Neurosyphilis: Still prevalent and overlooked in an at risk population

Background Neurosyphilis (NS) presents with a variety of clinical syndromes that can be attributed to other aetiologies due to difficulties in its diagnosis. We reviewed all cases of NS from the “Top End” of the Australian Northern Territory over a ten-year period to assess incidence, clinical and laboratory manifestations. Methods Patient data (2007–2016) were extracted from hospital records, centralised laboratory data and Northern Territory Centre for Disease Control records. Clinical records of patients with clinically suspected NS were reviewed. A diagnosis of NS was made based on the 2014 US CDC criteria. Results were also recategorized based on the 2018 US CDC criteria. Results The population of the “Top End” is 185,570, of whom 26.2% are Indigenous. A positive TPPA was recorded in 3126 individuals. A total of 75 (2.4%) of TPPA positive patients had a lumbar puncture (LP), of whom 25 (35%) were diagnosed with NS (9 definite, 16 probable). Dementia was the most common manifestation (58.3%), followed by epilepsy (16.7%), psychosis (12.5%), tabes dorsalis (12.5%) and meningovascular syphilis (8.3%). 63% of probable NS cases were not treated appropriately due to a negative CSF VDRL. Despite increased specificity of the 2018 US CDC criteria, 70% of patient in the probable NS group were not treated appropriately. The overall annual incidence [95%CI] of NS was 2.47[1.28–4.31] per 100 000py in the Indigenous population and 0.95[0.50–1.62] in the non-Indigenous population (rate ratio = 2.60 [1.19–5.70];p = 0.017). Conclusion Neurosyphilis is frequently reported in the NT, particularly in Indigenous populations. Disturbingly, 60% of probable neurosyphilis patients based on the 2014 criteria, and 70% based on the 2018 criteria with were not treated appropriately. It is critical that clinicians should be aware of the diagnosis of NS and treat patients appropriately.

at those figures. Other studies have either determined NS through chart reviews, referrals to tertiary centers or reporting to their local CDC. These cases have already met criteria for NS. It is unclear from these studies how many patients had an LP for a query of NS but did not met criteria, and therefore not diagnosed with NS.
Is there a take home message for practicing clinicians, given the low rate of appropriate treatment ?
We have added a line to the conclusion to try address this. Line 386 Could you add a line or two telling us what you think the limitations are of your study, please ?
We have added a paragraph explaining the limitations of the study, including the difficulties in making a definitive diagnosis of NS and our reliance on mostly retrospective data. Line 373 The authors estimate incidence rates in their population. While they comment on the risk of overestimate in their discussion due to the inclusion of probable cases, they should also comment on the possibility that these are underestimates (at least using the diagnostic criteria that have been employed). Neurosyphilis is often asymptomatic or not considered in the differential diagnosis. Unless the entire population was screened by anti-treponemal antibody, one really does not know the true prevalence of infection.
We agree that this is an important point. We have now raised the issues in the discussion along with the comments regarding asymptomatic cases. This is on line 280 of the discussion.
The authors state that the incidence of syphilis declined significantly following the introduction of penicillin. While this is true, contact tracing also played a significant role in the decline in incidence in the U.S. as case rates began falling long before the widespread institution of penicillin.
Thank you for this addition and something we have now mentioned with appropriate referencing on line 90 The increased incidence in the indigenous population, presumably individuals in a lower socioeconomic group relative to the white population, parallels the experience in the U.S. where rates of syphilis is substantially higher in the African-American and Hispanic populations.
Yes, this is indeed true, the Australian Indigenous population are individuals who have a considerably lower socioeconomic background compared to Whites, with poorer health outcomes. We have added an additional line regarding this in our discussion (line 313).
Two of their patients were HIV-infected. Can they state what percentage of their patients were men who have sex with men among both the indigenous and nonindigenous populations in their study? Unfortunately, we do not have that data as it was not consistently or clearly documented in the medical records. It was not available to us from the data provided from the CDC either.
Were there not any asymptomatic neurosyphilis cases? If not, why not? Also, can the authors identify whether the neurosyphilis manifestations were early or late manifestations of the infection. Whereas dementia, tabes, and psychosis are typically late manifestations (tertiary syphilis), meningitis, seizures and uveitis are often seen with secondary syphilis, particularly, in the HIV infected population.
There were no cases of asymptomatic NS, which was likely due to our methodology of screening cases based on LPs performed, all of who had neurological symptoms. It also likely relates to the practice of not consistently testing for NS with an LP in patients with syphilis that fail to have a four-fold decrease in RPR despite treatment. There is probably a much higher prevalence of asymptomatic cases that we have failed to detect. This has been added to the discussion (line 282) as well as the results section (line 222). We have categorized cases into early and late manifestations and added it to the results section (line 220), 79% of cases were late manifestations.
How comfortable are the authors in attributing the neurological manifestations to syphilis? After all, "a man can have as many diseases as he damn well pleases." If the clinicians caring for these patients felt comfortable that there were better explanations for their neurological manifestations might that not explain why those with "probable neurosyphilis" were not treated? This is an excellent point that was raised amongst our group. We have added our thoughts regarding this in the discussion (line 340), specifically around the issue of dementia, which is a common condition in the general population. If the clinician felt that the cause of dementia was due to a more common or likely cause then they may not have treated a diagnosis of probable NS. As we do not have in-depth cognitive testing results on our patient cohort, it is difficult for us to say that there was not another dementing syndrome. Further prospective studies are planned to address this.
"the test carries a high sensitivity and low sensitivity" (not "now") line 129 Thank you, correction made How do the authors explain the apparent profound decline in syphilis in the indigenous population between 2007 and 2015 with a nadir in 2013, a year that the incidence in the non-indigenous increased?
The vast majority of cases between 2007-2015 were late latent cases of syphilis in both the Indigenous and non-Indigenous population. These cases were on the decline in the indigenous population, which we believe was in part due to the work by the CDC, though this is difficult to say in retrospect. The small rise in 2013 in the non-indigenous population is in late latent cases as well, though we are not sure why there was a rise. The sharp increase in cases from 2013 onwards are all new cases of syphilis in the Indigenous population, with late latent cases remaining stable in both groups. We have added a table with a breakdown of the raw numbers as a supplementary table for additional clarity.
How does the application of the 2018 CDC criteria change the numbers with respect to the percentage of probably neurosyphilis in which there was a failure to treat?
Thank you for this suggestion as it was not something we had thought to look at. In fact with the 2018 CDC criteria only 3/10 cases received IV BP and 3 cases received IM benzathine. This is less than with the 2014 criteria. We have added these findings to the abstract and discussion section (line 74 & 369 We would like to thank you for reviewing our manuscript for PLOS ONE. The suggestions made by yourself and the reviewer have been addressed and we feel that they have further improved the manuscript. Of note, after re-categorizing cases with the new 2018 US CDC NS criteria, 70% of patients in the probable NS group still did not receive appropriate therapy, indicating significant underdiagnosis and undertreatment despite the improved specificity of the new 2018 criteria. We have tried to highlight the recurring theme of under treatment as one of the take home message for clinicians. We have also added a further supplementary table with raw data on regional syphilis cases over the last 10 years. We assessed the incidence of neurosyphilis (NS) in Northern Australia, where there is currently 34 an outbreak of syphilis. There was a higher incidence of NS in the Indigenous population using Syphilis continues to cause a major burden of disease due to its systemic manifestations and 72 long-term neurological sequelae. After the introduction of contact tracing, followed by penicillin, 73 the incidence of syphilis declined significantly, falling from 447 per 100 000-person years (py) in 74 the United Stated in 1947 to 11.2 per 100 000py in 2000(1, 2). However since the turn of the 75 millennium, the global incidence has increased markedly, more than doubling in at risk 76 populations in North America, Australia and Europe (1,3,4). In 2014 alone there were an 77 estimated 6 million new cases of syphilis (5) the majority occurring in Africa and South East 78 Asia(6). The risk of syphilis is expected to be particularly high in certain ethnic groups, 79 indigenous populations and men having sex with men (7). 80 Many countries have instituted mandatory reporting of all new cases of syphilis, to ensure early 81 detection of outbreaks and to guide patient management and public health interventions. 82 However, neurosyphilis (NS) is not a reportable disease as a separate entity and therefore 83 epidemiological data are sparse. In the antibiotic era, several studies have estimated the 84 incidence of NS to range from 0.08 to 2.2 per 100 000py(8-11). However, the true incidence of 85 NS is difficult to quantify due to frequent misdiagnoses, arising from a paucity of accurate 86 microbiological tests, protean manifestations of the disease and varied clinical diagnostic 87 criteria (12,13). of the ten-year study period was 185,570, of whom 48,632 (26.2%) were Indigenous. The 104 midpoint population over the ten-year period was used for incidence calculations. which has a high sensitivity, allowing the capture of all cases, however, these criteria lack 111 specificity, most notably when serological tests are negative (15). The revised 2018 US CDC NS 112 criteria were published during manuscript preparation and were included into subsequent 113 epidemiological analysis (Table 1)

(16). 114
Clinical syndromes consistent with NS were defined as meningitis with or without cranial nerve 115 abnormalities, meningovascular disease, dementia (general paresis) and tabes dorsalis, and 116 isolated presentations of epilepsy, psychosis, ocular and otic disease. Late manifestations of NS 117 were categorized as dementia, tabes dorsalis and psychosis. Early manifestations of NS were 118 categorized as meningitis, meningovascular disease, ocular disease, otic disease and isolated 6 epilepsy. As per the US CDC 2014 NS criteria, a definite diagnosis of NS was defined as at least 120 one clinical syndrome compatible with NS with a positive serum Treponema pallidum particle 121 agglutination assay (TPPA) test and a positive CSF Venereal Disease Research Laboratory 122 (VDRL). Probable NS was defined as a clinical syndrome suggestive of NS, a positive serum 123 TPPA test, an elevated CSF white cell count (WCC) >5 cell/microL and/or CSF protein >0.45mg/L 124 with no other cause found (Table 2). Possible NS was defined as a patient who was treated for 125 NS due to high suspicion, but who did not meet CDC criteria for lack of a CSF examination, or a 126 diagnosis based on an alternate criterion. The test carries a high sensitivity but low specificity, and is often used to rule out NS when 129 negative. However due to its poor negative predictive value, some argue that it should not be 130 used when a patient's pre-test probability of NS is high (17). We did not use either CSF TPPA or 131 FTA-Ab in our selection criteria as it was not part of the US CDC 2014 criteria. 132 133

Data sources 134
Patients were identified and matched using three separate databases; i) The Northern Territory 135 (NT) Centre for Disease Control notifications of syphilis, including classification (either less than 136 2 years duration or greater than 2 years or unknown duration) and register of prior treatment; ii) 137 Hospital discharge coding from each of the three district hospitals; and iii) laboratory records of 138 syphilis serology testing. 139 Demographic and clinical information from all eligible patients were gathered from the electronic 140 medical records and the clinical classification of NS, its treatment and outcomes recorded in a 141 standardised database using Excel (v16.19,Microsoft). Patients' hardcopy medical records were 142 retrieved when further information was required. Regional population data were obtained from the 143 Top End Health Department data which maintains yearly Indigenous and Non-Indigenous 144 population data for the region. NT CDC data was used for confirmation of treatment history and 7 serological titre response. NT CDC data for new cases of syphilis were divided into three 146 categories during collection: Indigenous, non-Indigenous and unknown. 147 The Top End syphilis diagnostic algorithm uses serum TPPA treponemal testing as the initial test 148 for syphilis and if positive, an RPR is done. CSF for suspected NS patients is analysed by VDRL, 149 TPPA and FTA-A alongside biochemistry analysis and microscopy. 150 151

Statistical analysis 152
All statistical analyses and graphs were generated using Graphpad Prism(v7, San Diego, 153 California). Normally distributed data were compared using Student's t-test, and non-parametric 154 comparisons were made using the Mann-Whitney U test. Proportions were examined using χ 2 155 with Yates' correction or Fisher's exact test. Correlations were assessed using the Pearson test 156 for correlated proportions for normal distributed variables and the Spearman rank test for non-157 normal distributed variables. Rate ratios and 95% confidence intervals were determined using Of the 25 patients with definite or probable NS, 19 (76%) were male, 12 (48%) were Indigenous 183 and 2(8%) were HIV positive. The median age of these patients was 64 years (range: 32 to 97). 184 The overall incidence of definite or probable NS was 1. Clinical data were available for 24 of the 25 patients with definite or probable NS. Dementia was 194 present in 14 (58.3%) of patients, seizures in 4 (16.7%), tabes dorsalis in 3 (12.5%), psychosis 195 in 3 (12.5%), meningovascular disease in 2 (8.3%) and ocular involvement (panuveitis) in 1 196 (4.2%) patient. Three patients had more than one clinical manifestation of NS; Figure 3 and 197 Table 2. There were no cases of meningitis or otic NS. It was not possible to characterise the 198 dementia syndrome or severity from the clinical records. Overall 79% (19/24) cases were late 199 manifestations of NS and 62% (5/8) of definite cases were late manifestations of the disease.  (Table 2). 207 In total 28.0% (7/25) patients with definite or probable NS had a raised WCC in their CSF with a 208 median count of 21 cell/microL (range 7 to 45 cell/microL), which was predominantly 209 lymphocytic. In patients with a definite NS diagnosis WCC was elevated in 55.6% (5/9) of 210 cases, compared to 12.5% (2/16) in patients with a probable diagnosis ; p= 0.144. The CSF 211 protein was elevated in 92% (24/25) of patients (median 0.74 mg/L, range 0.49-1.37mg/L). 212 Two patients with definite NS had borderline CSF results: with a CSF protein of 0.49mg/L and 213 0.5mg/L respectively with no pleocytosis. One patient with definite NS had a normal CSF protein 214 with CSF WCC of 5 cell/microL which should be considered as a normal CSF. Two of these 215 patients had dementia, there were no clinical data available for the third. 216 Serum RPR was reactive in all (9/9) patients with a definite diagnosis of NS and 62.5% (10/16) 217 of those with a probable diagnosis, with a significantly higher titre in those with definite NS; 218 p<0.0001 (Table 2). 219 220

Treatment 221
Prior to a diagnosis of NS, two patients with definite NS had been treated for late latent syphilis 222 with intramuscular benzathine penicillin without a pertinent four-fold decrease in their RPR. Nine 223 (56.3%) of the patients with probable NS had also received prior intramuscular benzathine 224 penicillin for latent syphilis of whom 4 (44.4%) did not have an appropriate four-fold decrease in 225 their RPR titre. This can be attributed to either treatment failure or potential re-infection. Of the 6 226 patients in the probable group who had a non-reactive serum RPR at the time of NS diagnosis, 227 data regarding prior treatment through the NT CDC was available for only 5. Of those 5 patients, 228 3 had no documented prior history of treatment for syphilis 229 All patients with definitive NS were treated with the recommended regimen of 14-15 days of IV 230 benzylpenicillin, but only 37.5% (6/16) of those with probable NS were treated with IV benzyl 231 penicillin. Of the 10 patients who were not treated appropriately, four were treated for late latent 232 syphilis (weekly IM benzathine penicillin for 3 weeks), three of whom had dementia with the only 233 mildly elevated CSF protein (range 0.57-0.74 mg/L). One patient treated for latent syphilis had 234 signs and symptoms of tabes dorsalis and a CSF WCC 21 cell/microL and a negative CSF 235

VDRL. 236
Only 2 of the 25 patients with definite or probable NS had a repeat lumbar puncture, one of 237 whom was a 60-year-old who had presented with psychosis and received appropriate therapy 238 for NS, but subsequently required re-treatment as there was still an elevated CSF VDRL without 239 a 4 fold decrease. 240

Discussion 241
Despite significant public health endeavours, syphilis continues to increase in incidence and 242 exert significant morbidity in at risk populations. Our analysis highlights that between 2007 and 243 2016, the incidence of syphilis in the Top End of Australia was 61.16 per 100,000 py with an 244 associated incidence of NS, based on the 2014 US CDC criteria, of 1.37 per 100,000 py. The 245 rate of syphilis was almost 7-fold higher in the Indigenous population than in the non-Indigenous 246 population and this was associated with a 2.5 fold higher rate of NS. However, the rate of NS 247 per case of syphilis was significantly lower in the Indigenous population in comparison. This is 248 likely due to the stringent follow up and treatment by the NT CDC of Indigenous cases 249 compared to non-Indigenous cases. This has likely limited the duration of syphilis exposure in 250 this community and prevented the development of NS. 251 Since two thirds of the cases were defined as probable diagnoses we may have overestimated 252 the true risk of NS however even when these patients were excluded the incidence of definitive 253 NS in the Top End was 0.49 per 100,000py, approximately 2.5 -5 fold higher than that reported 254 from UK and Denmark. We may also have underestimated the true incidence. Our 255 methodology screened patients based on those who underwent lumbar punctures, all of whom 256 had presented with neurological symptoms. This approach would have failed to capture 257 asymptomatic NS patients, and thus underestimated the true incidence. It is not current practice 258 to perform an LP for patients with syphilis who do not have a four-fold decline in serum RPR 259 despite adequate treatment and this may have missed detection of asymptomatic cases .. Our 260 proposed incidence of NS also assumes that clinicians have considered the diagnosis of NS in 261 all patients presenting with suspicious symptoms and sent appropriate testing. 262

12
In view of the rising number of cases of early syphilis it is likely that NS will increase over the 264 coming decades. Our study is unique in applying almost complete capture of reported cases 265 from the region. It is also the first study to detail the incidence and clinical characteristics of NS 266 in any region of Australia. Similar to previous studies in the US that demonstrate racial 267 disparities with the rates of syphilis between Whites, Blacks and Hispanic persons, our study 268 demonstrates marked ethnic disparities in the incidence of both syphilis and neurosyphilis (7, 269 18), which further undermines the already poor health outcomes and lower life expectancy of 270

Indigenous Australians.(18) 271
The diagnosis of NS is challenging, since there is no gold standard microbiological assay and 272 varying, non-specific diagnostic criteria. Alarmingly more than half of the patients in our study 273 with a probable diagnosis of NS did not receive appropriate treatment. Both false negative and 274 false positive diagnoses contribute to the misdiagnosis of NS. The former arises from the 275 misconception that a negative CSF VDRL has a high negative predictive value for ruling out NS. 276 Whilst CSF VDRL is highly specific, its sensitivity is between (30 and 70%)(19  (14). We conjecture that there is 289 either a reluctance to treat an already very demented patient or ambivalence over the 290 significance of a mildly elevated protein as the only abnormal CSF finding. The clinical 291 experience of many clinicians is that a mildly elevated CSF protein (0.46-0.55 mg/L) as an 292 isolated finding is of little clinical utility. (21) We were unable to extract data on the cognitive 293 profile of our patients with dementia and it is therefore not possible to differentiate between 294 patients with probable NS who truly had dementia related to NS or due to another aetiology. 295 Given the increasing prevalence of dementia in the general community, the treating clinicians 296 may have decided not to treat for probable NS as another cause of dementia was more likely. 297 During the preparation of the manuscript, the US CDC released revised diagnostic criteria which 298 included an additional requirement that a reactive serum RPR result was needed for a diagnosis 299 of definite and probable NS. A four-fold decrease in RPR titres suggests a treatment response 300 in syphilis, however RPR titres can fall and normalise even without treatment. Previous studies 301 have demonstrated that in patients with definite NS, the serum RPR is almost always reactive, 302 but approximately 10% of patients with probable NS have a non-reactive serum RPR(22). In our 303 cohort all patients with definite NS had a reactive serum RPR, but over a third of patients with 304 probable NS group did not have a reactive serum, of whom three had no history of prior syphilis 305 treatment and only had an elevated CSF protein. According to more stringent criteria these 306 serum RPR negative patients would be considered false positives. Although some of these 307 patients may have been correctly categorized as not having NS, the diagnosis in those who 308 have never received any form of prior syphilis treatment is unclearthey could have potentially 309 Our study has several limitations . Although we attempted to capture all cases of syphilis and 321 NS searching multiple databases, given the uncertainty of the diagnosis and our reliance on LP 322 and clinical suspicion, we may have either over-or under-estimated the true burden of disease. 323 Furthermore whilst we included prospective data collection the majority of the data were 324 gathered retrospectively , and we were unable to document detailed phenotypes of many of the 325 patients from the medical notes. 326 327

Conclusion 328
Our study highlights the ongoing difficulties with the diagnosis and management of NS. Despite 329 these challenges, there continues to be significant disparities between Indigenous and non-330 Indigenous Australians in the incidence of syphilis and NS. Similar to other studies, the key 331 dilemma is whether patients with probable NS truly have the disease and would benefit from 332 appropriate treatment (23). Until the advent of a new and accurate assay, NS will continue to be 333      There is currently a syphilis outbreak in Northern Australia. We assessed the incidence of may be attributed to other aetiologies due to difficulties in its diagnosis. We reviewed all cases 52 of NS from the "Top End" of the Australian Northern Territory over a ten-year period to assess 53 incidence, clinical and laboratory manifestations. The overall incidence of definite or probable NS was 1.

Clinical Manifestations 202
Clinical data were available for 24 of the 25 patients with definite or probable NS. Dementia was 203 present in 14 (58.3%) of patients, seizures in 4 (16.7%), tabes dorsalis in 3 (12.5%), psychosis 204 in 3 (12.5%), meningovascular disease in 2 (8.3%) and ocular involvement (panuveitis) in 1 205 (4.2%) patient. Three patients had more than one clinical manifestation of NS; Figure 3 and 206 Table 2. There were no cases of meningitis or otic NS. It was not possible to characterise the 207 dementia syndrome or severity from the clinical records. Overall 19/24 79% (19/24) cases were 208 late manifestations of NS and . 5/8 (62% (5/8) of definite cases were late manifestations of the 209 disease. There were nNo patients with cases of asymptomatic NS were identified.  (Table 2). 216 In total 28.0% (7/25) patients with definite or probable NS had a raised WCC in their CSF with a 217 median count of 21 cell/microL (range 7 to 45 cell/microL), which was predominantly 218 lymphocytic. In patients with a definite NS diagnosis WCC was elevated in 55.6% (5/9) of 219 cases, compared to 12.5% (2/16) in patients with a probable diagnosis ; p= 0.144. The CSF 220 protein was elevated in 92% (24/25) of patients (median 0.74 mg/L, range 0.49-1.37mg/L). 221 Two patients with definite NS had borderline CSF results: with a CSF protein of 0.49mg/L and 222 0.5mg/L respectively with no pleocytosis. One patient with definite NS had a normal CSF protein 223 with CSF WCC of 5 cell/microL which should be considered as a normal CSF. Two of these 224 patients had dementia, there were no clinical data available for the third. 225 Serum RPR was reactive in all (9/9) patients with a definite diagnosis of NS and 62.5% (10/16) 226 of those with a probable diagnosis, with a significantly higher titre in those with definite NS; 227 p<0.0001 (Table 2). 228 229

Treatment 230
Prior to a diagnosis of NS, two patients with definite NS had been treated for late latent syphilis 231 with intramuscular benzathine penicillin without a pertinent four-fold decrease in their RPR. Nine 232 (56.3%) of the patients with probable NS had also received prior intramuscular benzathine 233 penicillin for latent syphilis of whom 4 (44.4%) did not have an appropriate four-fold decrease in 234 their RPR titre. This can be attributed to either treatment failure or potential re-infection. Of the 6 235 patients in the probable group who had a non-reactive serum RPR at the time of NS diagnosis, 236 data regarding prior treatment through the NT CDC was available for only 5. Of those 5 patients, 237 3 had no documented prior history of treatment for syphilis 238 All patients with definitive NS were treated with the recommended regimen of 14-15 days of IV 239 benzylpenicillin, but only 37.5% (6/16) of those with probable NS were treated with IV benzyl 240 penicillin. Of the 10 patients who were not treated appropriately, four were treated for late latent 241 syphilis (weekly IM benzathine penicillin for 3 weeks), three of whom had dementia with the only 242 mildly elevated CSF protein (range 0.57-0.74 mg/L). One patient treated for latent syphilis had 243 signs and symptoms of tabes dorsalis and a CSF WCC 21 cell/microL and a negative CSF 244

VDRL. 245
Only 2 of the 25 patients with definite or probable NS had a repeat lumbar puncture, one of 246 whom was a 60-year-old who had presented with psychosis and received appropriate therapy 247 for NS, but subsequently required re-treatment as there was still an elevated CSF VDRL without 248 a 4 fold decrease. 249

Discussion 250
Despite significant public health endeavours, syphilis continues to increase in incidence and 251 exert significant morbidity in at risk populations. Our analysis highlights that between 2007 and 252 2016, the incidence of syphilis in the Top End of Australia was 61.16 per 100,000 py with an 253 associated incidence of NS, based on the 2014 US CDC criteria, of 1.37 per 100,000 py. The 254 rate of syphilis was almost 7-fold higher in the Indigenous population than in the non-Indigenous 255 population and this was associated with a 2.5 fold higher rate of NS. However, the rate of NS 256 per case of syphilis was significantly lower in the Indigenous population in comparison. This is 257 likely due to the stringent follow up and treatment by the NT CDC of Indigenous cases 258 compared to non-Indigenous cases. This has likely limited the duration of syphilis exposure in 259 this community and prevented the development of NS. 260 Since two thirds of the cases were defined as probable diagnoses we may have overestimated 261 the true risk of NS however even when these patients were excluded the incidence of definitive 262 NS in the Top End was 0.49 per 100,000py, approximately 2.5 -5 fold higher than that reported 263 from UK and Denmark . However wWe may also There is also a risk that we have 12 underestimated the true incidence. Our methodology screened for patients based on those who 265 underwent received lumbar punctures, all of whom had presented with neurological symptoms. 266 This approach method would have faileds to capture asymptomatic NS patients, and thus 267 understimatedunderestimated the true incidencewhich would have led to higher incidence 268 numbers. It is not current practice to perform obtain an LP for patients with syphilis who do not 269 have fail to have an appropriate a four-fold decline in serum RPR despite adequate treatment 270 and this may have therefore our missed capture of detection of asymptomatic cases may not 271 be complete. Unless all cases of syphilis are tested for NS, regardless of neurological 272 symptoms, we would not be able to truly gauge the real incidence of the disease. Our proposed 273 incidence of NS also assumes that clinicians have considered correctly queried the diagnosis of 274 NS in all patients presenting with suspicious symptoms and sent appropriate testing, which may 275 not always be the case. 276

277
In view of the rising number of cases of early syphilis it is likely that NS will increase over the 278 coming decades. Our study is unique in applying almost complete capture of reported cases 279 from the region. It is also the first study to detail the incidence and clinical characteristics of NS 280 in any region of Australia. Similar to previous studies in the US that demonstrate racial 281 disparities with the rates of syphilis between Whites, Blacks and Hispanic persons, our study 282 demonstrates marked ethnic disparities in the incidence of both syphilis and neurosyphilis (7, 283 18), which further undermines the already poor health outcomes and lower life expectancy of 284

Indigenous Australians.(18) 285
The diagnosis of NS is challenging, since there is no gold standard microbiological assay and 286 varying, non-specific diagnostic criteria. Alarmingly more than half of the patients in our study 287 with a probable diagnosis of NS did not receive appropriate treatment. Both false negative and 288 false positive diagnoses contribute to the misdiagnosis of NS. The former arises from the 289  (14). We conjecture that there is 303 either a reluctance to treat an already very demented patient or ambivalence over the 304 significance of a mildly elevated protein as the only abnormal CSF finding. The clinical 305 experience of many clinicians is that a mildly elevated CSF protein (0.46-0.55 mg/L) as an 306 isolated finding is of little clinical utility.(21) We were unable to extract did not have data on the 307 cognitive profile of our patients with dementia and it is therefore not possible difficult to clearly to 308 differentiate between patients with in the probable NS group who truly had dementia related to 309 NS or due to another aetiologycommon cause of dementia. Given the increasing prevalence of 310 dementia in the general community, the treating clinicians may have decided not to treat for 311 probable NS as another cause of dementia was more likely. 312 During the preparation of the manuscript, the US CDC released revised diagnostic criteria which 313 included an additional requirement that a reactive serum RPR result was needed for a diagnosis 314   Footnote: The percentage of Probable cases decreases from 64% to 40% respectively. The 476 Category of not NS subsequently increases to 24% respectively. 477 The rate of lumbar puncture was 2.4% -do you know how this compares with other countries ? Thank you for this excellent suggestion. Unfortunately, it is not clear after reviewing the literature what LP rates were in other countries. As we one of our methods of screening for NS patients was by reviewing all LP performed in the region we were able to arrive at those figures. Other studies have either determined NS through chart reviews, referrals to tertiary centers or reporting to their local CDC. These cases have already met criteria for NS. It is unclear from these studies how many patients had an LP for a query of NS but did not met criteria, and therefore not diagnosed with NS.
Is there a take home message for practicing clinicians, given the low rate of appropriate treatment ?
We have added a line to the conclusion to try address this. Line 386 Could you add a line or two telling us what you think the limitations are of your study, please ?
We have added a paragraph explaining the limitations of the study, including the difficulties in making a definitive diagnosis of NS and our reliance on mostly retrospective data. Line 373 The authors estimate incidence rates in their population. While they comment on the risk of overestimate in their discussion due to the inclusion of probable cases, they should also comment on the possibility that these are underestimates (at least using the diagnostic criteria that have been employed). Neurosyphilis is often asymptomatic or not considered in the differential diagnosis. Unless the entire population was screened by anti-treponemal antibody, one really does not know the true prevalence of infection.
We agree that this is an important point. We have now raised the issues in the discussion along with the comments regarding asymptomatic cases. This is on line 280 of the discussion.
The authors state that the incidence of syphilis declined significantly following the introduction of penicillin. While this is true, contact tracing also played a significant role in the decline in incidence in the U.S. as case rates began falling long before the widespread institution of penicillin.
Thank you for this addition and something we have now mentioned with appropriate referencing on line 90 The increased incidence in the indigenous population, presumably individuals in a lower socioeconomic group relative to the white population, parallels the experience in the U.S. where rates of syphilis is substantially higher in the African-American and Hispanic populations.

Response to Reviewers
Yes, this is indeed true, the Australian Indigenous population are individuals who have a considerably lower socioeconomic background compared to Whites, with poorer health outcomes. We have added an additional line regarding this in our discussion (line 313).
Two of their patients were HIV-infected. Can they state what percentage of their patients were men who have sex with men among both the indigenous and non-indigenous populations in their study?
Unfortunately, we do not have that data as it was not consistently or clearly documented in the medical records. It was not available to us from the data provided from the CDC either.
Were there not any asymptomatic neurosyphilis cases? If not, why not? Also, can the authors identify whether the neurosyphilis manifestations were early or late manifestations of the infection. Whereas dementia, tabes, and psychosis are typically late manifestations (tertiary syphilis), meningitis, seizures and uveitis are often seen with secondary syphilis, particularly, in the HIV infected population.
There were no cases of asymptomatic NS, which was likely due to our methodology of screening cases based on LPs performed, all of who had neurological symptoms. It also likely relates to the practice of not consistently testing for NS with an LP in patients with syphilis that fail to have a four-fold decrease in RPR despite treatment. There is probably a much higher prevalence of asymptomatic cases that we have failed to detect. This has been added to the discussion (line 282) as well as the results section (line 222). We have categorized cases into early and late manifestations and added it to the results section (line 220), 79% of cases were late manifestations.
How comfortable are the authors in attributing the neurological manifestations to syphilis? After all, "a man can have as many diseases as he damn well pleases." If the clinicians caring for these patients felt comfortable that there were better explanations for their neurological manifestations might that not explain why those with "probable neurosyphilis" were not treated? This is an excellent point that was raised amongst our group. We have added our thoughts regarding this in the discussion (line 340), specifically around the issue of dementia, which is a common condition in the general population. If the clinician felt that the cause of dementia was due to a more common or likely cause then they may not have treated a diagnosis of probable NS. As we do not have in-depth cognitive testing results on our patient cohort, it is difficult for us to say that there was not another dementing syndrome. Further prospective studies are planned to address this.
"the test carries a high sensitivity and low sensitivity" (not "now") line 129 Thank you, correction made How do the authors explain the apparent profound decline in syphilis in the indigenous population between 2007 and 2015 with a nadir in 2013, a year that the incidence in the nonindigenous increased?
The vast majority of cases between 2007-2015 were late latent cases of syphilis in both the Indigenous and non-Indigenous population. These cases were on the decline in the indigenous population, which we believe was in part due to the work by the CDC, though this is difficult to say in retrospect. The small rise in 2013 in the non-indigenous population is in late latent cases as well, though we are not sure why there was a rise. The sharp increase in cases from 2013 onwards are all new cases of syphilis in the Indigenous population, with late latent cases remaining stable in both groups. We have added a table with a breakdown of the raw numbers as a supplementary table for additional clarity.
How does the application of the 2018 CDC criteria change the numbers with respect to the percentage of probably neurosyphilis in which there was a failure to treat?
Thank you for this suggestion as it was not something we had thought to look at. In fact with the 2018 CDC criteria only 3/10 cases received IV BP and 3 cases received IM benzathine. This is less than with the 2014 criteria. We have added these findings to the abstract and discussion section (line 74 & 369).