Differential diagnosis in a patient with a suspected paraneoplastic disorder varies, depending on whether cancer is a known diagnosis. Among patients with neurologic dysfunction and a known cancer diagnosis, the level of suspicion for a paraneoplastic disorder depends on the neurologic syndrome, tumor histology, and presence of antineuronal antibodies. For example, small cell lung carcinoma or Hodgkin lymphoma are much more often associated with paraneoplastic disorders than squamous cell lung carcinoma or non-Hodgkin lymphoma. Tumor metastases and neurotoxicity of cancer treatments are far more common than paraneoplastic disorders and should always be considered, as should metabolic derangements and CNS infection.
For patients without a previous cancer diagnosis, the level of suspicion for a paraneoplastic disorder depends on patient age, gender, risk factors (especially cigarette smoking), the neurologic syndrome, and the presence of antineuronal antibodies. Several syndromes should always raise the possibility of a paraneoplastic etiology, including Lambert-Eaton myasthenic syndrome, subacute cerebellar degeneration, severe sensory neuronopathy, limbic encephalopathy, and opsoclonus-myoclonus (Graus et al 2004; Giometto et al 2010). For each of these syndromes, a rather short list exists of possible associated neoplasms. None of the clinical syndromes, however, have an absolute association with neoplasia, and each can occur in patients without tumors.
There is no clinical neurologic syndrome that is absolutely associated with neoplasia; every neurologic "paraneoplastic syndrome" can occur in patients without a tumor, with the proportion of paraneoplastic versus nonparaneoplastic etiologies varying among individual syndromes. With increasing recognition of patients with autoantibodies such as anti-NMDAR antibodies, anti-GABAB receptor antibodies, and antibodies against the VGKC complex, CNS paraneoplastic disorders can be viewed as a subset of autoimmune encephalitides in adults and children (Hacohen et al 2013; Hoftberger et al 2013; Irani et al 2014).