Subjects and Methods

Eligibility criteria and proceduresSubjects were eligible for enrollment if they had syphilis, defined as reactive serum nontreponemal and treponemal serological test results, and met one of the following criteria (based on the 1993 CDC guidelines) [10]: neurological or ophthalmological symptoms or signs during any stage of syphilis, late latent syphilis or syphilis of unknown duration (particularly if the serum nontreponemal antibody titer was ⩾1:32), treatment failure, HIV infection (particularly in patients with late latent syphilis), or intent to begin nonpenicillin therapy. Subjects were enrolled at 8 sites: University of Washington, Seattle (132 subjects); Johns Hopkins University, Baltimore (53 subjects); Emory University, Atlanta (41 subjects); Washington University, St. Louis (37 subjects); State University of New York, Brooklyn (23 subjects); Cook County Hospital, Chicago (17 subjects); University of Mississippi, Jackson (15 subjects); and the Chicago Department of Health, Chicago (8 subjects). Subjects were recruited from outpatient sexually transmitted diseases, infectious diseases, ophthalmology, lumbar puncture, and general medical clinics; emergency departments; and in-patient medicine and neurology wards. The study protocol was reviewed and approved by the institutional review boards of each participating site. After written informed consent was obtained, a standardized medical history was obtained, and subjects underwent neurological examination, venipuncture, and lumbar puncture. The syphilis stage was determined according to CDC guidelines; patients with syphilis of unknown duration were included in the group with late latent syphilis [9]. Results of HIV serological tests, plasma HIV-1 RNA copy number, and peripheral blood CD4⁺ T lymphocyte count were obtained from a review of medical records. Only plasma HIV-1 RNA copy number and peripheral blood CD4⁺ T lymphocyte count obtained within 90 days of the lumbar puncture (entry date) were used. We used a lower limit of detection for HIV-1 RNA of ⩽500 copies/mL, because that was the lower limit of detection when the study began

Laboratory methodsSerum nontreponemal (VDRL or rapid plasma reagin [RPR]) and treponemal (microhemagglutination–T. pallidum fluorescent treponemal antibody–absorbed test, T. pallidum particle agglutination or T. pallidum ELISA [Captia Syphilis IgG; Trinity Biotech]) serological tests and CSF glucose, protein, WBC count, red blood cell count, and CSF VDRL tests were performed using standard methods at each of the 8 participating study sites. RPR test titers were determined for all available baseline serum samples in a central laboratory using standard methods, and these values were used in the analysis. In 7 instances, no sample was available for testing in the central laboratory, and the RPR titer obtained by the participating study site was used in the analysis. The presence of T. pallidum in CSF was determined by reverse-transcriptase (RT) PCR, according to published methods [19], with minor modifications. Specifically, the primers were shortened in March 1999 (sense, 5′-CTCTTTTGGACGTAGGTCTT; antisense, 5′-TTACGTGTTACCGCGGCT), and the procedure was optimized for use in a Lightcycler (Roche) in October 2001. For all assays, the limit of detection was 1–10 T. pallidum organisms/mL

Design and statistical analysisThe analysis in the present article is restricted to subjects who met the following criteria: (1) the baseline serum sample was collected within 2 weeks of the lumbar puncture, and (2) the subject had never received a diagnosis of or been treated for neurosyphilis before study entry. We excluded individuals with a history of neurosyphilis because of concern that CSF abnormalities, particularly reactive CSF VDRL test results, might be related to a previous episode of neurosyphilis. Because HIV infection itself may be associated with mild CSF pleocytosis [20], for the purposes of our study, neurosyphilis was defined as a reactive CSF VDRL test result or CSF WBC count >20 cells/μL, which is above the mild level of pleocytosis usually seen with HIV alone

Associations between categorical variables were assessed by the χ² test or by Fisher’s exact test. Associations between continuous variables and categorical variables were assessed by the Mann-Whitney U test. Associations between neurosyphilis and clinical or laboratory parameters were also evaluated using logistic regression to control for potential confounding variables. Two-tailed P values <.05 were considered to be significant

Results

Characteristics of study subjectsThe characteristics of the 326 subjects are shown in table 1. Most subjects were men, were infected with HIV, and had late latent syphilis. The median age was 37 years (range, 18–89 years). Two hundred five subjects had received treatment for an episode of nonneurological syphilis before study entry. Of these 205, 49 subjects were treated within 14 days of study entry. As expected, early-stage syphilis was significantly associated with a serum RPR titer ⩾1:32 (P<.001). Early-stage syphilis was also associated with HIV infection (P<.001). Specifically, only 5 (6.0%) of 84 evaluable subjects not infected with HIV had early-stage syphilis, but 93 (44.7%) of 208 evaluable subjects with HIV infection had early syphilis. This disparity was expected, because patients with early-stage syphilis are more likely to have a lumbar puncture if they are HIV infected

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 1

Characteristics of the 326 subjects with syphilis

Symptoms and signs were categorized into those most consistent with syphilitic meningitis (headache, stiff neck, photophobia, subjective hearing loss, or abnormal finger friction test result in either ear) and with syphilitic ocular disease (subjective decrease in vision, ocular inflammation, or abnormal visual acuity). Eighty-seven subjects (26.7%) met the definition of meningitis alone, 13 (4.0%) met the definition of ocular disease alone, and 25 (7.7%) met the definition for both conditions. The proportion of subjects with ocular disease was significantly greater in the HIV-uninfected group (26.4% vs. 6.0%; P<.001), but the proportion of subjects with meningitis was similar between the 2 groups. The greater proportion of HIV-uninfected subjects with ocular disease likely reflects the CDC criteria for lumbar puncture [9–11]—specifically, HIV-uninfected patients are less likely than HIV-infected individuals to undergo lumbar puncture if they do not have ocular or neurological symptoms

Neurosyphilis in HIV-uninfected subjectsFifteen HIV-uninfected subjects (16.5%) met our definition of neurosyphilis. Of these 15, 6 subjects had a reactive CSF VDRL test result only, 7 subjects had both a reactive CSF VDRL test result and CSF WBC count >20 cells/μL, and 2 had a CSF WBC count >20 cells/μL only. The median CSF WBC count in HIV-uninfected subjects with neurosyphilis was 29 cells/μL (range, 0–221 cells/μL). Neurosyphilis was significantly more likely in subjects with early-stage syphilis (primary, secondary, or early latent disease) and in those with a serum RPR titer ⩾1:32 (table 2). In a multivariate logistic regression model, syphilis stage was not significant after controlling for serum RPR titer ⩾1:32. Neurosyphilis remained significantly more common in subjects with a serum RPR titer ⩾1:32, with an odds ratio (OR) of 10.85 (95% confidence interval [CI], 2.69–43.80) (table 2). Neurosyphilis was not more common in subjects with meningitis or ocular syphilis (P=.59 and P=.75, respectively)

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 2

Clinical and laboratory features in 91 HIV-uninfected subjects with and without neurosyphilis

The above analyses were repeated using reactive CSF VDRL test result as the definition of neurosyphilis, and identical conclusions were obtained. In the univariate analysis, neurosyphilis was significantly more likely in subjects with early-stage syphilis (OR, 10.35; 95% CI, 1.54–69.77; P=.02) and in those with a serum RPR titer ⩾1:32 (OR, 10.29; 95% CI, 2.77–38.20; P<.001). Serum RPR titer ⩾1:32 was the only variable that remained significant in the multivariate analysis (OR, 10.28; 95% CI, 2.38–44.40; P=.002)

Neurosyphilis in HIV-infected subjectsFifty HIV-infected subjects (21.5%) met our definition of neurosyphilis. Of these 50, 16 subjects had a reactive CSF VDRL test result only, 12 subjects had both a reactive CSF VDRL test result and CSF WBC count >20 cells/μL, and 22 had CSF WBC count >20 cells/μL only. The median CSF WBC count in HIV-infected subjects with neurosyphilis was 29 cells/μL (range, 0–349 cells/μL). Neurosyphilis was significantly more common in subjects with early syphilis, a serum RPR titer ⩾1:32, and a peripheral blood CD4⁺ T cell count ⩽350 cells/μL or a plasma HIV-1 RNA level >500 copies/μL (table 3). We reasoned that higher plasma HIV-1 RNA levels might be associated with higher CSF WBC counts because of HIV infection itself [21], rather than infection with T. pallidum and that this could mislead our interpretation of the association between neurosyphilis and plasma HIV-1 RNA level. We did not expect a similar problem with CD4⁺ T cell count, because HIV-related CSF pleocytosis is less common with lower CD4⁺ T cell counts [21]. In keeping with our prediction, CSF WBC counts were significantly higher in subjects with plasma HIV-1 RNA levels >500 copies/mL (P<.001) but were not higher in subjects with peripheral blood CD4⁺ T cell counts ⩽350/μL (P=.43). Thus, in the multivariate analysis shown in table 3, we did not consider plasma HIV-1 RNA levels. After adjusting for syphilis stage and previous treatment for syphilis, neurosyphilis remained significantly more common in HIV-infected subjects who had a serum RPR titer ⩾1:32 (OR, 5.98; 95% CI, 2.43–14.73) or a peripheral blood CD4⁺ T cell count ⩽350 cells/μL (OR, 3.10; 95% CI, 1.40–6.86) (table 3). Thus, in HIV-infected subjects with both a serum RPR titer ⩾1:32 and a peripheral blood CD4⁺ T cell count ⩽350 cells/μL, neurosyphilis was 18.6 times more likely than in individuals with neither of these features. Neurosyphilis was not more common in subjects with meningitis (P=.21) but was more common in those with ocular syphilis (P=.01)

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 3

Clinical and laboratory features in 233 HIV-infected subjects with and without neurosyphilis

The above analyses were repeated using reactive CSF VDRL test result as the definition of neurosyphilis, with identical conclusions. In the univariate analysis, neurosyphilis was significantly more common in subjects with serum RPR titers ⩾1:32 (OR, 3.57; 95% CI, 1.46–8.78; P=.004) and in those with peripheral blood CD4⁺ T cell counts ⩽350/μL (OR, 3.45; 95% CI, 1.32–8.99, P=.008). In the multivariate analysis, neurosyphilis remained significantly more common in HIV-infected subjects with serum RPR titers ⩾1:32 (OR, 4.09; 95% CI, 1.40–11.96; P=.01) or peripheral blood CD4⁺ T cell counts ⩽350 cells/μL (OR, 3.67; 95% CI, 1.33–10.12; P=.01). Multivariate analysis was repeated for subjects with a nonreactive CSF VDRL test result, in whom CSF WBC >20 cells/μL would be the only indication of neurosyphilis. Neurosyphilis remained significantly more common in subjects with serum RPR titers ⩾1:32 (OR, 7.25; 95% CI; 1.78–29.47; P=.006), but neurosyphilis was no longer more common in those with CD4⁺ T cell counts ⩽350 cells/μL. This loss of significance is likely due to the small number of subjects who met the restricted definition of neurosyphilis

Identification of T. pallidum in CSFT. pallidum was identified by RT-PCR in CSF from 26 (8.0%) of 326 subjects (table 4). The identification of T. pallidum in CSF was significantly more likely in subjects with early syphilis and in those with a serum RPR titer ⩾1:32, a reactive CSF VDRL test result, or a CSF profile consistent with neurosyphilis. CSF WBC counts were significantly higher in subjects with a positive RT-PCR result, compared with those with a negative RT-PCR result (median, 34.5 vs. 2.0 cells/μL; P<.001). The identification of T. pallidum in CSF was not more likely in those who were infected with HIV and did not differ among the 3 treatment groups. Ten of 26 subjects with positive RT-PCR results did not meet our definition of neurosyphilis; 8 had <10 CSF WBCs/μL, and all had normal CSF protein concentrations and nonreactive CSF VDRL test result. Nine of 10 individuals were HIV infected, which was consistent with the proportion of HIV-infected subjects in the study (table 1). Among the HIV-infected subjects, the identification of T. pallidum in CSF was not significantly associated with plasma HIV-1 RNA levels ⩾500 copies/mL or peripheral blood CD4⁺ T cell counts ⩽350 cells/μL (P=.78 and P=.16, respectively)

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 4

Clinical and laboratory features of 326 subjects with and without identification of Treponema pallidum in cerebrospinal fluid (CSF) by reverse-transcription polymerase chain reaction (RT-PCR)

Discussion

The frequent invasion of the CNS by T. pallidum has been recognized for decades, and routine examination of the CSF was conducted during the prepenicillin era to guide the duration of heavy-metal therapy. Today, universal lumbar puncture is both impractical and unnecessary. However, continuing uncertainty exists regarding which patients with syphilis should undergo lumbar puncture. For this reason, we sought to identify clinical and laboratory parameters that predict neurosyphilis in both HIV-uninfected and -infected individuals. To our knowledge, our multicenter investigation is the largest modern study to have examined this question. Regardless of stage of syphilis, previous syphilis treatment, and HIV status, neurosyphilis is significantly more common when the serum RPR titer is ⩾1:32. HIV-infected individuals with both serum RPR titers ⩾1:32 and peripheral blood CD4⁺ T cell counts ⩽350 cells/μL are at an even higher risk of neurosyphilis. We did not find an association between meningitis and our laboratory-based definition of neurosyphilis in all subjects nor between ocular syphilis and our definition of neurosyphilis in HIV-uninfected subjects. These unexpected findings likely reflect imprecision inherent in the clinical definitions. Both were based on symptoms or a simple bedside test. The inclusion of specific otological or ophthalmological findings would likely have increased the diagnostic accuracy

How can our findings be best used to guide clinicians in the selection of patients with syphilis who are most likely to benefit from CSF examination? On the one hand, all individuals with high (⩾1:32) serum RPR titers or low (⩽350 cells/μL) peripheral blood CD4⁺ T cell counts could be examined by lumbar puncture at the time of their diagnosis of syphilis. Those with CSF abnormalities would then be treated with a regimen recommended for neurosyphilis [9], to prevent progression to symptomatic or more severe disease. This approach would be the most comprehensive but would result in treatment of some patients who might ultimately be able to control their CNS infections without neurosyphilis therapy. However, because it is not possible to predict in which patients CSF abnormalities will or will not resolve, it can be argued that this approach is the safest

On the other hand, an alternative approach would be to postpone CSF evaluation for 6–12 months after conventional syphilis treatment and then to use the same criteria to select those individuals who are at the highest risk for neurosyphilis. Such an approach would likely result in fewer lumbar punctures and thus decrease costs because of the decline in serum RPR titers after conventional syphilis therapy. The number of patients put at risk by the latter approach because of loss to follow-up, combined with the subsequent development of symptomatic or more severe neurosyphilis, is not known. This risk may be greatest in those with concomitant HIV infection, in whom early neurorelapse after treatment with im BPG may be more likely [12, 18]. This issue is particularly relevant at present, because the rates of syphilis in men who have sex with men, a group with a high risk of HIV infection, in the United States and Europe are dramatically increasing [22–26]

Regardless of the approach, routine laboratory CSF tests will fail to identify some patients with CNS invasion. In our study, T. pallidum was identified in 10 CSF samples that were otherwise normal, which likely reflects very recent CNS invasion [27]. The clinical implications of such a finding have been debated; experts recommend the consideration of treatment for neurosyphilis in HIV-infected but not in HIV-uninfected individuals in whom T. pallidum is identified in otherwise normal CSF [28] because of concern about HIV-associated impaired clearance of CNS T. pallidum. Consistent with the results of previous studies, we found no association between the identification of T. pallidum in CSF by RT-PCR and HIV status [6, 7]. The significant association that we observed between a peripheral blood CD4⁺ T cell count ⩽350 cells/μL and neurosyphilis argues that, once invasion has occurred, the persistence of CNS infection with the development of CSF pleocytosis and reactive CSF VDRL test result may be more likely in those with HIV-mediated immunosuppression

The limitations of our study should be acknowledged. Our results are applicable to individuals who met the criteria that we specified for lumbar puncture and who did not have a prior episode of neurosyphilis. We did not collect information on antiretroviral therapy in the HIV-infected subjects. Instead, we focused on peripheral blood CD4⁺ T cell counts and plasma HIV-1 RNA concentrations, which are markers of the efficacy of therapy. Because of the demographics of syphilis in the United States and the criteria that we used for lumbar puncture, most of our subjects were HIV infected. Nonetheless, our study population included lumbar puncture data on 91 HIV-uninfected subjects. This number is quite similar to the number of HIV-uninfected subjects in a recent and well-regarded syphilis study that also included CSF examination [7]. It could also be argued that our laboratory criteria for neurosyphilis include individuals with CSF pleocytosis caused by HIV rather than by T. pallidum infection. We think that this is unlikely to be the case. Twenty-eight of 50 HIV-infected subjects who met our criteria for neurosyphilis also had a reactive CSF VDRL test result, which would not be expected solely on the basis of HIV infection. Furthermore, our cutoff for defining abnormal CSF WBC counts is above the mild pleocytosis caused by HIV itself [20]. When we more stringently defined neurosyphilis as a reactive CSF VDRL test result, thus removing diagnostic uncertainty due to concomitant HIV infection, the associations between neurosyphilis and serum RPR titer or peripheral blood CD4 cell count remained significant

We demonstrate that a serum RPR titer ⩾1:32 is predictive of neurosyphilis in all individuals with syphilis and that a peripheral blood CD4⁺ T cell count ⩽350 cells/μL is an additional risk factor for neurosyphilis in HIV-infected individuals. The risk associated with these parameters is independent of previous syphilis therapy and stage of syphilis. Because of the increased risk of neurorelapse after treatment with im BPG in HIV-infected individuals, we recommend that immediate lumbar puncture be performed in HIV-infected patients with syphilis, regardless of syphilis stage, particularly when the serum RPR titer is ⩾1:32 or the peripheral blood CD4⁺ T cell count is ⩽350 cells/μL. It is more difficult to judge when to perform lumbar puncture in HIV-uninfected patients with syphilis, because CSF abnormalities in the setting of early syphilis may or may not resolve spontaneously after treatment with im BPG. However, the 10.28-fold increased risk of a reactive CSF VDRL test result in HIV-uninfected subjects with a serum RPR titer ⩾1:32, regardless of syphilis stage and previous syphilis treatment, suggests that it is also reasonable to perform immediate lumbar puncture in such individuals. Further study to define factors that predict resolution of CSF abnormalities in early syphilis is needed