 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
To the Editor.—
Although there exists considerable interest in the development of new recombinant and peptide-antigen assays for Lyme disease (LD),1 the research community has committed few resources toward improving our ability to select patients for these tests. Only a handful of studies have explored clinical prediction rules, focusing on Lyme arthritis, facial palsy, and post–Lyme syndrome.2–5 An article by Avery et al6 in the January 2006 issue of Pediatrics describes the first clinical predictive model for Lyme meningitis in an endemic area. Prolonged headache, cranial neuropathy, and mononuclear cerebrospinal fluid (CSF) pleocytosis each contributed significantly to their quantitative model. Patients with Lyme meningitis presented a mean of 7.5 days after disease onset, compared with 2.8 days for patients with aseptic meningitis. The authors qualify their findings by noting the retrospective nature of their study, and they propose a prospective validation study before widespread adoption of their model; if confirmed, they suggest that antibiotics might be withheld from patients with low scores, pending additional serology and cultures. The authors also point out that their model may not be valid in communities with a low prevalence of LD.
In addition to the study limitations identified by the authors themselves, there are additional methodologic problems that make a prospective validation study premature. These problems include elements of the study design, choice of diagnostic tests, case definition, and data management.
The Centers for Disease Control and Prevention (CDC) case-surveillance definition used in this study states that in an LD-endemic area, erythema migrans (EM) is considered pathognomonic and that additional diagnostic tests are unnecessary; for all other disease manifestations, specific clinical criteria must be accompanied by laboratory evidence of Borrelia burgdorferi infection.7 Approximately 67% of the authors' patients with Lyme meningitis demonstrated EM.6 The proposed predictive model was standardized by using meningitis case subjects with and without EM. For meningitis case subjects with EM, such a model was unnecessary because the diagnosis was already established. For meningitis case subjects without EM, we do not know if the model applied equally to them. From a design perspective, it is unclear why it was necessary to include both groups to develop a predictive model rather than just meningitis case subjects without EM.
The most difficult problem with identifying patients with Lyme meningitis without EM is the need to rely on laboratory criteria for diagnosis.7,8 It is well known that patients with LD may remain positive for serum immunoglobulin G and immunoglobulin M antibodies to B burgdorferi for many years after the initial infection,9,10 reducing the specificity of those findings in an LD-endemic community. In the setting of acute neurologic disease, serum antibody to B burgdorferi may assist with clinical decisions; however, positive results may not prove a causal relationship.10 Although not standardized, the most widely accepted laboratory method for diagnosis of acute neuroborreliosis is intrathecal antibody production to B burgdorferi8,10–19; studies have shown variable sensitivity (33%–92%) but good specificity (93%–100%) for acute neuroborreliosis. False-positive assays are most commonly caused by neurosyphilis or previous Lyme meningitis.17 Additional laboratory diagnostic methods for Lyme meningitis include CSF culture and polymerase chain reaction for B burgdorferi DNA in CSF, although both tests are relatively insensitive.20 Intrathecal antibody production, acute and convalescent serology, or any of the other studies mentioned above were not systematically collected during the current investigation, possibly leading to patient misclassification.
The authors' model uses the same predictors (ie, cranial neuropathy and lymphocytic meningitis) that are part of the CDC case definition7; this is an example of incorporation bias (ie, circular reasoning).21 Because of the absence of a gold-standard diagnostic test,8 the diagnosis of Lyme neuroborreliosis has been traditionally clinical, and the CDC case definition reflects this approach. However, a largely clinical case definition prevents one from using the same criteria for clinical prediction. Standardization of a clinical prediction model will require an independent means of verifying cases.
Two problems with data management may have skewed the group labeled as aseptic meningitis. It is uncertain whether the authors included some patients with bacterial meningitis in the aseptic group; bacterial meningitis partially treated with oral outpatient antibiotic therapy might be confused with viral meningitis and, as such, should have been excluded from analysis. Mixing bacterial with aseptic meningitis may have artificially decreased the average number of lymphocytes in CSF. Finally, there was a group of 10 patients who were excluded from the aseptic group because they were treated for Lyme meningitis despite negative serology. If those excluded from the aseptic group had clinical manifestations similar to those who met the case definition, then the authors' predictive model may have been artificially strengthened by their exclusion.
Although the goal of a clinical prediction model for Lyme meningitis remains worthwhile, the means to develop that model require further refinement. On the basis of the considerations discussed above, a prospective validation study of the authors' predictive model should be deferred.
REFERENCES
This article has been cited by other articles:
 |
|
 |
|
  R. A. Avery, G. Frank, and S. C. Eppes Predictive Model for Lyme Meningitis: A Reply Pediatrics, January 1, 2007; 119(1): 219a - 220. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
