Article Text

PDF

Cerebrospinal fluid study in paraneoplastic syndromes
  1. D Psimaras1,
  2. A F Carpentier2,
  3. C Rossi3
  1. 1
    AP-HP, Groupe Hospitalier Pitié Salpetrière, Service of Neurology Mazarin, UPMC, INSERM, Paris, France
  2. 2
    AP-HP, Service de Neurologie de l’Hôpital Avicenne, and UFR de Santé, Médecine et Biologie Humaine de Bobigny, Université Paris, France
  3. 3
    Department of Public Health, Laboratory of Clinical Epidemiology, Ranica, Bergamo, Italy
  1. Correspondence to Dr A Carpentier, Service de Neurologie de l’Hôpital Avicenne, 125 route de Stalingrad, 75013 Bobigny, France; antoine.carpentier{at}avc.ap-hop-paris.fr

Abstract

Objective: Paraneoplastic neurological syndromes (PNS) probably result from an immune reaction against antigens shared by the nervous system and tumour cells. To characterise CSF alterations in these syndromes, we studied a large series of paraneoplastic patients.

Methods: Using the PNS European database which includes patients diagnosed with PNS in Europe, we reviewed the clinical data of all patients included between 2000 and 2007 for which information on CSF was available. Patients were studied if they met the following inclusions criteria: (1) definite paraneoplastic disease with anti-Hu, anti-Yo, anti-CV2, anti-Ri anti-Ma/Ta and anti-Tr antibodies; (2) clinical information available; and (3) at least one CSF study.

Results: 295 patients met the inclusion criteria. Abnormal CSF (pleiocytosis and/or high protein level and/or oligoclonal bands) was found in 93% of patients. Pleiocytosis, but not hyperproteinorachia, was more frequently seen in patients in whom the CSF study was done early in the evolution. In 24 patients, oligoclonal bands were the only abnormality found in the CSF (10%). Elevated numbers of cells were found in 47% of patients before the third month compared with 28% after the third month (p<0.01). This evolution might suggest a subacute inflammation phase within the nervous system, followed by a non-inflammatory phase. The inflammation profile was similar in all antibody types, cancers or neurological syndromes of the PNS. Surprisingly, anti-Hu patients with high pleiocytosis at the time of diagnostic had a better survival in this study than those without pleiocytosis (572 days vs 365 days; p = 0.05).

Conclusion: CSF inflammation is a common finding in PNS patients and can be a helpful tool for diagnosis, especially if this analysis is done within 3 months after neurological onset.

Statistics from Altmetric.com

Paraneoplastic neurological syndromes (PNS) are characterised by inflammation of the nervous system, indolent tumour growth and an immune reaction against antigens shared by the nervous system and the tumour cells.1 Antibodies (Abs) against these antigens are sometimes identified in patient sera and are strongly associated with paraneoplastic syndromes.2 3 4 5 6 7 8

Abnormalities of the CSF have been reported in paraneoplastic patients, and probably reflect the inflammatory nature of these diseases. However, the frequency, relationship with clinical features and evolution over time of the CSF abnormalities has never been thoroughly described. We therefore studied CSF features in a large series of paraneoplastic patients with anti-Hu, anti-Yo, anti-CV2, anti-Ri anti-Ma/Ta and anti-Tr antibodies.

Patients and methods

The PNS European database includes patients diagnosed with PNS in Europe after 2000. The database records information on neurological symptoms, tumour diagnosis, treatment and outcome.

We reviewed the clinical data of all patients included in this PNS European database before November 2007, for whom CSF information was available. Patients were studied if they met the following inclusion criteria: (1) definite paraneoplastic disease with anti-Hu, anti-Yo, anti-CV2, anti-Ri, anti-Ma/Ta or anti-Tr antibodies in serum; (2) clinical information available (type of neurological syndrome, date of neurological onset); and (3) at least one CSF study with records of cell count (white cells/mm3: normal ⩽5 white cells/mm3) and protein content (mg/dl: normal 20–50 mg/dl). When available, the presence of oligoclonal bands (OCB) in CSF was collected although the paired serum OCB status was usually not stated in the medical records. In patients who had several lumbar punctures, only the first CSF study was considered for global analysis.

Neurological syndromes were classified as limbic encephalitis, encephalomyelitis, cerebellar degeneration, brainstem encephalitis (including opsoclonus) and peripheral nerve disorders (sensory neuronopathy, polyneuropathy, chronic inflammatory demyelinating polyneuropathy, Guillain–Barre syndrome, dysautonomia and motor neuron disease) following the diagnostic criteria of the PNS Euronetwork.9 Tumours were classified as small cell lung carcinoma (SCLC), non-small cell lung carcinoma (NSCLC), breast cancer, testicular cancer, gynaecological (ovary and uterus) cancer, Hodgkin diseases and other.

Categorical variables were expressed by proportions and continuous variables were described with mean, median and interquartile range. The Pearson χ2 test was used to compare qualitative variables while the Mann–Whitney–Wilcoxon rank test was used for quantitative variables. Survival analysis was performed with the Kaplan–Meier method to compare patients with normal versus those with abnormal cell counts. Differences in survival time from syndrome onset were evaluated using the log rank test. All tests were two tailed, with 0.05 as the level of significance. Data were analysed using SAS software (V.9.02, SAS System, Cary, North Carolina, USA).

Results

Of 829 patients registered in the database in November 2007, 295 met the inclusion criteria. Patient characteristics are summarised in tables 1 and 2. The most frequent antibody was anti-Hu (58%) followed by anti-Yo (20%), anti-CV2 (8%), anti-Ma/Ta (6%), anti-Ri (5%) and anti-Tr (3%). Among the 170 patients with anti-Hu Abs, nine had anti-CV2 Abs, eight had anti-Ri Abs, five also had antiamphiphysine Abs, one had anti-VGCC Abs, two had anti-Yo Abs, one had anti-Tr Abs and five had other or atypical Abs. Most patients with anti-Hu Abs had peripheral nerve disorders and SCLC (44% and 64%, respectively) (see appendices 1 and 2 online). Cerebellar degeneration was the main clinical syndrome in anti-Yo (92%) and anti-Tr (89%) patients. Paraneoplastic encephalomyelitis was almost always associated with anti-Hu antibodies (91%) and very often with SCLC tumours (72%). The frequency of brainstem encephalitis was higher in the anti-Ri group (60%) and limbic encephalitis in the anti-Ma/Ta group (58%). No tumour was found in 41 patients (14%).

Table 1

Patient characteristics

Table 2

Frequency of antibodies

Cells were studied in 290 patients and found to be superior to 5 white cells/mm3 in 112 patients (39%) (median cell level 3 white cells/mm3). Proteins were studied in 281 patients and found to be higher than 50 mg/dl in 187 patients (67%) (median protein level 68 mg/dl). The presence of OCB in the CSF was studied in 135 patients and found to be positive in 85 (63%). In 24 patients, OCB were the only abnormality found in the CSF (10%). In patients in whom pleiocytosis, proteinorachia and oligoclonal bands where all studied, completely normal CSF was found in only 7% of patients (18/260).

To describe the CSF abnormalities over time, we plotted the cell and protein values on the y axis and the delay from neurological onset to the CSF study on the x axis in patients who did not receive any treatment before the CSF study (n = 206) (fig 1). The CSF study was done within the first 3 months after neurological onset in 162 patients. Elevated numbers of white cells were found in 47% of these patients versus 28% after the third month (Pearson χ2 test, p<0.01). The median level was 5 cells/mm3 before the third month and 2 cells/mm3 after (Mann–Whitney–Wilcoxon test; p<0.01). Hyperproteinorachia was observed in 71% of patients (median 73 mg/dl) before the third month and in 61% (median 59 mg/dl) after the third month (Mann–Whitney–Wilcoxon test; p<0.01). OCB were found in 63% (42 out of 67) of patients and the same after the third month (43 out of 68). Within 3 months from neurological onset, in patients in whom pleiocytosis, proteinorachia and oligoclonal bands were all studied, completely normal CSF was found in only in 3% of patients (11.9% of patients had completely normal CSF after 3 months from neurological onset).

Figure 1

Evolution over time after neurological onset of cells and proteins in CSF.

To further study the possible decrease over time of pleiocytosis, we focused on the few patients who had repeated CSF studies without any treatment between two CSF tests. There were 15 patients with two CSF studies done before day 90 and 17 patients in whom the second CSF study was done after day 90 (see appendix 5 online). The median time points of the first CSF study were 18 and 87 days, respectively, for both groups. Pleiocytosis appeared or increased in seven of 15 patients when both CSF studies were made before the first 3 months but only in one of 17 patients when the second lumbar puncture was done after the first 3 months (p = 0.009). Similarly, when the analysis was limited to the 23 patients showing at least one abnormal CSF, the decrease in pleiocytosis was more frequent when the second CSF study was done after the first 3 months (p = 0.03)

To investigate whether oncological or immunological treatments could impact on CSF abnormalities, we looked at the 12 patients who had two CSF studies and were treated in between. The difference between cells and proteins before and after treatment was not statistically significant (median cell level: 9 cells/mm3 before treatment and 4.5 cells/mm3 after treatment; median protein level 86 mg/dl before treatment and 86 mg/dl after treatment).

We then focused on the pattern of CSF abnormalities within 3 months, depending on the antibody or cancer types. Surprisingly, no significant difference was seen in protein levels or in the cell reaction among the various antibodies or cancers although there was a trend for higher protein levels in patients with anti-Hu and anti-CV2 Abs or lung cancer (SCLC and NSLC) (see appendices 3 and 4 online).

We compared the CSF in patients with central or peripheral (190 vs 103 patients) PNS. When the CSF study was done within 3 months after neurological onset (central n = 115; peripheral n = 45), pleiocytosis was more pronounced in central PNS than in peripheral PNS (median 6 vs 3 cells/mm3; Mann–Whitney–Wilcoxon test, p = 0.14), while proteinorachia showed the opposite trend (median 68 vs 100 mg/dl; Mann–Whitney–Wilcoxon test, p<0.01). The percentage of positive OCB was the same between central and peripheral PNS (35% vs 41%).

We then wondered whether pleiocytosis and protein at the time of diagnosis could have a prognostic value in paraneoplastic patients (fig 2). To address this issue, we selected the anti-Hu patients (the other antibodies were not studied because of the low numbers of patients) who had a CSF study within 3 months from clinical onset (n = 92). Patients were divided into two groups; those with a normal cell count in the CSF (n = 56) and those with more than 5 cells/mm3 (n = 35). The clinical characteristics that are known to be of prognostic value were similar in both groups. The median Rankin score was 3 at the time of neurological onset in both groups. Paraneoplastic encephalomyelitis was as frequent in both groups (23% vs 18%, respectively). Peripheral disorders were slightly lower in patients with pleiocytosis (37% vs 47%, respectively). Median age was the same in both groups (65 vs 66 years). Surprisingly, the group with CSF pleiocytosis showed a better median survival than the group without (572 days vs 365 days; log rank test, p = 0.05). Protein levels were not correlated with survival (data not shown).

Figure 2

Survival of patients with anti-Hu antibodies who had a CSF study within 90 days after neurological onset. Patients were divided into two groups; those with a normal cell count in the CSF (black curve; n = 56) and those with more than 5 white cells/mm3 (grey curve; n = 35). The group with CSF inflammation showed a better median survival than the group without (572 days vs 365 days; p = 0.05).

Discussion

Paraneoplastic neurological syndromes (PNS) are believed to be immune mediated. Inflammation of the CSF is a common finding in that setting but no dedicated study has been published to date. Here we report the largest CSF study in PNS showing that abnormal CSF is almost always seen in these patients, especially early in the course of the disease.

Abnormal CSF was recorded in most patients (93%). This value fits with the limited CSF data published in series of paraneoplastic patients with anti-Hu,2 anti-Yo,6 anti-CV2/CRMP5,10 anti-Ta,11 anti-Tr12 or anti-Ri antibodies,13 with a frequency of abnormal CSF ranging from 59% to 86% of patients. The presence of OCB in CSF has been reported in approximately 75% of patients with paraneoplastic diseases,2 6 11 14 and their specificity against the paraneoplastic antigens has been documented.15 16 However, the unclear status of the serum OCB in the medical records of our CSF study represents a methodological limitation, as intrathecal antibody synthesis cannot be stated. Indeed, positive OCB in serum, related or not to the paraneoplastic disease, could have passively diffused into the CSF.17

In our series, pleiocytosis was more frequent in patients in whom the CSF study was done early after neurological onset. Median level of cells was 5 white cells/mm3 before the third month and 2 white cells/mm3 after the third month. The incidence of proteinorachia also decreased but stayed at abnormal levels (median level 73 mg/dl before the third month and 59 mg/dl after). This evolution fits with the acute or subacute course of the disease, with early neuronal loss during the first months followed by a non-inflammatory phase when all neurons have disappeared. However, a potential bias in the decrease over time of pleiocytosis can be advocated. One can imagine that patients with highly inflammatory diseases have a CSF study early in the evolution while those with a poorly active or atypical disease have a CSF study later on. Serial CSF analysis in a prospective study should easily exclude this hypothesis. In our study, when several CSF studies were available, increase in pleiocytosis was more frequent in patients in whom both CSF studies were done within the first 3 months compared with patients with a second CSF done after 3 months. However, the median time point of the first CSF study was different between both groups and might have introduced bias.

The inflammatory CSF, especially during the acute stage, is in line with radiological and pathological studies. Hypermetabolism in FDG-positron emission tomography scan and increased perfusion on single photon emission computed tomography have been described in paraneoplastic limbic encephalitis and paraneoplastic cerebellar degeneration,18 attributed by the authors to inflammatory changes. Neuropathological studies of paraneoplastic cerebellar degeneration performed within 1 year after the onset of neurological symptoms generally show severe loss of Purkinje cells in combination with various degrees of inflammatory, mainly lymphocytic, infiltrates, while studies performed more than 1 year after the onset showed absence of inflammatory infiltrates.19 This early inflammation suggests that early treatment might be beneficial for the clinical outcome.

Several unexpected results were found in our study. The CSF profile was similar in all antibody types, cancers and neurological features of the PNS. Although there was a trend for higher protein levels in patients with anti-Hu Abs, SCLC and peripheral disorders, this did not reach statistical significance. In addition, the CSF features were similar regardless of whether or not the patients were treated (cancer or immunomodulatory treatments) before the CSF study. This probably reflects the poor efficacy of the currently available treatments.

However, the most intriguing finding of our study is the shorter survival of anti-Hu patients with low CSF inflammation compared with patients with pleiocytosis although both groups showed similar clinical characteristics, including Rankin scores. This cannot be explained by the imbalance in delays from onset to CSF study between the groups as we selected only patients who had a lumbar puncture within 3 months of neurological onset. Obviously, due to the limited number of patients, this result needs to be confirmed in a large prospective study. The better survival could be related to slower tumour growth and/or better tumour response in patients who developed a strong immune response. This could also be related to greater sensitivity to immunological treatment in patients with a high inflammatory response, as reported in neurosarcoidosis20 or even in multiple sclerosis.21 22

In conclusion, CSF inflammation is very common in PNS patient and could be a helpful diagnostic tool, especially if this analysis is done within 3 months after neurological onset. Preliminary data show that patients with pleiocytosis at the time of diagnosis might have a better survival but this finding should be confirmed in a larger prospective study.

Acknowledgments

We thank Professor Jean Yves Delattre and Dr Guido Bertolini for helpful advice and review of the manuscript.

REFERENCES

View Abstract

Footnotes

  • PNS Euronetwork participants: JC Antoine, G Bertolini, F Blaes, R Fazio, F Graus, W Grisold, I Hart, J Honnorat, L Lorusso, S Rauer, J Rees, Z Rodi, P Sillevis-Smitt, P Stourac, P Tonali, C Vedeler, J Verschuuren, MC Vigliani, A Vincent, A Mazzeo, R Vitaliani, R Voltz, B Giometto.

  • See Editorial Commentary, p 3

  • Funding This work was supported by funding from the European Community’s Sixth Framework Programme (contract LSSM-CT-2005-518174). It reflects only the authors’ views; the Community is not liable for any use that may be made of the information contained herein.

  • Competing interests None.

  • PNS Euronetwork participants: JC Antoine, G Bertolini, F Blaes, R Fazio, F Graus, W Grisold, I Hart, J Honnorat, L Lorusso, S Rauer, J Rees, Z Rodi, P Sillevis-Smitt, P Stourac, P Tonali, C Vedeler, J Verschuuren, MC Vigliani, A Vincent, A Mazzeo, R Vitaliani, R Voltz, B Giometto.

  • Provenance and Peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles