Neurosyphilis Treatment Outcomes After Intravenous Penicillin G Versus Intramuscular Procaine Penicillin Plus Oral Probenecid

Abstract

Background

Data comparing neurosyphilis treatment regimens are limited.

Methods

Participants were enrolled in a study of cerebrospinal fluid (CSF) abnormalities in syphilis that was conducted at the University of Washington between April 2003 to May 2014. They were diagnosed with syphilis and referred by their providers due to concerns for neurosyphilis. We evaluated 150 people with CSF abnormalities who were treated with either intravenous aqueous penicillin G (PenG) or intramuscular aqueous procaine penicillin G plus oral probenecid (APPG-P). An abnormal CSF diagnosis was defined as a white blood cell (WBC) count >20/µL, a CSF protein reading >50 mg/dL, or a reactive CSF–Venereal Disease Research Laboratory test (VDRL). Hazard ratios for normalization of CSF or serum measures were determined using Cox regression.

Results

In individuals treated with either PenG or APPG-P, CSF WBCs and CSF-VDRL reactivity normalized within 12 months after treatment, while protein normalized more slowly and less completely. There was no relationship between treatment regimen or human immunodeficiency virus (HIV) status and likelihood of normalization of any measure. Among those living with HIV, CSF WBC counts and CSF-VDRL reactivity were more likely to normalize in those treated with antiretrovirals. Unexpectedly, CSF WBCs were more likely to normalize in those with low CD4+ T cell counts. When neurosyphilis was more stringently defined as a reactive CSF-VDRL, the relationship with the CD4+ T cell count remained unchanged.

Conclusions

In the current antiretroviral treatment era, neurosyphilis treatment outcomes are not different for PenG and APPG-P, regardless of HIV status. The relationship between the normalization of CSF WBC counts and CD4+ T cell counts may indicate continued imprecision in neurosyphilis diagnostic criteria, due to HIV-related CSF pleocytosis.

In this study of individuals diagnosed with neurosyphilis, there was no significant difference in laboratory outcomes when comparing treatment with intravenous penicillin G versus intramuscular procaine penicillin given with oral probenecid, regardless of human immunodeficiency virus status.

In the United States and other developed countries, syphilis rates have continued to rise, particularly among men who have sex with men, many of whom also live with human immunodeficiency virus (HIV) [1–3]. Neurosyphilis is a serious complication of syphilis that may be more common in persons living with HIV (PLWH) [4]. The US Centers for Disease Control and Prevention (CDC) has recommended that neurosyphilis be treated with 18–24 million units of intravenous (IV) aqueous penicillin G (PenG) for 10–14 days. Intramuscular aqueous procaine penicillin G at 2.4 million units daily, plus oral probenecid at 500 mg, taken orally 4 times daily (APPG-P) for 10–14 days, is recommended as an alternative if compliance can be assured [5]. These recommendations are based on limited data that indicate penicillin reaches therapeutic levels in the cerebrospinal fluid (CSF) with the aforementioned regimens [6–8] and on many years of clinical experience. APPG-P may be easier to use than PenG, because it does not require a peripherally inserted central catheter or hospitalization for IV therapy, costs less, has less risk of infection or illicit use of peripherally inserted central catheters, and does not require insurance coverage for IV therapy.

Neurosyphilis treatment responses are determined by the reexamination of CSFs. We previously reported that, after neurosyphilis treatment, PLWH were significantly less likely than those living without HIV to normalize CSF-VDRL reactivity, regardless of the neurosyphilis treatment regimen [9]; the power of the study was not sufficient to compare treatment regimens. In this analysis, we evaluated the neurosyphilis treatment responses of 150 individuals who were enrolled after our previous report, to determine whether there were significant differences between treatment responses to PenG and APPG-P and to identify other factors that could influence responses.

METHODS

Participants were enrolled in a study of CSF abnormalities in syphilis that was conducted at the University of Washington between April 2003 to May 2014. Study eligibility included clinical or serological evidence of syphilis and an assessment by the referring provider that the patient was at risk for neurosyphilis, either due to neurological symptoms or other factors associated with increased risks [10], such as a serum rapid plasma reagin (RPR) titer ≥1:32 or a CD4+ T-cell count ≤350/uL in PLWH [11–13]. Other inclusion criteria included fluency in the English language, no contraindication to a lumbar puncture, no receipt of IV or intramuscular antimicrobials active against neurosyphilis within 3 months of enrollment (except for a dose of 250 mg of ceftriaxone for gonorrhea), and no evidence of another central nervous system process that could lead to abnormal CSF findings. All participants had reactive serum RPR and confirmatory treponemal tests.

CSF abnormalities considered to be consistent with neurosyphilis included a white blood cell (WBC) count >20/µL, a protein reading >50 mg/dL, or a reactive CSF-VDRL test. CSF-VDRL reactivity is considered as the gold standard for neurosyphilis diagnosis, but sensitivity ranges from 27–72% [14–16]. Thus, elevated CSF WBC counts and protein concentrations are also used in the diagnosis of neurosyphilis. A CSF WBC count >20/µL (rather than >5/µL) improves the diagnostic specificity of neurosyphilis for PLWH, because HIV infection, especially in absence of antiretroviral therapy, can lead to CSF pleocytosis [17, 18]. Similarly, HIV can result in the mild elevation of CSF protein, but concentrations above 50 mg/dL are rarely due to HIV; this cut-off may increase the diagnostic accuracy for neurosyphilis [11, 17–21]. We did not use CSF treponemal antibody tests, such as Treponema pallidum particle agglutination, because the performance of these tests can be variable, and a negative test does not always exclude the diagnosis of neurosyphilis, particularly in neurologically symptomatic individuals with syphilis [22].

Participants had asymptomatic neurosyphilis if they had 1 or more CSF abnormalities, but no neurologic abnormalities. Participants with symptomatic neurosyphilis had 1 or more CSF abnormalities, including symptomatic meningitis, vision or hearing loss, or stroke. Participants underwent study visits at entry and at Weeks 12, 24, and 52 after therapy. At each study visit, a standardized history, neurological examination, and blood draw were performed. Lumbar punctures were performed at entry and Week 12, as well as at Weeks 24 and 52 if the previous CSF profile was abnormal, as defined above. If a participant missed a study visit, they were allowed to complete the visit late. The study protocol was reviewed and approved by the University of Washington Institutional Review Board, and all study participants provided written informed consent.

Cerebrospinal fluid WBC enumerations, protein concentrations, and VDRL reactivity results were determined in a Clinical Laboratory Improvement Amendments–approved hospital clinical laboratory. Serum RPR tests were performed in a research laboratory, using previously published methods [23]. Medical record reviews provided results of plasma HIV RNA copy numbers and peripheral blood CD4+ T lymphocyte counts. Only HIV RNA copy numbers and CD4+ T cell counts obtained within 90 days of a lumbar puncture were included in the analysis.

Descriptive results are expressed as number (percent) or median (interquartile range [IQR]). Relationships between categorical variables were determined by a Chi-square or Fisher’s exact test. The time to normalization of each measure is graphically represented by a Kaplan Meier plot. Hazard ratios (HRs) for the normalizations of CSF WBC counts (decline to <20/µL), CSF protein measures (decline to <50 mg/dL), and CSF-VDRL or serum RPR reactivity results (4-fold decline or reversion to nonreactive status) were determined using Cox regression. Antiretroviral use was included as a time-dependent variable. For multivariable analyses, covariates with P values ≤ .10 in univariate analyses were included. P values ≤ .05 were considered significant.

RESULTS

Characteristics of Participants

Study participants with asymptomatic or symptomatic neurosyphilis who were treated with recommended doses of PenG or APPG-P for at least 10 consecutive days were included in the analysis (Figure 1), a total of 223 participants received some treatment for neurosyphilis. Of these, 4 were excluded because blood could not be obtained and 5 were excluded due to a prior diagnosis of neurosyphilis. There were 173 participants who had 1 or more qualifying CSF abnormalities; others were treated for neurosyphilis by their providers based on clinical findings, such as vision or hearing loss, in the absence of CSF abnormalities. Of the 173, 150 were treated with 1 of the 2 CDC-recommended treatment regimens; individuals treated with ceftriaxone or oral doxycycline, or who did not complete 10 days of PenG or APPG-P, were excluded. Neurosyphilis treatment regimens were chosen by the participant and primary care provider and were not provided by the study. No diagnosis associated with CSF abnormalities, other than syphilis (or HIV), was identified in any participant.

Figure 1.

Flow diagram of study participants included in the analysis. An abnormal cerebrospinal fluid (CSF) diagnosis included a white blood cell count greater than 20/µL, a protein concentration greater than 50 mg/dL, or a reactive CSF–Venereal Disease Research Laboratory test. Abbreviations: APPG-P, intramuscular aqueous procaine penicillin G with oral probenecid; CDC, Centers for Disease Control and Prevention; NS, neurosyphilis; PenG, intravenous aqueous penicillin G.

The entry-visit characteristics of the 150 participants are shown in Table 1. There were 32 participants treated with PenG, and 118 treated with APPG-P. This difference reflects the fact that most participants were recruited from a publicly funded sexually transmitted diseases clinic, where APPG-P was the standard neurosyphilis treatment and was provided at a low cost. Most participants were White men, the median age was 41 years (IQR 32–52), and most participants were living with HIV. Compared to those treated with PenG, more individuals treated with APPG-P were treated for uncomplicated (non-neurological) syphilis before their lumbar puncture. More people treated with APPG-P also had early stage syphilis, and their CSF protein concentrations were lower, on average (Table 1). There were no other significant differences between the 2 treatment groups.

Table 1.

Participant Characteristics

Characteristic	PenG, n = 32	APPG-P, n = 118	P Value
HIV	28 (87.5%)	88 (74.6%)	NS
Age	43 (32–52)	40 (33–46)	NS
Male	31 (96.9%)	118 (100%)	NS
Race
White	24 (75.0%)	97 (82.2%)	NS
Black	4 (12.5%)	14 (11.9%)
Other	4 (12.5%)	7 (5.9%)
RPR titer	128 (64–512)	128 (64–256)	NS
Early syphilisa	15 (46.9%)	84 (71.2%)	.01
Late syphilisa	17 (53.1%)	34 (28.8%)
Treated for uncomplicated syphilis before LPb	8 (25.0%)	52 (44.1%)	.05
CSF WBC >20/µL	23 (71.9%)	72 (610%)	NS
CSF WBC/µL	31(10–77)	26 (13–46)	NS
CSF protein >50/dL	24 (75.0%)	68 (57.6%)	NS
CSF protein mg/dL	69 (48–107)	53 (39–66)	.003
CSF-VDRL reactive	21 (65.6%)	69 (58.5%)	NS
CSF VDRL titer	2 (NR–8)	1 (NR–2)	NS
Symptomatic eurosyphilisc	16 (61.5%)	50 (48.1%)	NS

Results are expressed as n (%) or median (interquartile range). Data were considered NS at P > .05.

Abbreviations: APPG-P, intramuscular aqueous procaine penicillin G and oral probenecid; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; LP, lumbar puncture; NR, nonreactive; NS, not significant; PenG, intravenous penicillin G; RPR, rapid plasma reagent; VDRL, Venereal Disease Research Laboratory; WBC, white blood cells.

^aEarly syphilis included the primary, secondary, and early latent stages. Late syphilis included late-latent syphilis and syphilis of an unknown duration.

^bTreatment started within 90 days before lumbar puncture.

^cSymptomatic neurosyphilis included symptomatic meningitis, vision or hearing loss, or stroke.

There were 116 PLWH included in the analysis. CD4+ T cell and plasma HIV RNA measurements, taken within 90 days of study entry, were available for 103 individuals. The median CD4+ T cell count within 90 days of study entry was 392 (IQR 259–522) cells/µL, and only 17 (16.5%) participants had values ≤200. The median log plasma HIV RNA copy number was 4.37 (IQR 1.88–4.97), and the value was ≤50 in 23 (22.3%) participants. As expected, HIV RNA copy numbers ≤50 were significantly more common in those on antiretroviral therapy (ARV), compared to those not on ARV (21 [61.8%] of 34 vs 0 of 60, respectively; P < .001).

Cerebrospinal Fluid and Serum Abnormalities Before Treatment

Of the 150 individuals included in the analysis, 95 had an elevated CSF WBC concentration, 92 had an elevated CSF protein concentration, and 90 had a reactive CSF-VDRL test. In 25 individuals, an elevated CSF WBC count was the only CSF abnormality; in 17 individuals, an elevated CSF protein concentration was the only abnormality. Having an elevated CSF WBC count or protein concentration as the only abnormality was not more common in PLWH, compared to those living without HIV, or in those with symptomatic neurosyphilis, compared to those with asymptomatic neurosyphilis. Serum RPR and treponemal tests were reactive in all participants.

Normalization in All Participants

Figure 2 shows the time to normalization of each CSF measure. As in our previous report [9], CSF WBC counts and CSF-VDRL reactivity normalized in almost all individuals by 12 months after treatment, while protein normalized more slowly and less completely. Table 2 shows univariate analyses of the HRs of normalization of CSF and serum measures in all study participants. There was no significant relationship found between treatment regimen and normalization of any measure. Similarly, there was no relationship between receiving treatment for uncomplicated syphilis within 90 days before a lumbar puncture and the likelihood of normalization of any measure. CSF protein was less likely to normalize in participants with higher pretreatment concentrations. CSF-VDRL reactivity was more likely to normalize in participants with symptomatic neurosyphilis and less likely to normalize in those with late-stage syphilis. Serum RPR was more likely to normalize in PLWH, participants with higher pretreatment titers, and those with symptomatic neurosyphilis, and less likely to normalize in those with late-stage syphilis.

Figure 2.

Kaplan–Meier plot of time to normalization of each cerebrospinal fluid measure. Abbreviations: CSF, cerebrospinal fluid; VDRL, Venereal Disease Research Laboratory; WBC, white blood cells.

Table 2.

Normalization of Cerebrospinal Fluid and Serum Measures in All Participants

Variable	WBC	Protein	CSF-VDRL	Serum RPR
Treated with APPG-P	NS	NS	NS	NS
HIV	NS	NS	NS	1.52 (.99–2.32), P = .06
Pretreatment outcome measure value ≥ mediana	NS	0.16 (.08–.34), P < .001	NS	2.28 (1.56–3.35), P < .001
Late-stage syphilisb	NS	NS	0.46 (.28–.76), P = .002	0.38 (.25–.56), P < .001
Symptomatic neurosyphilisc	NS	NS	1.74 (1.07–2.83), P = .03	1.48 (1.03–2.14), P = .04

Data are shown as univariate hazard ratios (95% confidence intervals). A normalization of CSF WBC is defined as a decline to <20/µL; a normalization of CSF protein is defined as a decline to <50 mg/dL; and a normalization of CSF-VDRL or serum RPR reactivity is defined as a 4-fold decline or reversion to nonreactive status. All P-values ≤ .10 are shown; NS, P > .10.

Abbreviations: APPG-P, intramuscular aqueous procaine penicillin G and oral probenecid; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; NS, not significant; RPR, rapid plasma reagent; VDRL, Venereal Disease Research Laboratory; WBC, white blood cells.

^aMedian of those with an abnormal value, as defined in the text.

^bLate syphilis, compared to early syphilis. Late syphilis included late-latent syphilis and syphilis of an unknown duration. Early syphilis included the primary, secondary, and early latent stages.

^cSymptomatic neurosyphilis, compared to asymptomatic neurosyphilis. Symptomatic neurosyphilis included symptomatic meningitis, vision or hearing loss, or stroke. Asymptomatic neurosyphilis included those with abnormal cerebrospinal fluid but no neurologic abnormalities.

In multivariate analyses, taking into account stages and symptomatic neurosyphilis, CSF-VDRL reactivity was more likely to normalize in participants with symptomatic neurosyphilis (HR 1.79, 95% confidence interval [CI] 1.09–2.93; P = .02) and less likely to normalize in those with late syphilis (HR 0.41, 95% CI .24–.69; P = .001). In multivariate analyses, taking into account HIV status, pretreatment titer, stage, and symptomatic neurosyphilis, serum RPR was more likely to normalize in those with higher pretreatment titers (HR 2.01, 95% CI 1.30–3.12; P = .002) and less likely to normalize in late syphilis (HR 0.47, 95% CI .30–.73; P = .001); there was no relationship with HIV or symptomatic neurosyphilis. We examined whether there was a significant relationship between stage and symptomatic neurosyphilis that might explain some of our findings. More participants with symptomatic neurosyphilis had early syphilis than late syphilis (49 [74.2%] of 66 vs 17 [25.8%] of 66, respectively; P = .11), but the interaction between stage and symptomatic neurosyphilis was not significant in multivariate analyses.

Normalization in Participants Living With Human Immunodeficiency Virus

As shown in Table 3, in univariate analyses, all CSF measures and serum RPR tests were more likely to normalize in PLWH who were actively using ARVs. There were no relationships between treatment for uncomplicated syphilis within 90 days before lumbar puncture or undetectable HIV RNA and the likelihood of normalization of any measure. CSF WBC counts were more likely to normalize in those with CD4+ T cell counts ≤200/µL, and were less likely to normalize in those who were treated with APPG-P, as compared to PenG. CSF protein concentrations were also more likely to normalize in those with low CD4+ T cells, and were less likely to normalize in those with higher pretreatment CSF protein concentrations. CSF-VDRL tests were more likely to normalize in those with symptomatic neurosyphilis and less likely to normalize in late-stage syphilis. Serum RPR was more likely to normalize in those with higher pretreatment titers, and was less likely to normalize in those with late-stage syphilis. In multivariate analyses, the above associations remained significant, except for the relationship between the normalization of WBC counts and treatment with APPG-P; the relationship between ARV use and low CD4+ T cell counts and the normalization of CSF protein concentrations; and the relationship between ARV use and serum RPR (Table 4).

Table 3.

Univariate Hazard Ratios of Normalization of Cerebrospinal Fluid and Serum Measures in Participants Living With Human Immunodeficiency Virus

Variable	WBC	Protein	CSF VDRL	Serum RPR
Treated with APPG-P	0.56 (.33–.95), P = .03	NS	NS	NS
Pretreatment outcome measure value ≥ mediana	NS	0.17 (.07–.39), P < .001	NS	2.27 (1.44–3.56), P < .001
Late-stage syphilisb	NS	NS	0.41 (.22–.74), P = .003	0.48 (.31–.73), P = .001
Symptomatic neurosyphilisc	NS	NS	1.69 (.97–2.96), P = .07	NS
Active ARV use over time	2.72 (1.64–4.53), P ≤ .001	2.13 (.95–4.76), P = .07	1.64 (.95–2.82), P = .08	1.46 (.98–2.19), P = .07
CD4+ ≤200/µL	2.14 (1.07–4.29), P = .03	2.90 (1.22–6.87), P = .02	NS	NS

Data are shown as univariate hazard ratios (95% confidence intervals). A normalization of CSF WBC is defined as a decline to <20/µL; a normalization of CSF protein is defined as a decline to <50 mg/dL; and a normalization of CSF-VDRL or serum RPR reactivity is defined as a 4-fold decline or reversion to a nonreactive status. All P-values ≤ .10 are shown; NS, P > .10. Abbreviations: APPG-P, intramuscular aqueous procaine penicillin G and oral probenecid; ARV, antiretroviral therapy; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; NS, not significant; RPR, rapid plasma reagent; VDRL, Venereal Disease Research Laboratory; WBC, white blood cells.

^aMedian of those with HIV and an abnormal value, as defined in the text.

^bLate syphilis, compared to early syphilis. Late syphilis included late-latent syphilis and syphilis of an unknown duration. Early syphilis included the primary, secondary, and early latent stages.

^cSymptomatic neurosyphilis, compared to asymptomatic neurosyphilis. Symptomatic neurosyphilis included symptomatic meningitis, vision or hearing loss, or stroke. Asymptomatic neurosyphilis included those with abnormal cerebrospinal fluid but no neurologic abnormalities.

Table 4.

Multivariate Adjusted Hazard Ratios of Normalization of Cerebrospinal Fluid and Serum Measures in Participants Living With Human Immunodeficiency Virus

Variable	WBC	Protein	CSF VDRL	Serum RPR
Treated with APPG-P	NS	NI	NI	NI
Pretreatment outcome measure value ≥ mediana	NI	0.14 (.05–.43), P = .001	NI	1.82 (1.10–3.01), P = .02
Late-stage syphilisb	NI	NI	0.36 (.19–.69), P = .002	0.60 (.38–.94), P = .03
Symptomatic neurosyphilisc	NI	NI	2.30 (1.27–4.18), P = .006	NI
Active ARV use over time	2.52 (1.46–4.36), P = .001	NS	2.10 (1.18–3.75), P = .01	NS
CD4+ ≤200/µL	2.77 (1.28–5.98), P = .01	NS	NI	NI

Data are shown as multivariate adjusted hazard ratios (95% confidence intervals). A normalization of CSF WBC is defined as a decline to <20/µL; a normalization of CSF protein is defined as a decline to <50 mg/dL; and a normalization of CSF-VDRL or serum RPR reactivity is defined as a 4-fold decline or reversion to nonreactive status. NI (not included), P-values > .10 in univariate analyses. As described in Methods, these were not included in the multivariate analyses. NS, P > .05.

Abbreviations: APPG-P, intramuscular aqueous procaine penicillin G and oral probenecid; ARV, antiretroviral therapy; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; NS, not significant; RPR, rapid plasma reagent; VDRL, Venereal Disease Research Laboratory; WBC, white blood cells.

^aMedian of those with HIV and an abnormal value, as defined in the text.

^bLate syphilis, compared to early syphilis. Late syphilis included late-latent syphilis and syphilis of an unknown duration. Early syphilis included the primary, secondary, and early latent stages.

^cSymptomatic neurosyphilis, compared to asymptomatic neurosyphilis. Symptomatic neurosyphilis included symptomatic meningitis, vision or hearing loss, or stroke. Asymptomatic neurosyphilis included those with abnormal cerebrospinal fluid but no neurologic abnormalities.

We expected that CSF abnormalities in individuals with HIV would be less likely, rather than more likely, to normalize if their CD4+ T cells were lower. Previous work in individuals without syphilis showed that HIV-related CSF pleocytosis is about 25-fold less likely in individuals with CD4+ T cell counts ≤200/µL, compared to those with CD4+ T cell counts >200/µL [17]. To further explore the relationship between low CD4+ T cell counts and the increased likelihood of normalization of CSF WBC counts, we repeated the multivariate analyses, restricting them to individuals with reactive CSF-VDRL results. The normalization of CSF WBC counts remained significantly more likely in those with CD4+ T cell counts ≤200/µL (HR 3.11, 95% CI 1.15–8.42; P = .03).

DISCUSSION

The goal of this analysis was to determine whether neurosyphilis treatment responses differed between individuals who were treated with PenG versus APPG-P, 2 regimens recommended by the CDC for neurosyphilis. We found no difference in the likelihood of normalization of CSF and serum measures between the 2 treatment groups in all individuals or in analyses restricted to PLWH. Because of the nature of referrals to this study, the treatment regimens could not be randomized. However, we took into account in our analyses variables such as the severity of CSF abnormalities, the presence or absence of symptomatic neurosyphilis, or HIV-related factors that might have influenced treatment choices.

Among PLWH, in time-dependent analyses, we found that individuals taking ARVs were about twice as likely to normalize CSF WBC counts and CSF-VDRL reactivity as those not using ARVs. Our finding that ARV use conveyed an increased likelihood of normalization of CSF WBC counts supports the hypothesis that CSF pleocytosis might have been due, at least in part, to HIV, and not to neurosyphilis. However, HIV alone would not explain CSF-VDRL reactivity. The benefits of ARV use on the normalization of CSF-VDRL are consistent with those shown in previous works showing that neurosyphilis and serological failures after the treatment of uncomplicated syphilis are both less likely in PLWH who take ARVs [11, 24].

Of note in our study is the counterintuitive finding that CSF WBC counts were more likely to normalize in PLWH with low CD4+ T cell counts, even when neurosyphilis was rigorously defined as a reactive CSF-VDRL. The number of individuals with CD4+ T cell counts ≤200 was low, so it is possible that this association is spurious. Another possibility is that the CSF WBC responses in individuals with low CD4+ T cell counts, who are much less likely to have HIV-associated pleocytosis than individuals with higher CD4+ T cell counts [17], truly reflected the antibiotic responses. CSF pleocytosis in PLWH with higher CD4+ T cell counts may be due to both HIV and neurosyphilis. Only CSF pleocytosis attributable to neurosyphilis would be expected to change in response to penicillin.

In univariate analyses, regardless of HIV status, CSF protein concentrations were less likely to normalize when the pretreatment concentration was high. Previous studies have shown that CSF protein concentrations may remain elevated despite the normalization of other CSF measures, independent of HIV status [9]. This finding may suggest that the definition of an elevated protein concentration is too conservative, or that the time to normalization of a CSF protein concentration after successful neurosyphilis treatment may be longer than the time to normalization of other measures.

In uncomplicated syphilis (non-neurosyphilis), serum RPR titers normalize faster and more completely after treatment in those with higher titers and in those with early syphilis, compared to late syphilis [25]. Our findings after neurosyphilis treatment also showed these associations. The more rapid normalization of CSF-VDRL reactivity in those with symptomatic neurosyphilis may also be related to the syphilis stage, because more individuals with early syphilis, compared to late syphilis, were neurologically symptomatic. However, we were not able to support this hypothesis statistically.

Our findings regarding the normalization of CSF-VDRL reactivity and serum RPR tests broadly reproduce our analysis of a different group of individuals, who were treated for neurosyphilis 15 years ago [9]. However, we did not see differences by HIV status itself, which were notable in our previous analysis. The number of individuals in the current investigation was more than twice that in the previous study and, based on dates of enrollment, more PLWH in the current analysis were likely to be taking combination ARVs than in the previous study. We were unable to test this hypothesis because we did not systematically collect HIV treatment data during the earlier study period. However, one-third of individuals enrolled before those included in the current analysis had CD4+ T cell counts ≤200/µL, and 85% had HIV RNA numbers ≥500 copies/mL, with a median of 11 300 copies/mL (unpublished data). While HIV RNA numbers may increase in acute syphilis, the difference is generally well below 1 log [26], consistent with the assertion that the earlier study population had advanced and untreated HIV.

It is commonly believed that CSF abnormalities in early syphilis may normalize after treatment for uncomplicated syphilis [4], although it is not clear whether this phenomenon applies to CSF-VDRL reactivity. We did not find a relationship between syphilis stage, defined as being either early or late, and the normalization of CSF pleocytosis or elevated protein concentrations. However, we did see a relationship between the stage and the normalizations of CSF-VDRL reactivity and serum RPR tests, which are 2 syphilis-specific measures. Even if the spontaneous normalization of CSF WBC counts or protein concentrations had occurred in some of our study participants, we hypothesized that the normalization would not differ by treatment regimen. To explore this possibility, we repeated the analyses of normalization of all measures and treatment regimens, restricting the stage to those with late disease, both among all participants and among those with HIV. Our results were unchanged (data not shown), and our original conclusions remained supported.

The limitations of our study, in addition to those addressed above, should be considered in interpreting our findings. The number of individuals without HIV was small, reflecting the demographics of syphilis in our community, and may not have been sufficient to demonstrate differences by HIV status. However, our previous study was much smaller and demonstrated such differences in a group that was, presumably, largely untreated with ARVs. In keeping with clinical practice, a neurosyphilis diagnosis was based on an abnormal CSF WBC count or protein concentration when the CSF-VDRL was nonreactive. Indeed, 25 participants only had an elevated CSF WBC count and 17 only had an elevated CSF protein concentration. However, there was no relationship between these 2 single CSF abnormalities and HIV or symptomatic neurosyphilis, and the results of our overall analysis (Table 2) were not changed when we excluded those with just abnormal WBC counts or just abnormal protein concentrations (data not shown). Moreover, the observation that CSF WBC concentrations and CSF-VDRL reactivity normalized after neurosyphilis therapy (Figure 2) adds further support to our contention that our participants had neurosyphilis.

Our results suggest that, in the current ARV treatment era, neurosyphilis treatment outcomes are similar for PenG and APPG-P. Moreover, while HIV may continue to confound neurosyphilis diagnoses with regard to CSF pleocytosis, HIV does not substantially affect the treatment outcome. Given previous shortages of APPG-P, future work should investigate outcomes after alternative neurosyphilis treatment regimens.

Notes

Presented in part: 2017 Conference on Retroviruses and Opportunistic Infections, Seattle, WA, 13–16 February 2017. Abstract number 750.

Financial support. This work was supported by the National Institute of Neurological Disorders and Stroke at the National Institutes of Health (grant number NS34235).

Potential conflicts of interest. C. M. M. has received royalties from Wolters Kluwer, outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.