Abstract
INTRODUCTION
CSF cytology and neuraxis MRI scanning are complementary examinations for diagnosing NM. Positive cytology is the current gold standard despite relatively low sensitivity. Although MRI is an accepted alternative diagnostic tool for both treatment and clinical trial inclusion, its sensitivity and specificity are poorly understood.
METHODS
We conducted an exhaustive, PRISMA-compliant systematic review of the literature from January 1990 to September 2016 comparing MRI to CSF cytology (the gold standard) for the diagnosis of NM. Relevant studies were assigned a level of evidence using AAN criteria. Pre-specified data was extracted, and summary statistics were calculated using the inverse variance method and random effects model. Survival of patients diagnosed with various combinations of CSF cytology and MRI was also extracted and summary statistics were calculated.
RESULTS
Thirteen studies (2 class-I, 5 class-II, 5 class-III, 1 class-IV, 1646 solid and hematologic tumor patients) provided data sufficient to calculate the diagnostic characteristics of MRI. Sensitivity and specificity were low (63.4% [56.5–71.2] and 40.1% [30.2–53.1] respectively). False positive and false negative rates were high (59.9% [46.9–69.8] and 36.6% [28.8–43.5] respectively). Various sensitivity analyses (recent studies, class I/II studies, only class I studies, specific tumor histologies) yielded very similar results. Overall survival in patients with MRI (+)/CSF (+), MRI (+)/CSF (-), and MRI (-)/CSF (+) disease (3 studies, 347 patients) were 50.8 [39.8–64.8], 115.4 [70.4–189.4] and 228.7 [130.5–400.7] days respectively (p < 0.001).
CONCLUSIONS
MRI scanning can diagnose leptomeningeal metastases, but is a very poor surrogate for CSF cytology. Patients with NM diagnosed by various combinations CSF cytology and MRI positivity have dramatically different survivals, suggesting that these categories define subsets of patients with NM with different natural histories. This finding has important implications for patient care and clinical trial design and interpretation. Novel diagnostic strategies for NM should be subjected to similar analysis.