Article Text

Download PDFPDF

Effect of delay in analysis on neonatal cerebrospinal fluid parameters
  1. N T Rajesh1,
  2. Sourabh Dutta1,
  3. Rajendra Prasad2,
  4. Anil Narang1
  1. 1
    Department of Pediatrics, Postgraduate Institute of Medical Education and Research (PGIMER), Chandigarh, India
  2. 2
    Department of Biochemistry, Postgraduate Institute of Medical Education and Research (PGIMER), Chandigarh, India
  1. Correspondence to Dr S Dutta, Department of Pediatrics, PGIMER, Chandigarh 160012, India; sourabhdutta{at}


Objectives: The effect of delayed analysis on cerebrospinal fluid (CSF) white blood cell (WBC) count and glucose has never been studied in neonates.

Design: Prospective cohort study.

Setting: Level III newborn unit.

Patients: Neonates undergoing lumbar puncture were enrolled after consent. CSF was analysed at baseline (30 minutes) for protein, WBC and glucose; and from the same sample for WBC and glucose after a lag of 2 h and 4 h after lumbar puncture. Those with traumatic/inadequate CSF were excluded. Subjects were classified in three groups (n  =  20 each) based on baseline WBC count: no WBC, 1–30 WBC and >30 WBC/μl. Analysis was by repeated-measures ANOVA.

Results: There was a significant decline in mean (SD) CSF glucose from baseline to 2 h and 4 h (41.0 (19) to 38.3 (19) and 36.2 (20) mg/dl, respectively) and WBC count (36 (45) to 28.6 (38) and 23.8 (34) cells/μl, respectively; both p<0.001). CSF glucose and WBC declined in all three groups (p<0.001). High baseline CSF WBC (p<0.001) and protein (p<0.001) was associated with a more rapid decline in the levels of CSF WBC, but not glucose. True CSF parameters could be predicted from 4-h parameters: “baseline glucose 5.4 + 0.98 (4-h glucose)” (adjusted R2 97.2%, p<0.001) and “baseline WBC 1.3 (4-h WBC) +0.05 (protein)” (adjusted R2 98.8%, p<0.001). In group 3, a diagnosis of meningitis (based on pleocytosis) would be missed in 52.6% and 78.9% subjects at 2 h and 4 h, respectively.

Conclusions: CSF WBC count and glucose decrease significantly with time. Reliance on WBC counts of delayed samples can result in underdiagnosis.

Statistics from

Bacterial meningitis is most common in the neonatal period.1 As many as 27% cases of culture-positive neonatal sepsis are complicated by meningitis; therefore, it is mandatory to exclude meningitis in all cases of suspected sepsis.2 Although a positive cerebrospinal fluid (CSF) culture is considered to be the gold standard, a rapid diagnosis of meningitis is usually based on cut-off values of CSF glucose, white blood cell (WBC) count and protein.3 Under ideal circumstances, the CSF sample should be transported to the laboratory and analysed immediately. Despite the best of intentions, there are many logistic reasons that may result in a delay in the analysis of CSF. To now, no study in humans (let alone neonates) has systematically evaluated the time-dependent changes that occur in various CSF parameters due to a delay in the processing of the sample. Falsely low CSF WBC counts and glucose levels may alter the diagnosis of neonatal meningitis, with potentially dangerous consequences. There are many differences between the composition of normal neonatal CSF and that in older age groups, the more obvious being higher WBC counts and protein levels and lower glucose levels in neonates. Data generated from older children or adults on this issue may thus not be applicable to neonates. Given this background, we conducted a study with the hypothesis that the glucose levels and WBC counts of neonatal CSF show a significant and predictable decline at 2 h and 4 h after a lumbar puncture.


This was a prospective, repeated measure, cohort study conducted in a level III newborn unit of a tertiary care institute in northern India. Subjects were enrolled from among the neonates (<28 days) who underwent a lumbar puncture as a part of the routine work-up for sepsis. As per unit policy, lumbar puncture was done in all symptomatic neonates with clinically suspected sepsis that merited antibiotics; the exceptions being neonates who were considered too unstable to undergo a lumbar puncture and any local pathology that precluded the performance of lumbar puncture. The exact time of the lumbar puncture was noted. The CSF sample was collected in separate aliquots for biochemistry (10 drops), WBC count (10 drops) and microbiological tests.

The CSF aliquots for biochemical analysis and cell count were immediately transported to the respective laboratories. The first analysis of CSF glucose, protein and WBC count was performed strictly at 30 minutes (± 5) after the lumbar puncture. The 30-minute values were taken as the “baseline” values for the CSF parameters. The persons who analysed the CSF were masked to the identity and clinical details of the subjects and to each other’s findings. From the same CSF aliquot, analysis was again done for glucose and WBC count strictly at 2 h (ie, 120 minutes, ± 15) and 4 h (ie, 240 minutes, ± 15) after collection. Protein estimation was not repeated at 2 h and 4 h because that would have required an unacceptably high volume of CSF sample.

What this study adds

  • There is a significant decline in CSF glucose, and total CSF WBC and neutrophil counts over time.

  • The decline in CSF WBC is related to the baseline WBC count and CSF protein. The decline in CSF glucose is unrelated to baseline WBC count and protein.

The WBC count and differential count for neutrophils was done in an improved Neubauer chamber, using 1% methylene blue for staining the CSF. Glucose in CSF was assayed by a commercially available kit that used the standard glucose oxidase/peroxidase method (Qualigens Fine Chemical, Bombay, India). Special measures for inhibiting glycolysis or special temperatures for storage and transport were deliberately not adopted, so as to mimic routine clinical practice vis-à-vis CSF analysis. The newborn care areas are air conditioned, with temperature maintained at 28°C. The temperature in the laboratory varied from 26 to 27°C.

The following individuals were excluded: those with any red blood cells in the CSF on microscopic examination, those with inadequate CSF sample volume and those whose first analysis could not be done (for logistic reasons) at the stipulated time.

In this study, subjects were consecutively recruited until 20 eligible subjects were included in each of the following groups, based on the WBC count at 30 minutes: group 1:no WBC in CSF; group 2: WBC count 1–30 cells/μl; group 3: WBC count greater than 30 cells/μl.

This was a sample size of convenience. The above groups were chosen to ensure adequate representation from subjects with different ranges of WBC counts. Subjects were recruited after written informed consent was obtained from one of the parents.

The study protocol was submitted to the Institute Ethics Committee of the Postgraduate Institute of Medical Education and Research. The study commenced after obtaining ethics approval. Neonates were enrolled after the nature of the study had been explained to the parents, who had read an information sheet and had provided written informed consent. The study did not entail any additional invasive or painful procedure or the collection of sensitive information. The CSF volume collected was not higher than that collected routinely in this unit, and no extra amount was collected for the study. Pilot observations showed this volume to be sufficient for performing the repeated timed analyses of glucose and WBC count.

Statistical analysis

Demographic data were compared by analysis of variance for normally distributed variables, by Kruskal–Wallis test for skewed distributions and the χ2 test for binomial variables. The WBC count and glucose and protein values were described by descriptive statistics. The temporal differences in the CSF glucose and WBC counts were analysed by one-way repeated measures analysis of variance (RM-ANOVA). The effects of baseline CSF WBC count and protein on the temporal differences were analysed by two-way RM-ANOVA, with protein as a covariate. The χ2 test for linear trends was used to analyse trends in proportions. Statistical analysis was done on SPSS version 10.0.


In the study period, 210 subjects underwent a lumbar puncture, of whom 70 had visible or microtraumatic CSF and were excluded. Only the first 20 subjects to fulfill eligibility criteria in each of the three groups were enrolled (total n  =  60), and the surplus subjects (n  =  80) were not included. The mean birth weight of the study population was 1268 g (SD 410), and the mean gestational age was 32 weeks (SD 3.9). Male infants accounted for 70% of subjects. Demographic data of the groups are compared in table 1. The gestational age, birth weight, age at performing the lumbar puncture, sex distribution and small-for-gestational age status were similar across groups. The 5-minute Apgar scores were statistically different, but no subject had a score of less than 7.

Table 1

Comparison of demographic data

The mean (SD) value of CSF protein in groups 1, 2 and 3 were 32.3 (9.7), 31.6 (10.4) and 98.7 (137.3) mg/dl, respectively. At half an hour the CSF glucose levels in groups 1, 2 and 3 were 41.35 (16.35), 34.55 (11.64) and 45.6 (27.17) mg/dl respectively. At half an hour the CSF WBC counts in groups 2 and 3 were 12 (6.47) and 59.95 (53.1), respectively; and the CSF neutrophil counts were 8.2 (3.9) and 52.5 (49.2), respectively. Group 1, by definition, had no WBC.

One-way RM-ANOVA was performed to study the changes in CSF glucose and WBC count in the full study population (table 2, fig 1).

Figure 1

Change in cerebrospinal fluid (CSF) parameters over time. PMN, polymorphonuclear leucocyte; WBC, white blood cell.

Table 2

Repeated analysis of CSF parameters in the study population

The mean (SD) CSF glucose declined from 41.0 (19.5) mg/dl at half an hour to 38.3 (19.4) mg/dl at 2 h and 36.2 (19.6) mg/dl at 4 h. This within-subject difference was statistically significant (p<0.001). Similarly, the mean (SD) CSF WBC count (per μl) declined from 36 (44.6) at half an hour to 28.6 (38.3) at 2 h and 23.8 (33.8) at 4 h. This within-subject difference was also statistically significant (p<0.001). The mean (SD) CSF neutrophil count (per μl) declined from 30.2 (41.2) at half an hour to 24.4 (35.5) at 2 h and 21.2 (32) per μl at 4 h. This within-subject difference was also statistically significant (p<0.001).

A two way RM-ANOVA was performed to analyse the interaction between the group and the change in CSF glucose with time, with CSF protein as a covariate (table 3). Irrespective of the group there was a significant within-subject decline in CSF glucose (p<0.001; F  =  22.1). There was no relationship between group membership and the magnitude of decline of glucose (p value of interaction 0.26). The relationship between CSF glucose and baseline protein was also not significant (p value of interaction 0.09).

Table 3

Groupwise repeated analysis of CSF glucose (CSF protein as covariate)

A two way RM-ANOVA was performed to analyse the interaction between the group and the temporal change in CSF total WBC and neutrophils, with CSF protein as a covariate (table 4).

Table 4

Groupwise repeated analysis of CSF total WBC count and neutrophil count (CSF protein as covariate)

Irrespective of the group there was a significant within-subject decline both in CSF WBC (p<0.001; F  =  23.3) and neutrophils (p<0.001; F  =  19.9). There was, in addition, a significant association between group membership and the decline in total CSF WBC (p value of interaction <0.001; F  =  22.2) and neutrophils (p value of interaction <0.001; F  =  13.9). The decline of WBC and neutrophils was higher in group 3. There was a significant interaction of the baseline CSF protein with decline in WBC count (p<0.001; F  =  13.1) and decline in neutrophil count (p = 0.005; F  =  5.6).

In group 3, of 20 subjects, 19 had CSF suggestive of bacterial meningitis at baseline, ie, a WBC count above 30 cells/μl and above 60% neutrophils (one subject had a neutrophil count less than 60%). We treated these 19 subjects for meningitis. Three of them had positive blood cultures. Based on the above WBC criteria, a diagnosis of meningitis would be missed in 10/19 (52.6%) subjects if analysed at 2 h and in 15/19 (78.9%) subjects if analysed at 4 h. This change in proportion of subjects that satisfied the WBC criteria of meningitis was significant (p<0.001; χ2 test for linear trends).

Apart from the above 19 subjects who were suspected to have meningitis based on cell counts, 18 others had isolated hypoglycorrhachia (defined as CSF : blood glucose <50% or absolute value of CSF glucose <20 mg/dl). No subject had isolated raised CSF protein (defined as >150 mg/dl in term and >180 mg/dl in preterm neonates). Two subjects had a positive Gram stain (both also had CSF pleocytosis) and no subject had a positive CSF culture. Of the subjects with presumed meningitis, seven had positive blood cultures.

To predict the true CSF WBC count and true CSF glucose from timed samples, linear regression models were generated for the study population (table 5).

Table 5

Multivariate linear regression models for predicting true CSF WBC count and glucose

  1. CSF WBC count at half an hour  =  1.17 × 2-h CSF WBC count + 0.02 × CSF protein (in mg/dl)

  2. CSF WBC count at half an hour  =  1.3 × 4-h CSF WBC count + 0.05 × CSF protein (in mg/dl)

  3. CSF glucose at half an hour (mg/dl)  =  3.03 + 0.99 × 2-h CSF glucose (in mg/dl)

  4. CSF glucose at half an hour (mg/dl)  =  5.4 + 0.98 × 4-h CSF glucose (in mg/dl)

The four models reported above had adjusted R2 values close to 100% and were statistically significant (all p values <0.001). Alternative models (not shown) that included a constant in the equation for predicting CSF WBC and CSF protein in the equation for predicting CSG glucose had non-significant predictors and lower adjusted R2 values; therefore those models were rejected.


The results of the present study have thrown light on the dramatic variation in the CSF parameters that can occur due to a delay in analysis. The magnitude of decline in the CSF glucose from baseline was independent of the baseline WBC count and protein content of the specimen. On the other hand, the magnitude of decline in the CSF WBC depended on the baseline cell count and the protein content.

An exhaustive literature search revealed that this study is the only one of its kind. Steele et al4 had conducted a study on a pooled CSF sample obtained from older children. After leucodepleting the pooled CSF sample, populations of neutrophils and monocytes were added in vitro to the CSF specimen to “adjust” their range artificially from 400 to 1250 cells/μl, and these were analysed after specific time intervals. The authors reported a decline in neutrophil count at 2 h and 4 h by 50% and 58% of the initial concentration, respectively, and they attributed it to the hypotonicity of CSF. The obvious limitation of the study by Steele et al4 was the in-vitro nature of the study and the rarity of such high cell counts in routine clinical practice. In the current study, the neutrophil count declined by 23% at 2 h and by a further 16% between 2 h and 4 h.

Fry et al5 conducted a study on canine CSF. They found that CSF samples with protein concentrations of 50 mg/dl or greater were less susceptible to cellular degeneration than those containing lower protein concentrations. In contrast, in the present study, the deterioration of cells was faster in the presence of higher CSF protein levels. Neonates are known to have very high CSF protein levels; and thus it is not clear what range of protein level is most suited for the survival of WBC.

Interestingly, although subjects in group 1 in this study had no cells in the CSF, they showed a significant decline in CSF glucose, which was comparable to that of the other two groups. This result suggests that the decline in CSF glucose cannot be solely ascribed to its consumption by cellular elements. None of the subjects had red blood cells in the CSF as this was an exclusion criterion. Further studies are required to identify the factors that cause a decline in CSF glucose, so that measures can be adopted to minimise such a decline in routine practice. The percent decline of WBC and neutrophils was significantly more rapid in the first 2 h after performing lumbar puncture, when compared with the next 2 h. One reason for this could be that the decline in WBC is directly related to the baseline value. By 2 h, the baseline value had itself dropped to a lower level, thus slowing down further decline. WBC thus display exponential decay kinetics.

An alarming finding from this study is the high proportion of subjects with elevated CSF WBC counts who could be diagnosed as not having bacterial meningitis if their samples were processed after a delay of a few hours.

The linear regression equations can help the neonatologist in “correcting” for a delay in analysis of CSF. All models derived from this study have very high adjusted R2 values, implying that they are robust and capable of explaining most of the variability in CSF WBC count and CSF glucose. Although these equations do not obviate the need for immediate analysis of CSF, they do give an accurate idea of the true values, provided the timing of performing the lumbar puncture and that of analysis are recorded. The equations can be further simplified to be used at the bedside without complex calculations. For instance, to get the true CSF WBC count one can simply multiply the 4-h count by 1.3 and add to 5% of the protein value; and to get the true CSF glucose one can add 5 to the 4-h CSF glucose.

The presence of WBC in the CSF in itself may contribute to the CSF protein, but this is difficult to quantitate. With modestly elevated WBC counts (as in our study) the contribution may not be much. One of the limitations of our study is that we cannot estimate how much of the increased protein in group 3 is because of leaky capillaries due to culture-negative meningitis and how much due to the presence of cellular protein. This must be kept in mind when interpreting the equations in which protein is a covariate.

The conclusion from this study is that CSF is a labile body fluid, which gets altered in its composition over time. It is strongly recommended that immediate analysis of the CSF specimen should be done to avoid wrong diagnoses. The strength of this study is that it addresses and quantitates the magnitude of a very common, but hitherto uninvestigated, clinical problem. It also provides prediction models as a solution to get around the problem. The conclusions of the study may not be valid for traumatic CSF samples. The results of the present study have raised several questions that need to be answered by further research. The factors influencing the decline in CSF glucose levels and the precise mechanism of interaction between CSF WBC and protein are not clear. By understanding the dynamics of these processes better, it may become easier to make an accurate diagnosis of neonatal meningitis.



  • Contributors: NTR collected the data and wrote the first draft, SD conceived the idea, planned the study, analysed the data and wrote the manuscript, RP performed biochemical analysis and AN supervised the study design and edited the manuscript.

  • Competing interests None.

  • Provenance and Peer review Not commissioned; externally peer reviewed.

  • Ethics approval The study protocol received ethics approval from the Institute Ethics Committee of the Postgraduate Institute of Medical Education and Research.

  • Patient consent Obtained.

  • Contributors: NTR collected the data and wrote the first draft, SD conceived the idea, planned the study, analysed the data and wrote the manuscript, RP performed biochemical analysis and AN supervised the study design and edited the manuscript.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Fantoms
    Ben Stenson