The use of polymerase chain reaction (PCR)–based testing of cerebrospinal fluid specimens is becoming the standard in diagnosing central nervous system (CNS) infections. This technology has excellent sensitivity and specificity and allows for a more rapid diagnosis than conventional culture techniques. However, PCR testing is expensive. For example, at the Mayo Clinic, PCR testing on cerebrospinal fluid (CSF) for seven major pathogens (herpes simplex virus [HSV], cytomegalovirus, Epstein-Barr virus, varicella-zoster virus, JC virus, Borrelia burgdorferi and Tropheryma whippelii) costs more than $1,200. It has been known for many years that CSF leukocyte counts and total protein levels, although nonspecific, are quite predictive of CNS infection. Tang and colleagues performed a retrospective analysis of the use of PCR testing at their institution.
Specimen requests submitted during a three-year period for PCR testing were reviewed. Specific clinical information about the patients from whom the CSF specimens were obtained was not available for analysis. Data on the CSF protein and leukocyte counts were available. An abnormal protein result was considered to be greater than 45 mg per dL (0.45 g per L), and an elevated leukocyte count was defined as greater than 5 nucleated cells per mm3 (5 × 106 per L).
During the three years included in the study, 16,125 CSF specimens were processed by the laboratory for PCR testing. Of these, 974 were from the researchers' home institution. Of this subset, only 24 (2.5 percent) were positive for one pathogen, and two (0.2 percent) were positive for two pathogens. There were 458 specimens submitted for two or more tests. HSV and B. burgdorferi were the pathogens most commonly requested for multispecimen testing. Multiple testing requests increased during each year of the study. It was observed, however, that the greatest number of tests was ordered during the third quarter of each year that correlated with the arrival of new house staff.
Looking specifically at CSF submitted for herpes viruses (HSV, cytomegalovirus, Epstein-Barr virus, and varicella-zoster virus), 24 of 523 specimens (4.6 percent) were positive when the protein or leukocyte count was abnormal. Of the 209 patients with normal CSF protein and leukocyte counts, none of the PCR results was positive for the herpes viruses. Overall, the greatest diagnostic yield occurred when protein and leukocyte counts were abnormal. None of the 471 CSF specimens for B. burgdorferi was positive, although 68 percent of these specimens did have abnormal protein levels or leukocyte counts. However, serologic tests were performed in only 30 percent of patients whose CSF was submitted for Lyme disease testing.
The authors conclude that PCR testing should not be performed if the CSF protein and leukocyte levels are normal. This criterion would significantly decrease costs without sacrificing test sensitivity. The PCR assay found to be most helpful and most likely to affect outcome is the one for HSV. HSV was the most common pathogen identified in this study. The authors also note that PCR testing is not a good screening tool for Lyme neuroborreliosis, even when the CSF is abnormal. In a patient with suspected Lyme disease, serologic testing should be performed first. Objective neurologic deficits generally occur later in the disease (weeks 3 to 4), allowing time for seroconverison. Early testing of CSF in seronegative patients may occasionally be appropriate if neurologic findings are present and a history compatible with tick exposure is obtained.