Author's response to reviews

Haematogenous Staphylococcus aureus meningitis A 10-year nationwide study of 96 Consecutive Title:Cases

Authors:

Michael Pedersen (m.pedersen@dadlnet.dk)Thomas L Benfield (tlb@dadlnet.dk)Peter Skinhoej (skinhoej@rh.dk)Allan G Jensen (allan.garlik@dadlnet.dk)

2 25 January 2006Version: Date:

see overAuthor's response to reviews:

1

Copenhagen, 2006-01-25

Dear BioMed Central Editorial Team

On behalf of my co-authors, I take the liberty of submitting the enclosed revision of the

manuscript “HAEMATOGENEOUS STAPHYLOCOCCUS AUREUS MENINGITIS. A 10-

year nationwide study of 96 Consecutive Cases”. We have revised the manuscript in

accordance with the reviewers fine and relevant comments and believe that they have

strengthened the manuscript.

All authors have read and approved the revised manuscript, and hope that you will find it

satisfactory.

We are looking forward to hearing from you.

Yours sincerely

Michael Pedersen

Corresponding author:

Dr. Michael Pedersen, Dept. of Infectious Diseases, Copenhagen University Hospital,

Rigshospitalet, Blegdamsvej 9, DK-2100, Denmark. Tel: +45 3545 7726; e-mail:

m.pedersen@dadlnet.dk

2

Point to Point comments

Reviewer: Elmer Villanueva:

Major Compulsory Revisions (that the author must respond to before a decision on

publication can be reached)

1. Case definitions

a. The Authors need to clarify their use of the term “community-acquired”. In the

Background section, they imply that all cases of S. aureus meningitis which is

haematogenously disseminated is community-acquired. How would the following case be

classified – a patient who had neurosurgical repair of an AV malformation but whose

meningitis arises because of underlying endocarditis?

Further on, the Authors further classify non-surgical meningitis into three groups based

timing of diagnosis and regularity of contact with hospitals. One of these groups is

classified as “community-acquired”. If these definitions are standard, the Authors need to

cite a specific reference. All definitions must be face valid.

Reply to 1a:

We agree that the text the text in the Background section could be clarified. The following

paragraph:

“Two different modes of S aureus meningitis pathogenesis have been

considered in the literature: postoperative meningitis i.e. associated with either head

trauma or neurosurgical procedures, or haematogenous dissemination i.e. secondary to

infection outside the central nervous system[1,2,4-6]. Most studies have evaluated both

surgically introduced and spontaneous S aureus meningitis cases[1,2,4-6]. However, there

is a considerable difference between the two groups, e.g. as regards mortality, age and

3

frequency of metastatic infections[1,2,4,7]. To our knowledge there has never been a

review of a large number of haematogenous i.e. community-acquired bacteraemic S

aureus meningitis cases exclusively.”

Has been changed to:

“Two different modes of S aureus meningitis pathogenesis have been

considered in the literature: hospital-acquired infection with a positive blood culture and

evidence of infection later than 48 hours after admission, often associated with either head

trauma or neurosurgical procedures, or a community-acquired infection with a positive

blood culture and evidence of infection less than 48 hours after admission, in the absence

of any recent head trauma or surgical procedures i.e. due to an infection outside the

central nervous system[1,2,4-6]. Most studies have evaluated both surgically introduced

and spontaneous S aureus meningitis cases[1,2,4-6]. However, there is a considerable

difference between the two groups, e.g. as regards mortality, age and frequency of

metastatic infections[1,2,4,7]. To our knowledge there has never been a review of a large

number of haematogenous i.e. community-acquired bacteraemic S aureus meningitis

cases exclusively.”

The reviewer asks how we would classify: a patient who had neurosurgical repair of an AV

malformation but whose meningitis arises because of underlying endocarditis? This would

depend upon different aspects e.g. is the patient still admitted to the hospital? If not, was

the neurosurgical repair performed more or less than four weeks ago?

4

The reviewer finds than we ”… classify non-surgical meningitis into three groups based on

the timing of diagnosis and regularity of contact with hospitals”. However, we classify

meningitis as ”… community, nosocomial or health care associated”, define the

classifications under Materials and Methods in the paragraph: Exclusions, definitions and

clinical data. In order to clarify the definition of community-acquired meningitis the

paragraph:

“Meningitis was classified as community, nosocomial or health care associated. Meningitis

was defined as community acquired, if diagnosed < 48 hr. after hospitalisation in patients

with no regular contact with the hospital. Nosocomial meningitis was defined as a hospital

acquired meningitis > 48 hr. after hospitalisation.”

Has been changed to:

“Meningitis was classified as community, nosocomial or health care associated. Meningitis

was defined as community acquired, if diagnosed with a positive blood culture < 48 hr.

after hospitalisation in patients with no regular contact with the hospital, with clinical

evidence of infection and if the patient had not received surgical intervention within the

recent four weeks. Nosocomial meningitis was defined as a hospital acquired meningitis >

48 hr. after hospitalisation.”

5

b. The section entitled “Selection of cases” only given the general definition of S. aureus

meningitis, not community-acquired meningitis. Presumably, this occurs much later in the

next, in which cases those sentences must be moved to this section. The Authors need to

cite a proper reference for this case definition.

Reply to 1b:

It is a good point. Initially we did not select only the community-acquired meningitis cases,

but all the S aureus meningitis cases. After that we classified the meningitis cases in the

above-mentioned three groups. In order to clarify the definition and oblige comments from

one of the other reviewers we have changed the section and added a relevant reference.

c. The Authors need to be specific about case definitions and exclusions. In the section

headed “Exclusions, definitions and clinical data”, there no exclusions described.

Reply to 1c:

The section headed “Exclusions, definitions and clinical data” is in the Materials and

Methods paragraph, which is why we find it most proper only to define the in- and

exclusion criteria here. We have described the exclusions under Results in the Patient

Characteristics section.

2. Definitions of variables

a. Overall survival or in-hospital death. How was this determined? Was the cause of

death reported? Why was consideration given only to the in-hospital period? This

outcome is not considered anywhere in the text. Patients are only classified as

having survived or died. If in-hospital deaths are not reported, then listing it as an

outcome is superfluous and must be dropped.

6

Reply to 2a:

Whether a patient survived or died was based on the medical records from the department

where the patient was treated. We only registered who died during the admission and who

was discharged. None of the patients were re-admitted to the hospital with meningitis. The

numbers of in-hospital deaths are reported under Results, Patient Characteristics last line.

To clarify this point we have changed the line:

“56% (54/96) of the patients died.”

To:

“56% (54/96) of the patients died during admittance.”

If the patient died, the medical records only described that fact. From the medical records

we could not se whether the non-survivors have had post-mortem autopsy. Consideration

was only given the in-hospital period, as only very few of the patients as far as we could

read in the medical records were referred to the outpatient clinic. The vast majority was

discharged and referred to the individual patients own general practitioner. Therefore, a

follow up was not possible and we chose only to consider the in hospital period.

b. Nursing home residence. What is the definition of a nursing home? How was this

determined? When was this determined? Do the Authors include both the pre-and

post-hospitalisation periods?

Reply to 2b:

We acknowledge our translation and definition is insufficient. Nursing home residence is

changed everywhere in the manuscript to old people’s home, which is defined as a home

outside of the hospital for people over 65 years of age who require full-time care. This was

determined from the medical records and did only include the pre-hospitalisation period.

7

c. Alcohol consumption and tobacco use. How were these determined? Was any

attempt made to consider the subject’s entire exposure experience (eg., pack-years

of smoking).

Reply to 2c:

As the medical records were variable in the information details, we decided to evaluate

whether there were any daily intake/use yes or no for both these categories.

d. Symptoms and objective findings. What is meant by “objective findings”? If they are

clinical signs, are there any of particular interest? At what point in the clinical course

were these determined?

Reply to 2d:

Yes, “objective findings” could be misleading. We do in fact mean clinical signs and have

changed it in the section Exclusions, definitions and clinical data. They were determined at

the point of admission.

e. Temperature, blood pressure, heart rate. These are highly variable vital signs. At

what point in the clinical course were these determined?

Reply to 2e:

We agree that they are highly variable signs. However, values were uniformly taken from

the time of admittance to the hospital. The section definitions and clinical data has been

change from:

“alcohol consumption, tobacco use, symptoms and clinical signs, temperature, blood

pressure, heart rate, “

8

To

“Alcohol consumption, tobacco use, symptoms and clinical signs, temperature, blood

pressure, heart rate, which was uniformly taken from the time of admittance to the

hospital.”

f. Primary and secondary focus. How were these determined?

Reply to 2f:

Primary and secondary focus should be better described. In the section Exclusions,

definitions and clinical data, we have added:

“Primary focus of infection was based on evident clinical signs and/or symptoms that were

later confirmed by cultivation of a bacterial strain with the same resistance pattern as the

blood culture strain. Endocarditis, osteomyelitis, arthritis and meningitis were considered

secondary foci if the patient had not received surgical intervention at these sites prior to

the onset of S aureus bacteremia.”

g. Co-morbidity. What is a definition of a co-morbidity? Who determined that these

were co-morbid factors? Why were specific co-morbidities not of interest? If they

were, how were they determined? In addition, the description of the classifications

used should state that three groups were used, not two.

Reply to 2g:

The authors based on the medical records determined the co-morbidity factors. We

divided co-morbidity in categories of competing illnesses. They were cardiac illness (eg.

previous myocardial infarction), pulmonary illness (eg. COPD), diabetes, alcoholism,

neurological illness (eg. stroke), reumatological illness (eg. arthritis), renal illness (eg. renal

9

insufficiency), gastrointestinal (eg. mb. Crohn), malignancy (eg. cancer) and other

immunosuppression (eg. HIV). Initially we found the intravenous drug users, cardiac illness

and the reumatologic patients of particular interest. In order to clarify the number of groups

we have changed the sentence:

“Co morbidity was registered and for the statistical calculations divided in one or more than

one predisposing condition.”

To

“Co morbidity was registered and for the statistical calculations divided in none, one or

more than one predisposing condition.”

h. Laboratory techniques. All procedures requiring laboratory work requires a

description of the technical method employed. In describing the routine aggregated

laboratory procedures (eg., blood biochemistry), the individual techniques need to

be listed (eg., potassium, sodium, etc.).

Reply to 2h:

We agree and have in the Exclusions, definitions and clinical data section added:

“…and the blood biochemistry variables pH, red and white blood cell count, platelets, CRP,

potassium, sodium, creatinine, carbamide was done.”

i. Cultures. The techniques used to perform cultures need to be described. There is

no mention of blood cultures. Was this performed?

Reply to 2i:

All S aureus strains in Denmark are sent from the 14 departments of microbiology to the

Statens Serum Institut to be phage typed. All strains were found in blood culture(s) from

10

the patient, which was an inclusion criterion. We are unable to describe the techniques at

the individual department, however they absolutely live up international standard.

j. There has been no description of the methods used in determining the

appropriateness or inappropriateness of antibiotic therapy. This requires correcting.

Reply to 2j:

We agree that this needs to be clarified. In the Exclusions, definitions and clinical data

section under Materials and Methods we have added:

“The treatment was considered appropriate if the S aureus strain was found susceptible to

the antimicrobial treatment administered.”

k. The Authors should consider sub-headings for variables sharing specific themes.

Reply to 2k:

We have sub headed the section Exclusions, definitions and clinical data in:

-Definitions and clinical data

-Classification of meningitis cases and exclusions

-Laboratory data

-Severity of illness

-Antimicrobial treatment

We find that this should clarify the text and make definitions easy accessible.

3. Statistical analyses

a. The Authors need to describe whether rates that use the underlying Danish population

as a denominator were adjusted for age and sex and by what method.

11

If rates were not adjusted for age and sex, they should, especially in the context of

temporal analyses. The method of deriving confidence intervals for rates needs to be

described, too.

Reply to 3a:

As the population demographics did not change between 1991-2000 in Denmark, we did

not find it necessary to adjust for age and sex.

b. A more complete description of the regression methods used is required. Which

regression method was used to answer which question? While it is clear in which instance

proportional hazards regression was used, the results for the logistic regression analyses

do not seem to have been reported. The process of model selection is particularly

important.

Reply to 3b, 3c and 3d:

We completely agree with the reviewer and have added the following to the Statistical

analysis section:

“Cox proportional hazard regression analysis was used to estimate risk factors associated

with outcome. The likelihood ratio test was used to test for interaction. There were no

significant interactions. The adequacy of the model was checked by testing the

proportional hazards assumption in different ways: by conducting the traditional graphics

check based on the log of the cumulative hazard and by performing a formal test of

proportionality based on Schoenfeld residuals according to the method of Hosmer and

Lemeshow [10]. Statistical analysis was performed with SPSS 11.5 (Statistical Package for

Social Sciences, Chicago, IL.) and Stata Statistical Software (version 9.0; Stata

Corporation, College Station, TX). P<0.05 on a two-sided test was considered significant.”

12

c. A more complete description of the survival experience of the subjects is required. The

Authors should consider semi-parametric methods in the initial examination of underlying

survival. Cox regression is more correctly specified as Cox proportional hazards

regression.

Reply:

As stated above.

d. The methods of checking the underlying assumptions (eg., independence, proportional

hazards, etc.) of the regression methods need to be described.

Reply:

As stated above.

4. Reporting of results

a. A few of the textual descriptions are presented more concisely in tables. This

duplication needs to be avoided.

Reply to 4a:

We are not in complete agreement with the reviewer. We find the results presented in a

concise and proper manner.

b. A few of the results reported in the text are better placed in tabular format.

Reply to 4b:

As stated immediately above we are not in concurrence with the reviewer, or at least it

would be helpful if the reviewer were more specific.

13

c. All results, regardless of statistical significance, must be reported.

Reply to 4c:

We are uncertain about what exactly the reviewer is referring to?

d. All p-values must be reported in full. We must stamp out the practice of reporting

“p=ns”.

Reply to 4d:

The reviewer has a very good point. We have changed this throughout the manuscript and

Tables.

e. The results for the proportional hazards regression should be reported as hazards

ratios.

Reply to 4e:

We agree, it has been corrected throughout the entire manuscript.

f. When applying the chi-squared test, the Authors only need to report a single p-

value for the entire group, unless several tests were performed sequentially (in

which case, a single global test should suffice).

Reply to 4f:

We agree, it has been corrected in Table 1.

5. Additional issues

14

a. Since there were no underlying hypotheses to be tested, this paper should be seen

as generating hypotheses. In this context, the hunt for statistically significant results

will need to be tempered by corrections for multiple testing.

Reply to 5a:

Yes, this has been the intention.

b. The Authors will need to describe the methodological limitations inherent in their

study.

Reply to 5b:

In Material and Methods we have added the following section:

Limitations of study design

The data were collected retrospectively. Therefore, we were dependent on the data

obtained routinely during hospitalisation from the pre-exiting medical records. These data

were not collected for the specific purpose of the present study, why potential confounding

factors might not have been available from the records. Additionally with regard to the

multivariate analysis the study has relatively small numbers.

Minor Essential Revisions (such as missing labels on figures, or the wrong use of a term,

which the author can be trusted to correct)

1. Abstract:

a. The Authors need to clarify what is meant by “population-adjusted”?

Reply to 1a:

15

With “population-adjusted” we meant the evaluation of whether the 96 cases were a high

number compared to the entire population. In order to avoid misinterpretations we have

deleted “population-adjusted” from the abstract.

b. Results of the proportional hazards regression should be reported as hazard, not

risk, ratios.

Reply to 1b:

We agree and have changed this throughout the entire manuscript.

2. Text:

a. Numerous typographical errors need to be corrected. I do not believe it is my place

to serve as copy-editor for the Authors or for the BMC.

Reply to 2a:

The manuscript has been controlled for spelling and grammatical errors. In addition a

native English speaking person has read it through.

Discretionary Revisions (which the author can choose to ignore)

What next?: Unable to decide on acceptance or rejection until the authors have responded

to the major compulsory revisions

16

Reviewer's report

Title: Community-acquired bacteraemic Staphylococcus aureus meningitis A 10-year

nationwide

study of 96 Consecutive Cases

Version: 1 Date: 29 December 2005

Reviewer: Sally Roberts

Reviewer's report:

General

This paper is a retrospective review of 96 cases of S. aureus meningitis. Cases were

identified from a large database of all referred S. aureus bacteraemia isolates. There may

have been a few patients missed who presented with positive CSF cultures but negative

blood cultures, however, I do not consider this of significance and the title of the paper and

comments by the authors in the text make it clear that the patient group were identitifed by

the presence of bacteraemia. The paper describes the demographic features and clinical

variables of prognostic influence on the outcome. Data collection is limited by the

retrospective nature of the study.

Major Compulsory Revisions (that the author must respond to before a decision on

publication can be reached)

17

The title of the paper is "Community-acquired bacteraemic S. aureus meningitis". The

authors give clear definitions for community-acquired, nosocomial and health care

associated infections. Of the 107 cases identified from the database 11 episodes were

excluded because they occurred post-neurosurgical procedure. Ten were found to be

nosocomially-acquired and 5 healthcare associated. The true number of community-

acquired cases is then only 81 not 96. The authors should present the data related only to

these cases and not all 96, alternatively they should consider changing the title to

"Haematogenous S. aureus meningitis".

Reply:

The initial scope of the work was to describe the haematogenous i.e. the non-surgical S

aureus meningitis cases, as we describe in the end of the Background section. We

believed that to often both surgical and haematogenous cases were described in the same

category, as the epidemiology is different. We agree that the title should be changed,

which is why we change it to:

“Haematogenous S. aureus meningitis.”

The authors describe the secondary foci of infection but give no details about the primary

focus of infection. It would be useful to know if primary infection at particular sites is more

likely to result in meningeal seeding.

Reply:

We could not agree more. We had hoped to be able to make a statement about a relation

between one or more primary focus/foci and the development of meningitis. However, in

the vast majority of cases the primary focus was unknown. The few primary foci noted in

the medical record were skin, pulmonary, teeth and urinary tract infections. Because of the

18

limited numbers of cases with known primary foci we were unable to predict possible

relations or risk foci.

Minor Essential Revisions (such as missing labels on figures, or the wrong use of a term,

which the author can be trusted to correct)

None noted

Discretionary Revisions (which the author can choose to ignore)

1. There was a low Gram stain positivity rate. How does this compare to other

community acquired episodes of meningitis? Was the use of antibiotics pre-lumbar

puncture responsible for this? Did clinicians respond appropriately, in terms of

choice of antibiotics, to the Gram stain result?

Reply to 1:

Of the 55 cases with the information available 15 (27%) had a positive gram stain of the

CSF. In Denmark it is customary to wait until after to lumbar puncture unless time to the

performance of that will be too long. The medical records are inexact as regards the

clinicians response to the gram stain answer. In Denmark these answers are normally

given by telephone through physician-to-physician contact.

2. The mortality rate was high and co-morbidity was associated with this on multivariant

analysis. The co-morbid conditions included in the demographics are no well described in

the paper, it would be useful to have this information.

Reply to 2:

19

We agree and have added this in the Exclusions, definitions and clinical data section:

“Co morbidity was divided in categories of competing illnesses. They were cardiac illness,

pulmonary illness, diabetes, alcoholism, intravenous drug use, neurological illness,

reumatological illness, renal illness, gastrointestinal malignancy and other

immunosuppression.”

3. Why was the inappropriate antibiotic choice not changed in the other 16%?

Reply to 3:

We have asked ourselves the exact same question. But we find no explanations or

indications in the medical records.

What next?: Unable to decide on acceptance or rejection until the authors have responded

to the major compulsory revisions

20

Reviewer's report

Title: Community-acquired bacteraemic Staphylococcus aureus meningitis A 10-year

nationwide

study of 96 Consecutive Cases

Version: 1 Date: 30 December 2005

Reviewer: Philip Hill

Reviewer's report:

General

This paper is from a group well known for their Staphylococcal research and other

infectious diseases publications.

There are a number of spelling mistakes in the document, easily addressed.

Major Compulsory Revisions (that the author must respond to before a decision on

publication canbe reached)

Methods:

The criteria for inclusion needs to be more clear and, for each of the criteria how many

patients were identified early in the results. This is possible to extract from the text, but

should be more explicit. It seems very dubious to include subjects in this study who have

21

only clinical symptoms and signs, especially if they have a relatively normal CSF result.

For example, it appears that individuals with a CSF WCC of 1 were included in this study.

This is dubious. Further it is not clear how those with blood contamination from a bloody

tap were handled-there are clear clinical guidelines as to how to interpret a CSF WCC in

the context of a bloody tap. Those with a bloody tap and a relatively low WCC, then with a

culture positive CSF, could simply have had the organism grown from the contaminating

blood. It is important to understand more clearly how the authors see this issue.

Reply:

We are in agreement with the reviewer; this needs clarification. We have sub headed the

section Exclusions, definitions and clinical data in:

-Definitions and clinical data

-Classification of meningitis cases and exclusions

-Laboratory data

-Severity of illness

-Antimicrobial treatment

We find that this should clarify the text and make definitions easy accessible.

No cases are included solely on their clinical symptoms and signs. There are three cases

with low CSF WCC where one had 1. They were all included because they apart from the

positive blood culture in addition had a positive CSF culture. We find that cultivating the

identical strain in the CSF is more diagnostic that the CSF WCC. In order to clarify this in

the Selection of cases section we have:

“Cases with clinical findings were included if an identical strain was identified in CSF and

blood cultures. “

22

A few of the cases had a bloody lumbar puncture. In these cases the WCC were

subtracted the final CSF WCC result by a red/white cell ratio 1000:1. On the basis of the

data material this showed not to result in any conflicts with our definition of the Selection of

cases.

It should be clearer what constitutes a co-morbidity, since this is an important part of the

results-presumably one of a relatively few possible predisposing conditions: therefore

these should be named clearly.

Reply:

We agree and have added to the Definitions clinical data section:

“Co morbidity was divided in categories of competing illnesses. They were cardiac illness,

pulmonary illness, diabetes, alcoholism, intravenous drug use, neurological illness,

reumatological illness, renal illness, gastrointestinal malignancy and other

immunosuppression. Co morbidity was registered and for the statistical calculations

divided in none, one or more than one predisposing condition.“

Results:

It is of interest that the majority of cases in this study are community acquired. How does

this relate to S. aureus bacteraemia in their setting? Similarly, there are more women than

men: how does this relate to S. aureus bacteraemia and to the authors’ suspicion of under-

reporting overall, as one would expect the opposite in S. aureus invasive infection of any

type? There is a very high death rate, how does this relate to S. aureus bacteraemia in

their setting?

23

Reply:

These are extremely interesting and relevant questions. A large epidemiological study of S

aureus bacteraemia in Denmark including our study period is currently conducted, but has

not been finished. To answer the questions is outside the scope of our analysis, but the

other study should be published later this year.

It is not clear how the ‘73 cases with available information’ relates to the overall results. It

would be useful to have, early on in the results, a description of how many people for each

diagnostic category, and how much missing information was there. Then for each

comparison have n=? in the text and in the tables.

Wrt ‘the treatment was changed in 81’, this is unhelpful. What one really wants to know is

what percentage of people had appropriate antibiotic treatment initially, and of those where

it was inappropriate, what percentage were changed to an appropriate antibiotic regimen

when the diagnosis was made.

Reply:

We believe that the CSF culture results in the 73 available cases should be view as

representative of our overall material, as missing cases were distributed randomly all over

the different departments.

Early under results in the Patient characteristics section we have added:

“Of the remaining 96 patients with haematogenous S. aureus meningitis, 81 cases were

community-acquired, ten cases were nosocomial and five cases health care associated.”

Likewise we have added n=? for all variables and comparisons, both in text and tables, in

order to clarify how much data was available.

24

We are writing how many patients initially received the appropriate treatment, but after

“…treatment was changed in 81” we have added:

“After bacterial diagnosis 41 (77%) of the initial 53 patient receiving an inappropriate

treatment were change to an appropriate treatment.”

Discussion:

How do the authors propose that clinicians search for osteomyelitis in patients with

meningitis due to S. aureus, assuming there maby be no evidence on clinical

examination? The mortality is unchanged compared to previous surveys makes one

wonder what the hospital clinicians are doing to improve their treatment of this problem in

the presence of researchers with a keen established interest in S. aureus disease.

Perhaps the authors could propose a plan to educate staff, especially with respect to early

appropriate treatment?

Reply:

In the hypothetical situation where clinicians search for osteomyelitis in patients with

meningitis due to S. aureus, assuming there may be no evidence on clinical examination,

they have perform a bone scientigraphy.

The mortality is unchanged. It might be a special problem that the disease is

seldom and perhaps overlooked. Additionally our material is limited to the hospitalisation

period only. The period prior to the admission could very well be more important, as we do

not find a difference in mortality between the patients that receive the appropriate and the

inappropriate initial treatment. Therefore, it is a difficult task to propose a better plan for the

staff. A special problem in Denmark is that many departments, according to the local

25

guidelines, administer penicillin as the initial antimicrobial treatment. It might be more

correct with ceftriaxone and ampicillin until microbial diagnosis.

I think it is overstated to say that advanced age is not a risk factor for death. This study

has relatively small numbers for a multivariable analysis and perhaps the statement about

advanced age not being a risk factor should be less dogmatic.

Reply:

We agree that our study has these limitations. Conversely, we have to report the results as

we find them.

Tables:

Table 1

The column headed ‘no’ does not have any information within it. It needs to be very clear

what the ‘n’ is for every comparison.

Reply:

This is a good point and we have corrected Table 1 according to that.

It should be clear what the numbers are eg. Are the numbers presented for age, the

median and range, or mean and range.

Reply:

We believe that with the above-mentioned correction and the legend to Table 1 this should

be clear. The legend states:

“Unless otherwise indicated, data are presented as No. or median (range), ns (non-

significant).”

26

Table 3

The authors should present the p values, not just RR and 95% CI.

Reply:

P values have been added in the Tables.

Minor Essential Revisions (such as missing labels on figures, or the wrong use of a term,

which the author can be trusted to correct)

Discretionary Revisions (which the author can choose to ignore)

The exclusion of those with recent surgery of less than 4 weeks seems too short a time

period to be sure surgery has no relationship to the meningitis. I would have thought 8

weeks or so would be more appropriate, considering the time some abscesses can take to

form before seeding to the CNS. However this may be decided on the basis of a standard

definition-then this needs to be referenced.

Reply:

We agree with the reviewer that our choice of 4 weeks is debateable. The group has used

4 weeks in prior studies. However, in this study no cases were excluded on the basis of

this criterion.

I do not necessarily see how it S. aureus meningitis can be reliably defined as health care

27

associated on the basis of regular visits to an outpatient department at a hospital, when

these subjects have by far the majority of their contacts, in terms of time and number,

outside the hospital. Again if there is a published definition justifying the authors’ position

on this issue, it should be referenced.

Reply:

All the patients in the health-care associated group were renal patients receiving dialysis.

This is not necessarily perform at the hospital department, however in a clinic associated

with the department. In our opinion this justifies the classification.

Overall mortality seems quite high in these patients: compared to 10% in Norgaard et al’s

study (Scand J Infect Dis 2003; 35:311-4) and 27% by Pintado et al (Eur J Clin Micro ID

2002; 864-8). The authors may want to discuss this.

Reply:

This is a relevant comment. However the mentioned mortality rates are overall mortality

rates. Both Nørgaard and Pintado analyse both surgical and non-surgical cases. Nørgaard

has a mortality of 83% (5/6) in community-acquired cases. Pintado finds a mortality rate of

50% for the spontaneous meningitis cases. They both have few cases of community-

acquired/spontaneous meningitis cases compared to our study.

The mortality appears higher for those with high blood glucose, presumably related to

being diabetic. Blood glucose, especially when low, needs to be considered in the light of

the overall clinical condition. Eg. Severely unwell people may have very low blood glucose

measurements as well as only mildly unwell patients. The authors might want to consider

some comment along these lines in addition to the present discussion.

28

Reply:

We believe that, with respect to the limitations of the study design, we ought not comment

further.