limited to the mediastinal lymph nodes.11–13⇓⇓ For instance, in a study on PNS three anti-Hu positive patients were reported with initial negative workup using radiologic methods, but an abnormal FDG-uptake in the mediastinum.13 In our series 4/5 patients (80%) with SCLC presented with a centrally/mediastinally localized cancer. In these patients FDG-PET was positive in 100%, while sensitivity of CT scanning was 50%. The low sensitivity of CT scanning for lung cancer in our series is comparable with that of others reported for mediastinally localized tumors.14,25⇓

In retrospect, all tumors which were correctly identified by FDG-PET could later be localized on CT images by second look, even if read as negative initially. Only in one, CT was positive whereas FDG-PET was FN (Patient 2). FN results are rare with FDG-PET and may correspond to tumors with slow growth or a diameter clearly less than 1 cm.13,14,26⇓⇓ Vice versa, infectious or inflammatory lesions in the lung, in the bowel, or in joints ought to be considered within the differential diagnosis of “hot spots” in FDG-PET, but may cause FP results when searching a malignant tumor (e.g., Patient 4).13,24,26⇓⇓ A combination of FDG-PET and CT may contribute to accurately distinguishing a true tumor or recurrence from benign lesions or physiologic or inflammatory uptake and helps localizing the tumor for further management of the patient such as surgery or mediastinoscopy.15,24⇓

In a recent series in patients with clinically suspected paraneoplastic syndromes 16/43 (37%) patients showed a pathologic hypermetabolism on FDG-PET, and a tumor was confirmed in only 7 of those 16 (44%).14 In contrast to this, a pathologic FDG-PET was found at higher rate of 10/13 (77%) in our series (10 tumors confirmed). The higher rate of pathologic FDG-PET results is explained by the well defined group of patients in our series harboring an antineuronal antibody in their serum. In contrast to patients with only clinically suspected PNS without known antibody status, in patients with such an antineuronal antibody, the likelihood of a tumor is close to 100%.1,3,4⇓⇓

Retrospective studies have obvious limitations. For instance, it is possible that patients with suspected PNS in whom no tumor could be identified within a follow-up of 4 to 5 years may still develop a tumor or tumor recurrence in the future. In the present study the number of patients is small and there is heterogeneity of the underlying tumors, but the reported cases present the typical diagnostic problems that occur. These limitations emphasize the need for a properly designed long-term prospective study to address the difficult problem of finding the associated tumor in a patient with antibody-positive PNS. Based on our study, a practical but preliminary recommendation would be to get a CT scan first, and if negative or equivocal for malignancy, then a FDG-PET scan, since PET procedure is not routinely available in many hospitals. Ideally, one should have both to verify CT results by PET and PET results by CT.