Central nervous system (CNS) manifestations of syphilis are generally divided into early neurosyphilis and late neurosyphilis. During the pre-HIV penicillin era, early neurosyphilis was rarely seen and late neurological manifestation of syphilis occurred in patients who were not treated or were inadequately treated (reviewed by Musher ). Early after the recognition of HIV, increasing reports of neurosyphilis among HIV-infected patients were noted – even among those who had received adequate penicillin therapy [2–5]. Additionally, surveillance data suggested that early neurosyphilis was more common than late neurosyphilis . Marra et al.[7,8] further defined risk factors and therapeutic responses when reporting that patients with CD4 counts less than or equal to 350 cells/ml and baseline rapid plasma regain (RPR) titer at least 1: 32 had increased odds of neurosyphilis and that after standard therapy for syphilis, patients with CD4 cell counts less than or equal to 200 cells/ml were less likely to normalize their cerebrospinal fluid (CSF) parameters after a median of 6.9 months of follow up. Our goal was to describe the risk factors, clinical presentation and long-term follow up of participants enrolled in a clinical cohort of HIV-1 infected patients who were diagnosed and treated for neurosyphilis.
Population and data abstraction
All HIV-infected patients who enroll in continuity care at the Johns Hopkins Moore Clinic are offered the opportunity to join the Johns Hopkins HIV Clinical Cohort. A detailed description of this dynamic cohort has been presented elsewhere . Maintenance of the database and use of its contents for analysis of patient outcomes are approved by the Institutional Review Board of the Johns Hopkins University School of Medicine. Data used for this analysis included syphilis serologies (see below), CD4 cell counts, HIV-1 RNA, and antiretroviral therapy use, including highly active antiretroviral therapy (HAART). Data on antibiotic use, other than that used for syphilis therapy, were available for azithromycin, clarithromycin, doxycycline, and oral penicillins. Data were not available for intravenous penicillins and cephalosporin use.
Patients in the cohort who were diagnosed and treated for syphilis between 1990 and 2006 were eligible. Patients were screened with the nontreponemal RPR test; reactive specimens were confirmed using a treponemal test, the fluorescence treponemal antibody absorption test (FTA-ABS). Inclusion into the study required at least two serological syphilis titers (an initial titer at the time of treatment and at least one follow-up titer) within 365 days from the date of treatment. Syphilis diagnoses were made by clinicians on the basis of the Centers for Disease Control and Prevention (CDC) criteria . Criteria for the diagnosis of neurosyphilis included positive serologies and one or more of the following: (a) one or more abnormalities on CSF examination [white blood cells >10/μl; protein >50 mg/dl; and or reactive CSF Venereal Diseases Research Laboratory (VDRL)]; (b) an otherwise unexplained neurological manifestation consistent with neurosyphilis.
Serological failure was defined as any four-fold rise in RPR titers at least 30 days following treatment, lack of four-fold drop in RPR titers at least 365 days after therapy, or clinical manifestations compatible with syphilis. Serological failures were owing to either reinfection or treatment failure. Because some patients with late disease may have very low pretreatment titers, individuals with baseline titers less than or equal to 1: 2 who did not serorevert (and who did not have clinical evidence of failure) were considered serofast and not serological failures .
Each individual patient (N = 180) may contribute one or more episodes of syphilis (N = 231). To account for these repeated measures, we used generalized estimating equations to determine the risk factors for developing neurosyphilis among the 231 syphilis cases in the cohort . We used an ‘exchangeable’ correlation structure and robust standard errors to estimate the 95% confidence limits. Univariable predictors with a P-value less than 0.2 were included in the final models; additionally, other variables that were not significant in the univariable analyses but were thought to be biologically relevant (age, race, and HIV risk factors) were also included.
When evaluating the impact of HAART and macrolide use on risk of serological failure in patients with neurosyphilis, we used the nonparametric Kaplan–Meier method to calculate time to serological failure . Because the number of neurosyphilis cases was relatively limited (N = 41), we did not attempt any multivariable comparisons. In this cohort, 40 unique patients contributed 41 cases of neurosyphilis. One patient contributed two episodes of neurosyphilis but the second case was excluded from Kaplan–Meier analyses as the model does not take into account multiple failure time data. To ensure that the exclusion of the neurosyphilis case did not affect the outcomes, we performed univariable analyses using the Cox proportional hazards model and the Andersen–Gill method  that adjusts for multiple failure time data. In the first model, we included all 41 cases of neurosyphilis while in the second model, we only included the 40 cases of neurosyphilis. Both models yielded similar results. Continuous variables were compared using the Wilcoxon signed-rank test; McNemar's test was used to compare categorical observations with repeated measures, and the K-sample equality of medians test was used to compare independent median values. Two-sided P-values were calculated and those less than 0.05 were assumed to indicate significance. STATA version 10.1 was used for all analyses (STATA Corporation, College Station, TX).
Between 1990 and 2006, 180 patients contributed 251 cases of syphilis. Follow-up data including information on syphilis therapy, immunological parameters, and medication use were available in 231 of the cases. Among these, 40 patients contributed 41 cases of neurosyphilis. The median duration of follow up after antibiotic therapy for neurosyphilis was 4.3 years and the mean number of follow-up RPR serologies was 4 (range 2–13). Thirty-four percent of cases were diagnosed between 1990 and 1995, 44% between 1996 and 2000, and 22% between 2001 and 2005.
In a multivariable model of all syphilis cases (N = 231), predictors of developing neurosyphilis (both symptomatic and asymptomatic) included a CD4 cell count of less than 350 cells/ml at the time of diagnosis [odds ratio (OR) 2.87; 95% confidence interval: 1.18–7.02], an RPR titer greater than 1: 128 vs. less than 1: 32 at the time of diagnosis (OR 2.83; 1.11–7.26), and male sex (OR 2.46; 1.06–5.70). Use of any HAART before syphilis infection reduced the odds of neurosyphilis by 65% (OR 0.35; 0.14–0.91). HIV risk factor (self-reported same sex contact, heterosexual contact, and injection drug use), age, HIV-1 RNA, and use of any macrolide prior to the diagnosis of syphilis were not associated with the risk of developing neurosyphilis. An analysis restricted to symptomatic neurosyphilis yielded similar trends but with broader confidence limits (results not shown).
The characteristics of the neurosyphilis cases are summarized in Table 1. Sixty-six percent (N = 27) of neurosyphilis patients were symptomatic on presentation. Among those, the main sign was uveitis in 33.3% (diagnosis confirmed by ophthalmology), altered cognition in 20.8%, motor weakness in 16.7%, headaches in 12.1%, gait abnormality in 9.0%, hearing loss in 4.2%, and Bell's palsy in 4.2%. Thirty-four percent (N = 14) were asymptomatic at the time of neurosyphilis diagnosis; a lumbar puncture was performed in these patients for the following reasons: lack of serological (RPR) response to therapy in 57.0%, and high baseline RPR titer (>1: 32) in 36.0%.
Median RPR titer at the time of diagnosis of neurosyphilis was 1: 512 [interquartile range (IQR) 1: 64–1: 2048]. Figure 1 summarizes the RPR titers among symptomatic and asymptomatic patients. Median baseline CD4 cell counts at the time of neurosyphilis diagnosis among asymptomatic and symptomatic patients were 331 (IQR 72–365) and 189 (87–324), respectively (P > 0.05).
Symptomatic patients had more abnormalities on initial lumbar puncture compared with asymptomatic patients (Table 2). Time from syphilis diagnosis to neurosyphilis could be determined in 27/41 (66%) patients; median time was 9 months (IQR 1–30); 63% (N = 17) occurred within 1 year of syphilis diagnosis (Fig. 2); of those 6 of 17 (35%) occurred in symptomatic patients within the first month.
Clinical data relating to neurosyphilis were available in 37 of 41 (90.2%) cases at 1 year. Of those, 43.2% (N = 16) had a follow-up lumbar puncture within 12 months (median time to lumbar puncture was 10 months); 37.5% (N = 6) had resolution of all cerebrospinal fluid abnormalities, 43.8% (N = 7) had improved parameters, and 19.0% (N = 3) had unchanged or worsening parameters. Of the latter three cases, two were symptomatic at the time of the lumbar puncture, were found to have worsening CSF parameters (elevation of CSF WBC count and an increase in the CSF VDRL titer as compared to initial lumbar puncture), and were retreated for neurosyphilis. Seventy-one percent of patients who received at least 1 month of HAART after treatment for neurosyphilis and 67% of those who did not receive any HAART had evidence of improvement or resolution of abnormal CSF parameters. At 1 year, 38% had persistence of their major symptom despite adequate treatment for neurosyphilis and 71% had at least one symptom that could be attributed to their previous neurosyphilis diagnosis. Among those with persistent symptoms at 1 year, 65% had received HAART during that year.
Retreatment was deemed necessary by managing clinicians in 12 of 41 (29.3%) patients. Median time to retreatment was 536 days. Of those, eight (66.7%) were asymptomatic and were retreated because they met the criteria for serological failure (five had a four-fold rise in titers and three did not manifest the expected four-fold drop in titers 1 year or more after appropriate therapy). These patients were treated with 7.2 million units of benzathine penicillin G over 3 weeks. Among the four who were retreated owing to symptoms, two had CNS symptoms and met the CSF criteria for recurrent neurosyphilis (see above) and two presented with secondary syphilis (both patients had no neurological symptoms and refused lumbar puncture). Retreatment consisted of 14 days of intravenous penicillin G for the former and 2.4 MU of benzathine penicillin for the latter. The effects of HAART and macrolide use (for opportunistic infection prophylaxis or therapy) following treatment of neurosyphilis are presented in Fig. 3. The use of HAART for more than 6 months and the cumulative use of at least 3 months of macrolides during the follow-up (i.e. after therapy for neurosyphilis) were associated with a trend towards decreased incidence of serological failure although none of the differences were statistically significant.
In this cohort, the degree of immunosuppression, as measured by CD4 cell count, was an independent risk factor for developing neurosyphilis, and the use of HAART reduced the odds of neurosyphilis by 65%. Similarly, among patients diagnosed and treated for neurosyphilis, there was a trend for decreased risk of serological failure among patients who received six or more months of HAART therapy during a median follow-up of 4.3 years. These findings highlight the importance of the immune response in controlling syphilis and the possible role of HAART in mitigating neurological complications of syphilis among co-infected patients. Owing to the smaller number of patients who had follow-up lumbar puncture data, we were unable to determine whether the use of HAART, macrolides, and the degree of immunosuppression at the time of syphilis diagnosis were associated with resolution of CSF abnormalities at 1 year.
We diagnosed early neurosyphilis in 17 of 27 (63%) patients in whom time from syphilis exposure to neurosyphilis could be determined (Fig. 2). Our findings are similar to those reported elsewhere [6,15]. On the basis of a few neurosyphilis case reports published at the time, Musher  concluded in 1991 that ‘an interesting analogy can be drawn between the prepenicillin era, when progression to early neurosyphilis resulted from treating immunologically normal hosts with insufficient amounts of a barely effective antimicrobial agent, and the modern era, when we use a far more effective antimicrobial agent in a severely immunodeficient host’.
Our study has several limitations. The number of patients with neurosyphilis was relatively small and, therefore, our analyses may not have statistical power to detect clinically relevant findings. Time from syphilis exposure to the diagnosis of neurosyphilis may be biased, especially for late onset neurosyphilis. For example, the assumption in Fig. 2 is that interim exposures did not occur. This would tend to underestimate the true proportion of early neurosyphilis cases. We do not believe this to be a major issue, as our findings are very similar to those of the other studies. We did not have data on penicillin and cephalosporin use (other than that used for syphilis therapy) during follow up other than the use of oral penicillin, amoxicillin, and amoxicillin/clavulanic acid. Data on medication use was on the basis of patient's self-reports to their clinicians and are thus potentially biased. Although most serologic testing was done at the Johns Hopkins Hospital laboratory, variation in test reagents may yield different results. This may have led to some misclassification of serologic failure but the misclassification should be nondifferential across groups. Finally, the interpretation of serological failure can be difficult and included patients experiencing either treatment failure or reinfection. In all studies of clinical syphilis, the distinction between these two events is very difficult to make. Thus, we decided to combine these outcomes to avoid misclassification. The 16-year study span made any distinction between the two events more difficult and less reliable. In a separate analysis evaluating serological failures among all stages of syphilis in this cohort, evaluating serological failure as early (<2 years) vs. late or lack of four-fold drop in titers vs. four-fold increases in titers (both strategies are attempts to separate treatment failure from reinfection), we found a consistent protective effect for both HAART use and macrolide use (data not shown). In addition, serologic response following syphilis therapy in HIV co-infected patients may be slower than HIV noninfected patients . Because this was an observational study, we could not influence the decision to treat. In our study, the median time to retreatment was 528 days. It is possible that the number of failures because of lack of a four-fold response would have been attenuated if clinicians had waited the full 730 days before treating. Although the study mirrors routine clinical practice, it may have overestimated the number of serologic failures.
The immune system is clearly vital in controlling the extent of syphilis involvement of the CNS. Since the early days of HIV, despite the use of recommended treatment regimens for neurosyphilis, increased rates of neurosyphilis and clinical relapse among co-infected patients have been well documented [16,17]. The recommended treatment regimens in HIV-infected patients are similar to those for HIV-negative patients . The follow-up schedule, however, is more aggressive. Ensuring that patients follow up in a timely manner is challenging, especially in public health settings . Our study demonstrates that targeting HIV-induced immunosuppression should be an important target of syphilis therapy to try and reduce the long-term morbidity among co-infected patients.
The authors wish to thank the staff and patients of the Johns Hopkins HIV Clinical Cohort.
National Institutes of Health K23-HD047395 to K.G.G. and National Institutes of Aging R01-AG026250 and Drug Abuse (K23-DA00523, K24-DA00432, and RO1-DA-11602) to (K.A.G. and R.D.M.) and JHU Richard S. Ross Clinician Scientist Award (K.A.G.).
K.G.G. and K.A.G. participated in the study design, data collection, data analysis, and manuscript preparation. R.D.M. participated in the data collection, data analysis, and manuscript preparation. A.M.R., E.J.E., and J.M.Z. participated in the study design, data analysis, and manuscript preparation.
All authors have seen and approved this version of the manuscript.
Portions of the data in this manuscript were presented as an oral presentation (abstract # 0-04) at the International Society for Sexually Transmitted Diseases Research conference in Seattle, Washington (July, 2007). The manuscript presented here is part of a Doctor of Philosophy thesis manuscript titled: Syphilis as an opportunistic infection in HIV (K.G.G.) at the Johns Hopkins Bloomberg School of Public Health.
There are no conflicts of interest.
1. Musher DM. Syphilis
, penicillin, and AIDS. J Infect Dis 1991; 163:1201–1206.
2. Berry CD, Hooton TM, Collier AC, Lukehart SA. Neurologic relapse after benzathine penicillin therapy for secondary syphilis
in a patient with HIV
infection. N Engl J Med 1987; 316:1587–1589.
3. Johns DR, Tierney M, Felsenstein D. Alteration in the natural history of neurosyphilis
by concurrent infection with the human immunodeficiency virus. N Engl J Med 1987; 316:1569–1572.
4. Katz DA, Berger JR. Neurosyphilis
in acquired immunodeficiency syndrome. Arch Neurol 1989; 46:895–898.
5. Berger JR. Neurosyphilis
in human immunodeficiency virus type 1-seropositive individuals. A prospective study. Arch Neurol 1991; 48:700–702.
6. Flood JM, Weinstock HS, Guroy ME, Bayne L, Simon RP, Bolan G. Neurosyphilis
during the AIDS epidemic, San Francisco, 1985–1992. J Infect Dis 1998; 177:931–940.
7. Marra CM, Maxwell CL, Smith SL, Lukehart SA, Rompalo AM, Eaton M, et al
. Cerebrospinal fluid abnormalities in patients with syphilis
: association with clinical and laboratory features. J Infect Dis 2004; 189:369–376.
8. Marra CM, Maxwell CL, Tantalo L, Eaton M, Rompalo AM, Raines C, et al
. Normalization of cerebrospinal fluid abnormalities after neurosyphilis
therapy: does HIV
status matter? Clin Infect Dis 2004; 38:1001–1006.
9. Moore RD. Understanding the clinical and economic outcomes of HIV
therapy: the Johns Hopkins HIV
clinical practice cohort. J Acquir Immune Defic Syndr Hum Retrovirol 1998; 17(Suppl. 1):S38–S41.
10. Wharton M, Chorba TL, Vogt RL, Morse DL, Buehler JW. Case definitions for public health surveillance. MMWR Recomm Rep 1990; 39:1–43.
11. Ghanem KG, Erbelding EJ, Wiener ZS, Rompalo AM. Serological response to syphilis
treatment in HIV
-positive and HIV
-negative patients attending sexually transmitted diseases clinics. Sex Transm Infect 2007; 83:97–101.
12. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:121–130.
13. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 3:457–481.
14. Andersen PK, Gill RD. Cox's regression model for counting processes: a large sample study. Ann Stat 1982; 10:1100–1120.
15. Centers for Disease Control and Prevention (CDC). Symptomatic early neurosyphilis among HIV-positive men who have sex with men – four cities, United States, January 2002–June 2004.MMWR Morb Mortal Wkly Rep
16. Gordon SM, Eaton ME, George R, Larsen S, Lukehart SA, Kuypers J, et al
. The response of symptomatic neurosyphilis
to high-dose intravenous penicillin G in patients with human immunodeficiency virus infection. N Engl J Med 1994; 331:1469–1473.
17. Walter T, Lebouche B, Miailhes P, Cotte L, Roure C, Schlienger I, et al
. Symptomatic relapse of neurologic syphilis
after benzathine penicillin G therapy for primary or secondary syphilis
-infected patients. Clin Infect Dis 2006; 43:787–790.
18. Workowski KA, Berman SM. Centers for Disease Control and Prevention. Sexually transmitted diseases treatment guidelines, 2006. MMWR Recomm Rep 2006; 55:1–94.