Study objective
We determine the optimal correction factor for cerebrospinal fluid WBC counts in infants with traumatic lumbar punctures.
Methods
We performed a secondary analysis of a retrospective cohort of infants aged 60 days or younger and with a traumatic lumbar puncture (cerebrospinal fluid RBC count ≥10,000 cells/mm 3 ) at 20 participating centers. Cerebrospinal fluid pleocytosis was defined as a cerebrospinal fluid WBC count greater than or equal to 20 cells/mm 3 for infants aged 28 days or younger and greater than or equal to 10 cells/mm 3 for infants aged 29 to 60 days; bacterial meningitis was defined as growth of pathogenic bacteria from cerebrospinal fluid culture. Using linear regression, we derived a cerebrospinal fluid WBC correction factor and compared the uncorrected with the corrected cerebrospinal fluid WBC count for the detection of bacterial meningitis.
Results
Of the eligible 20,319 lumbar punctures, 2,880 (14%) were traumatic, and 33 of these patients (1.1%) had bacterial meningitis. The derived cerebrospinal fluid RBCs:WBCs ratio was 877:1 (95% confidence interval [CI] 805 to 961:1). Compared with the uncorrected cerebrospinal fluid WBC count, the corrected one had lower sensitivity for bacterial meningitis (88% uncorrected versus 67% corrected; difference 21%; 95% CI 10% to 37%) but resulted in fewer infants with cerebrospinal fluid pleocytosis (78% uncorrected versus 33% corrected; difference 45%; 95% CI 43% to 47%). Cerebrospinal fluid WBC count correction resulted in the misclassification of 7 additional infants with bacterial meningitis, who were misclassified as not having cerebrospinal fluid pleocytosis; only 1 of these infants was older than 28 days.
Conclusion
Correction of the cerebrospinal fluid WBC count substantially reduced the number of infants with cerebrospinal fluid pleocytosis while misclassifying only 1 infant with bacterial meningitis of those aged 29 to 60 days.
Introduction
Background and Importance
The evaluation of febrile infants in the first 2 months of life is challenging because up to 10% of well-appearing infants can have bacterial infections such as meningitis. Consequently, many febrile infants undergo a lumbar puncture as part of their initial clinical evaluation. Of infants who have a lumbar puncture performed, a substantial proportion (15% to 40%) will have a traumatic lumbar puncture. Because introduction of blood into the cerebrospinal fluid complicates the interpretation of cerebrospinal fluid results, infants with traumatic lumbar punctures have higher rates of hospitalization and are more likely to receive parenteral antibiotics than infants with nontraumatic lumbar punctures. Although several methods for correction of the cerebrospinal fluid WBC count for the presence of RBCs have been investigated, these approaches have not been extensively evaluated in the youngest infants and remain controversial.
What is already known on this topic
Traumatic lumbar punctures are common and difficult to interpret in febrile infants. Correction factors have been advocated but not extensively studied.
What question this study addressed
The authors determined an optimal correction factor from a multicenter retrospective cohort of 20,000 infants younger than 60 days and with lumbar punctures performed. Fourteen percent of lumbar punctures were traumatic; 1.1% had bacterial meningitis.
What this study adds to our knowledge
Using a cerebrospinal fluid RBC:WBC correction ratio of 877:1 would reduce from 78% to 33% the proportion of infants receiving a diagnosis of cerebrospinal fluid pleocytosis while missing 7 cases of bacterial meningitis. The simpler ratio 1,000:1 performed similarly.
How this is relevant to clinical practice
A correction factor of 1,000 RBCs:1 WBC in traumatic lumbar punctures should help identify a low-risk group of infants.
Goals of This Investigation
Our primary objective was to determine the optimal correction factor for cerebrospinal fluid WBC counts in infants aged 60 days or younger and with traumatic lumbar punctures. We derived a traumatic lumbar puncture correction factor to adjust the cerebrospinal fluid WBC count for the presence of cerebrospinal fluid RBCs and compared this derived correction factor, as well as other commonly used correction factors, with the uncorrected cerebrospinal fluid WBC count for the detection of bacterial meningitis.
Materials and Methods
Study Design and Setting
We performed a retrospective cohort study of infants who underwent lumbar punctures at 20 participating centers that contributed cerebrospinal fluid cell counts and culture results for analysis. This study was a planned secondary analysis of the Pediatric Emergency Medicine Clinical Research Committee herpes simplex virus study. The parent herpes simplex virus study sought to determine the incidence of herpes simplex virus disease in infants aged 60 days or younger and undergoing evaluation for central nervous system infection. Of the 23 pediatric hospitals included in this parent study, we limited our analysis to the 20 sites that contributed cerebrospinal fluid WBC and RBC counts. The study protocol was approved by each of the participating institutional review boards, with permission for data sharing.
Selection of Participants
Infants were eligible for inclusion in our study if they were aged 60 days or younger, presented to a participating emergency department (ED) between January 1, 2005, and December 31, 2013, had a cerebrospinal fluid culture obtained within 24 hours of ED presentation, and had cerebrospinal fluid cell counts available for analysis. We identified eligible infants with electronic search strategies optimized for each study site. Infants with greater than one eligible ED visit in which a lumbar puncture was performed were included for each eligible encounter. Infants without cell counts available were excluded. We defined a traumatic lumbar puncture as a cerebrospinal fluid RBC count greater than or equal to 10,000 cells/mm 3 to reflect a level of blood contamination that made the cerebrospinal fluid WBC truly difficult to interpret. We categorized lumbar punctures with a cerebrospinal fluid RBC count less than 10,000 cells/mm 3 as nontraumatic.
Methods of Measurement
Data were abstracted electronically at 17 sites and manually at 3. All site abstractors received standardized training through conference calls and biannual in-person meetings. The following data elements were abstracted: demographics (age and sex), visit-level data (study site, date of visit, and disposition), triage temperature, and laboratory data (CBC count with differential, cerebrospinal fluid cell counts, cerebrospinal fluid glucose and protein levels, and cerebrospinal fluid Gram’s stain). We manually abstracted the following microbiology results: bacterial cultures (blood, urine, and cerebrospinal fluid), viral cultures (cerebrospinal fluid), and herpes simplex virus polymerase chain reaction cerebrospinal fluid testing.
Outcome Measures
Our primary outcome was a case of bacterial meningitis, defined as an infant with a cerebrospinal fluid culture result (standard or broth) positive for a pathogenic organism. The following bacterial organisms were classified a priori as contaminants: coagulase-negative Staphylococci, Streptococcus viridans, Propionibacterium acnes, and Corynebacterium species.
Urinary tract infection was defined as greater than or equal to 50,000 colony forming units/ml or 10,000 to 50,000 colony forming units/ml with an abnormal urinalysis (positive nitrite or pyuria). We defined bacteremia as growth of a bacterial pathogen in blood culture. Herpes simplex virus meningoencephalitis and enteroviral meningitis were defined as a positive cerebrospinal fluid herpes simplex virus and enteroviral polymerase chain reaction test, respectively.
Primary Data Analysis
The unit of analysis was the lumbar puncture. Standard descriptive statistics were used to describe infants with and without a traumatic lumbar puncture. The relationship between cerebrospinal fluid RBCs and cerebrospinal fluid WBCs was explored with linear regression, with cerebrospinal fluid RBC count as the independent variable and cerebrospinal fluid WBC count as the dependent variable, using our a priori understanding of the relation between these cerebrospinal fluid parameters. The slope of the regression line (ie, β-coefficient) was used to derive the cerebrospinal fluid WBC correction factor. We then used a generalized estimating equation to adjust for postnatal age and to cluster by hospital center.
We next evaluated the performance of cerebrospinal fluid pleocytosis for the detection of bacterial meningitis in traumatic lumbar punctures. We defined cerebrospinal fluid pleocytosis as a cerebrospinal fluid WBC count greater than 20 cells/mm 3 for infants aged 0 to 28 days and greater than 10 cells/mm 3 for infants aged 29 to 60 days. We evaluated the following 5 approaches to interpretation of the cerebrospinal fluid WBC count: (1) uncorrected cerebrospinal fluid WBC count, (2) the regression-based correction factor, (3) a 1,000:1 cerebrospinal fluid RBCs:cerebrospinal fluid WBCs correction factor, (4) a 500:1 cerebrospinal fluid RBCs:cerebrospinal fluid WBCs correction factor, and (5) a correction factor based on using the peripheral RBC to WBC ratio (cerebrospinal fluid WBC corrected=cerebrospinal fluid WBC–[cerebrospinal fluid RBC×peripheral WBC/peripheral RBC]). The area under the curve for the receiver operator characteristic curve for cerebrospinal fluid WBC count for the prediction of bacterial meningitis was determined for each of these 5 strategies. For purposes of comparison, we assessed the ability of the cerebrospinal fluid WBC count to detect bacterial meningitis in infants with a nontraumatic lumbar puncture. Given the potential for outpatient management for low-risk febrile infants aged 29 to 60 days, we planned a priori to also report the performance of the 5 approaches to cerebrospinal fluid WBC correction in this subgroup.
Last, we calculated the number needed to treat defined as the number of infants without bacterial meningitis who would need to be treated for every misclassified infant with bacterial meningitis if we corrected the cerebrospinal fluid WBC count for the presence of cerebrospinal fluid RBCs. This number represents the number of infants without bacterial meningitis who would need to be treated to avoid one misclassified case of bacterial meningitis, using the uncorrected cerebrospinal fluid WBC count instead of the derived correction factor.
Statistical analyses were performed with SPSS (version 23.0; IBM Corporation, Armonk, NY).
Materials and Methods
Study Design and Setting
We performed a retrospective cohort study of infants who underwent lumbar punctures at 20 participating centers that contributed cerebrospinal fluid cell counts and culture results for analysis. This study was a planned secondary analysis of the Pediatric Emergency Medicine Clinical Research Committee herpes simplex virus study. The parent herpes simplex virus study sought to determine the incidence of herpes simplex virus disease in infants aged 60 days or younger and undergoing evaluation for central nervous system infection. Of the 23 pediatric hospitals included in this parent study, we limited our analysis to the 20 sites that contributed cerebrospinal fluid WBC and RBC counts. The study protocol was approved by each of the participating institutional review boards, with permission for data sharing.
Selection of Participants
Infants were eligible for inclusion in our study if they were aged 60 days or younger, presented to a participating emergency department (ED) between January 1, 2005, and December 31, 2013, had a cerebrospinal fluid culture obtained within 24 hours of ED presentation, and had cerebrospinal fluid cell counts available for analysis. We identified eligible infants with electronic search strategies optimized for each study site. Infants with greater than one eligible ED visit in which a lumbar puncture was performed were included for each eligible encounter. Infants without cell counts available were excluded. We defined a traumatic lumbar puncture as a cerebrospinal fluid RBC count greater than or equal to 10,000 cells/mm 3 to reflect a level of blood contamination that made the cerebrospinal fluid WBC truly difficult to interpret. We categorized lumbar punctures with a cerebrospinal fluid RBC count less than 10,000 cells/mm 3 as nontraumatic.
Methods of Measurement
Data were abstracted electronically at 17 sites and manually at 3. All site abstractors received standardized training through conference calls and biannual in-person meetings. The following data elements were abstracted: demographics (age and sex), visit-level data (study site, date of visit, and disposition), triage temperature, and laboratory data (CBC count with differential, cerebrospinal fluid cell counts, cerebrospinal fluid glucose and protein levels, and cerebrospinal fluid Gram’s stain). We manually abstracted the following microbiology results: bacterial cultures (blood, urine, and cerebrospinal fluid), viral cultures (cerebrospinal fluid), and herpes simplex virus polymerase chain reaction cerebrospinal fluid testing.
Outcome Measures
Our primary outcome was a case of bacterial meningitis, defined as an infant with a cerebrospinal fluid culture result (standard or broth) positive for a pathogenic organism. The following bacterial organisms were classified a priori as contaminants: coagulase-negative Staphylococci, Streptococcus viridans, Propionibacterium acnes, and Corynebacterium species.
Urinary tract infection was defined as greater than or equal to 50,000 colony forming units/ml or 10,000 to 50,000 colony forming units/ml with an abnormal urinalysis (positive nitrite or pyuria). We defined bacteremia as growth of a bacterial pathogen in blood culture. Herpes simplex virus meningoencephalitis and enteroviral meningitis were defined as a positive cerebrospinal fluid herpes simplex virus and enteroviral polymerase chain reaction test, respectively.
Primary Data Analysis
The unit of analysis was the lumbar puncture. Standard descriptive statistics were used to describe infants with and without a traumatic lumbar puncture. The relationship between cerebrospinal fluid RBCs and cerebrospinal fluid WBCs was explored with linear regression, with cerebrospinal fluid RBC count as the independent variable and cerebrospinal fluid WBC count as the dependent variable, using our a priori understanding of the relation between these cerebrospinal fluid parameters. The slope of the regression line (ie, β-coefficient) was used to derive the cerebrospinal fluid WBC correction factor. We then used a generalized estimating equation to adjust for postnatal age and to cluster by hospital center.
We next evaluated the performance of cerebrospinal fluid pleocytosis for the detection of bacterial meningitis in traumatic lumbar punctures. We defined cerebrospinal fluid pleocytosis as a cerebrospinal fluid WBC count greater than 20 cells/mm 3 for infants aged 0 to 28 days and greater than 10 cells/mm 3 for infants aged 29 to 60 days. We evaluated the following 5 approaches to interpretation of the cerebrospinal fluid WBC count: (1) uncorrected cerebrospinal fluid WBC count, (2) the regression-based correction factor, (3) a 1,000:1 cerebrospinal fluid RBCs:cerebrospinal fluid WBCs correction factor, (4) a 500:1 cerebrospinal fluid RBCs:cerebrospinal fluid WBCs correction factor, and (5) a correction factor based on using the peripheral RBC to WBC ratio (cerebrospinal fluid WBC corrected=cerebrospinal fluid WBC–[cerebrospinal fluid RBC×peripheral WBC/peripheral RBC]). The area under the curve for the receiver operator characteristic curve for cerebrospinal fluid WBC count for the prediction of bacterial meningitis was determined for each of these 5 strategies. For purposes of comparison, we assessed the ability of the cerebrospinal fluid WBC count to detect bacterial meningitis in infants with a nontraumatic lumbar puncture. Given the potential for outpatient management for low-risk febrile infants aged 29 to 60 days, we planned a priori to also report the performance of the 5 approaches to cerebrospinal fluid WBC correction in this subgroup.
Last, we calculated the number needed to treat defined as the number of infants without bacterial meningitis who would need to be treated for every misclassified infant with bacterial meningitis if we corrected the cerebrospinal fluid WBC count for the presence of cerebrospinal fluid RBCs. This number represents the number of infants without bacterial meningitis who would need to be treated to avoid one misclassified case of bacterial meningitis, using the uncorrected cerebrospinal fluid WBC count instead of the derived correction factor.
Statistical analyses were performed with SPSS (version 23.0; IBM Corporation, Armonk, NY).
Results
Characteristics of Study Subjects
We identified 23,618 lumbar punctures with cerebrospinal fluid culture obtained, of which 20,319 (86%) had cerebrospinal fluid cell counts available. Of these, 14% (2,880/20,319) were traumatic ( Figure 1 ). Among infants with a traumatic lumbar puncture, 1.1% (33/2,880) had bacterial meningitis. We identified the following bacterial meningitis pathogens in infants with traumatic lumbar punctures: Group B Streptococcus (8), Staphylococcus aureus (8), Escherischia coli (7), Enterococcus species (3), Klebsiella species (2), Listeria (1), Neisseria meningitidis (1), Moraxella catarrhalis (1), Proteus species (1), and Streptococcus pneumoniae (1). Rates of traumatic lumbar puncture varied significantly by study center (range 6% to 23%). Overall, children with a traumatic lumbar puncture were younger and more likely to be hospitalized ( Table 1 ). Infants aged 29 to 60 days with traumatic lumbar punctures had higher rates of hospitalizations than those with nontraumatic lumbar puncture (87.7%; 1,055/1,203 versus 82.0%; 7,137/8,704; difference 5.7%; 95% confidence interval [CI] 3.5% to 7.7%).
| Characteristic | Nontraumatic LP, N=17,439 | Traumatic LP, N=2,880 |
|---|---|---|
| Demographics | ||
| Age, n/N (%), days ∗ | 29.0 (16.0–42.0) | 25.0 (13.0–39.0) |
| 0–28 | 8,628/17,439 (49.5) | 1,665/2,880 (57.8) |
| 29–60 | 8,811/17,439 (50.5) | 1,215/2,880 (42.2) |
| Male infant | 9,813/17,439 (56.3) | 1,578/2,880 (54.8) |
| Clinical | ||
| Triage temperature, °C ∗ | 99.8 (98.7–100.9) | 99.7 (98.7–100.8) |
| CBC WBC count, cells/mm 3 ∗ | 10.4 (7.8–13.5) | 10.7 (7.9-13.6) |
| Admitted, n/N (%) | 15,482/17,232 (89.8) | 2,682/2,847 (94.2) |
| Admitted to ICU, n/N (%) | 2,529/15,482 (16.3) | 394/2,682 (14.7) |
| Length of stay for admitted, days ∗ | 2.2 (1.7–3.0) | 2.2 (1.8–3.0) |
| Microbiology results by age category † | ||
| Urinary tract infection, n/N tested (%), days | ||
| 0–29 | 548/7,911 (6.9) | 90/1,519 (5.9) |
| 29–60 | 702/8,000 (8.8) | 84/1,108 (7.6) |
| Bacteremia, n/N tested (%), days | ||
| 0–29 | 210/8,214 (2.6) | 32/1,566 (2.0) |
| 29–60 | 198/8,318 (2.4) | 30/1,144 (2.6) |
| Bacterial meningitis, n/N (%), days | ||
| 0–29 | 84/8,628 (1.0) | 26/1,665 (1.6) |
| 29–60 | 78/8,811 (0.9) | 7/1,215 (0.6) |
| HSV meningoencephalitis, n/N tested (%), days | ||
| 0–29 | 29/3,532 (0.8) | 3/766 (0.4) |
| 29–60 | 4/1,700 (0.2) | 1/306 (0.3) |
| Enteroviral meningitis, n/N tested (%), days | ||
| 0–29 | 382/1,614 (23.7) | 33/293 (11.3) |
| 29–60 | 440/1,641 (26.8) | 35/209 (16.8) |
Related posts:
Full access? Get Clinical Tree
