As in most bacterial infections, leucocytosis (>11000-12000leucocytes/mm3) is frequently detected in acute BM.23,24 Classic studies associate higher levels of peripheral leucocytes with marked neutrophila with infections due to Streptococcus pneumoniae.25 The mean peripheral leucocyte count in children with BM is approximately 17000leucocytes/mm3 (values vary by study), as opposed to a mean of 9100leucocytes/mm3 measured in VM, with a statistical significance of P<.001.26 A recent Spanish study in adults27 found mean values of 17500leucocytes/mm3 for BM and 11096leucocytes/mm3 for VM, which translates into an area under the ROC curve (AUC) of 0.756 (P=.003). Nevertheless, we also found studies reporting leucocyte counts below the leucocytosis range, even in BM.28 In both cases, this was the blood test with the lowest AUC for distinguishing between BM and VM, as confirmed by a recent article.29 One study even reports that blood leucocyte count in elderly patients (>75 years) does not reveal statistically significant differences for distinguishing between BM and VM.11 Therefore, blood leucocyte count is confirmed to have the lowest predictive power for BM, below that of C-reactive protein (CRP) and PCT.20 However, it should be noted that leucocyte count (>12000leucocytes/mm3 or <4000leucocytes/mm3) is one of the 4 diagnostic criteria for systemic inflammatory response syndrome and therefore, for sepsis, with the associated prognostic implications.23,24

Biomarkers are defined as measurable molecules in a biological sample which may be used as an indicator of the status (normal or pathological) of a biological process, allowing monitoring of the process based on concentration of the biomarker.16 Increased, decreased, or stable biomarker values reflect the progression of the disease and the response to treatment.16,30

To date, many studies have published data on different BII involved in diagnosing AM and establishing aetiology and prognosis.16,20,30 We describe below the most relevant biomarkers used in everyday clinical practice, divided as measured in the blood (in this section) or in the CSF (in a subsequent section).

CRP is an acute-phase protein synthesised in the liver in response to interleukin-6 (IL-6) and IL-8, and increases both in infectious (viral and bacterial) and inflammatory (chronic and acute) processes.16,17,30

Synthesis starts in the first 4 to 6hours following process onset, although peak synthesis is recorded at approximately 36 to 48hours. CRP levels may remain elevated for days despite correct treatment and improvement of the patient's condition.16,17,30 However, detection is easy, reproducible, and inexpensive. Normal values (0-8mg/L) depend on age, sex, and race.16,31

Several studies report increased values in the context of bacterial infections,32 including BM; however, its poor kinetics and the increase observed in the context of other inflammatory and viral processes limit its usefulness as an independent BII, especially in elderly patients (diagnostic usefulness decreases with age) and in patients with oncohaematological or autoimmune disease or cirrhosis, among others.11,16 In any case, most of the studies place CRP above leucocyte count for predicting BM vs VM.11,16,27 Before the use of PCT measurement was generalised, CRP was the most widely studied BIIs in cases of BM.33

In patients with negative CSF cultures, CRP achieves a sensitivity of 86% and a specificity of 84%, with a cut-off point (CP) of 37mg/L for distinguishing between BM and VM.11 Another recent study reports an AUC of 0.916 with a 95% confidence interval (95% CI) of 0.838 to 0.994 (P<.001); despite this diagnostic power being lower than that of PCT, it is higher than that of leucocyte count.27

CRP remains a stable and reliable biomarker, performing similarly to PCT in paediatric populations, especially in cases with isolated S. pneumoniae or Neisseria meningitidis.34 However, other studies report a poorer ability to distinguish between VM and BM and demonstrate that concentrations overlap with those observed in non-infectious aetiologies, reducing the specificity of this parameter for distinguishing between types of AM.35 In adults older than 75, increased CRP values in cases of BM are not statistically significant for distinguishing between VM and BM, with an AUC of only 0.514 and a specificity of 43%.11

Therefore, we should use and interpret CRP values with caution in elderly patients with such severe processes as suspected BM. Over 50% of patients with fever in EDs undergo CRP tests (but not PCT tests)19.36 to distinguish between viral and bacterial infections, and more than 40% of all infectious processes in EDs affect elderly patients.3 We must therefore be aware of the diagnostic limitations of CRP measurements for confirming or ruling out bacterial aetiology of AM in this population.37,38

Procalcitonin, a polypeptide precursor of calcitonin, is a protein containing 116 amino acids, and is mainly synthesised in the thyroid gland and the lungs (Kultschitzky or neuroendocrine cells). PCT remains almost undetectable in healthy subjects in normal conditions; values below 0.05ng/mL are considered normal.16,17,39 Many tissues have been observed to produce PCT during bacterial infections and sepsis in response to the stimulus of tumour necrosis factor α, IL-6, and IL-8, after recognition of bacterial components (such as lipopolysaccharides in gram-negative cells or lipoteichoic acid in gram-positive cells).16,17 Increases in PCT concentration, which depend directly on bacterial load and/or presence of endotoxins, are measurable within just 3 to 4hours of BM onset; peak values are observed at approximately 12hours and half-life is 20 to 36hours.16,40 Decreasing PCT values, on the other hand, may confirm positive response to an appropriate antimicrobial treatment and a good progression at 12 to 24hours. This particular kinetic behaviour is very useful for emergency decisions when there is initial suspicion of BM. This, together with the excellent diagnostic power of this biomarker (higher than that of leucocytes and PCR) and the fact that its predictive ability is maintained in elderly patients,11,38 oncohaematological and neutropenic patients,42 and those with kidney failure,41 cirrhosis, or autoimmune diseases,43 have made it the ideal serum BII for aetiological diagnosis, prognostic assessment, and management of patients with BM.11,16,18,20,27,29 Previous studies describe and validate this role of PCT in other severe infectious processes (sepsis,16,23,44 bacteraemia,16,45 severe pneumonia,46,47 etc.).

Early studies, conducted more than 10 years ago, already suggested that the diagnostic power of PCT was significantly higher than that of leucocyte count or PCR concentration, although the small sample sizes used and the study methodologies limit the reliability of these results.48,49 However, there is controversy regarding the optimal CP, which varies greatly from study to study (mainly due to differences in sample size and the technique used for measuring PCT), ranging from 0.23ng/mL to 5ng/mL (to obtain sensitivity and specificity values above 90% in all studies).11,16,18,20,27,29,39 According to one of the most relevant studies, conducted by Viallon et al.18 and including 254 patients with AM (35 with BM and 181 with VM), a CP≥0.28ng/mL for PCT provided the greatest diagnostic power, achieving a sensitivity of 95%, a specificity of 100%, a positive predictive value (PPV) of 100%, and a negative predictive value (NPV) of 97%, with an AUC of 0.99 (95% CI, 0.99-1). These results are very similar to those obtained by 2 more recent studies, which reported higher CPs (PCT≥0.8 and 0.74ng/mL, respectively) obtaining similar sensitivity and specificity. In a very recent meta-analysis including 9 studies and 725 patients, Vikse et al.20 confirmed these excellent results, obtaining a sensitivity of 90% (95% CI, 84%-94%), a specificity of 98% (95% CI, 97%-99%,), and an odds ratio (OR) of 287 (95% CI, 55-1409). In the light of these results, ED physicians should regard initial PCT levels >0.25 to 0.5ng/mL as potentially diagnostic of BM.16,20 The necessary microbiology tests should be performed8 and appropriate antibiotic treatment should be started immediately.5

By age group, PCT diagnostic power is higher than that of CRP in children,11,34,50 adults, and elderly patients,11,20 and is not affected by the limitations of CRP.

In addition to being of great help in aetiological diagnosis, some studies report the use of PCT to monitor the progression of infection, since it exponentially decreases after the first 24 to 48hours in patients receiving an appropriate antibiotic treatment.51,52 Furthermore, PCT concentrations higher than 1ng/mL are suggestive of bacteraemia,16,27 with the associated implications for prognosis.

Blood lactate is considered the best marker of tissue hypoperfusion and hypoxia, and is included in all assessment recommendations for patients with sepsis, severe sepsis, and septic shock in EDs.23,24 However, it has no capacity to distinguish between bacterial and viral aetiology or other causes of non-infectious systemic inflammatory response syndrome. Therefore, determination of blood lactate concentration is recommended in AM as a prognostic factor of severity, mortality, and treatment response.16,17,30 Patients with any infectious process showing blood lactate levels >2 to 2.5mmol/L should be more intensively monitored since they present higher mortality rates during admission and at 30 days of presentation,16,53 even in situations of haemodynamic stability with or without bacteraemia.54,55 This excellent prognostic power for severity and mortality was also confirmed in elderly patients.56

Normal CSF leucocyte count is <5 to 10leucocytes/mm3, comprising mainly mononuclear cells; in newborns, levels may reach 30leucocytes/mm3. Pleocytosis (100-10000leucocytes/mm3) occurs in the vast majority of cases of BM with predominance of PMN, especially after the first 24hours from bacterial infection.1,59 Its diagnostic power for identifying bacterial aetiology has been widely studied for a number of years. Several CPs have been established; one study reports a CP of 118leucocytes/mm3, with a sensitivity of 80% and a specificity of 85% to identify BM,11 although the majority of authors consider it necessary to increase this CP to a minimum of 300leucocytes/mm3 to consistently obtain sensitivity and specificity values above 80%.1,28,58,59 In any case, and regardless of the CP used, the total number of CSF leucocytes presents lower diagnostic power and AUCs compared with the remaining CSF parameters analysed.11,18,29,59 Furthermore, it should be noted that this occurs at all ages11 (especially in newborns, elderly and immunocompromised patients, in whom its discriminant ability for BM decreases).59,60 Previous antibiotic treatment is known to interfere with CSF leucocyte count, causing them to vary greatly.59 Low CSF leucocyte count is associated with a poorer prognosis in patients with BM.61 There are no specific differences to be taken into account with regards to elderly patients.11 A recent study in children (≤14 years) establishes an optimal CP of 321leucocytes/mm3, with a sensitivity of 80.6% and a specificity of 81.4% for distinguishing between BM and VM.62

PMN account for >50% of total leucocyte count in acute BM cases.1,58,59 A recent study reports that figures >60% would achieve sensitivity and specificity values of 69% and 77%, respectively, with a PPV of 89%.63 Decreased CSF glucose and lactate are statistically significant predictors of BM vs VM, as is a PMN percentage >50%, with an OR of 20.19 (95% CI, 8.31-49.09; P=.002).29

Another classically studied CSF parameter is protein level. Levels should not exceed 45 to 50mg/dL in healthy individuals.1,58 They are slightly elevated in cases of VM (mean of 56mg/dL) and clearly higher in BM (mean of approximately 135mg/dL).63 According to Viallon et al.,18 protein levels above 188mg/dL are associated with sensitivity and specificity of 87% and 93%, respectively, a PPV of 67%, and an AUC of 0.93 for distinguishing between BM and VM. However, the great majority of recent studies in children, adults, and elderly patients report a lower predictive ability and discrete AUC in comparison with the percentage of PMN, glucose CSF level, or lactate concentration.11,27,29,50,59,61,63

Decreased CSF glucose concentration is caused by bacterial metabolism, and is a typical finding in cases of BM.1 It is also dependent on simultaneously measured blood glucose level, and has been shown to represent 60% to 80% of that value under normal circumstances.1,59 Therefore, we assume that CSF should contain >40 to 50mg/dL of glucose. Thus, in presence of normal blood glucose levels, it has been suggested that a CP<40mg/dL of glucose would provide a sensitivity of 97% and a specificity of only 49% for distinguishing between BM and VM18 (with no significant variations between age groups50,60). However, CSF glucose level is not the most appropriate parameter for differentiating types of AM; rather, we should consider the CSF/blood glucose ratio, since blood glucose levels can be modified by concomitant illnesses, time of day, and other factors related with alcohol consumption and stress.1,64 A recent study establishes that a CP of 0.36 for the CSF/blood glucose ratio yields excellent results, with sensitivity and specificity both at 92.9%.64 Another novel study establishes that a CSF glucose level below 60% of blood glucose level is observed in 95% of patients with BM, with an OR of 20.82 (95% CI, 8.86-48.96; P=.001). This parameter, CSF lactate concentration, PMN percentage, and blood PCT level are the 4 individual factors found to be predictive of BM.29

Lactate is an end product of cellular anaerobic metabolism; therefore, increased CSF lactate levels are to be expected in bacterial processes.65 A CSF lactate concentration of 0 to 35mg/dL is considered normal; one important characteristic is that this does not depend on the blood lactate level.66 Blood lactate concentration should be taken into account, as is the case with CSF glucose levels.59 However, as in the case of pleocytosis, it should be noted that CSF lactate levels considerably decrease when antibiotic treatment has been administered previously.22,66,67

For several decades, increased lactate concentration has been associated with BM,1,58 although there is controversy regarding its true diagnostic yield.59 In recent years, lactate has again been reported to be the diagnostic marker with the highest sensitivity for BM in CSF analyses,68 with an almost incomparable yield, achieving an AUC of 1 and sensitivity and specificity both at 100%, with an established CP>35mg/dL.26

Lactate level measurement to distinguish bacterial from viral aetiology is equally useful in children and in adults, with the controversy also focusing on the most suitable CP. A CP of >33mg/dL is currently recommended, since it obtains a sensitivity of 95% and a specificity of 93.6%,69 although other studies consider a wider interval, ranging between 23 and 48mg/dL; these provide better AUC results, reflecting predictive power for bacterial aetiology of AM.21,70 Studies in elderly patients are more limited, for which reason no specific CP for that age group has been established to date.21

Although the most frequently selected CP is 35mg/dL, some studies in adults have also obtained better results with a CP of 33mg/dL: AUC of 0.942 (95% CI, 0.886-0.999; P<.001), 89.8% sensitivity, and 86.9% specificity.29

Two recent meta-analyses21,22 have confirmed the excellent predictive power of lactate concentration and its higher diagnostic yield in comparison with other CSF measurements. A meta-analysis by Huy et al.21 of 25 articles (1703 patients) reported excellent predictive power for lactate, with an AUC of 0.984 for CSF lactate concentration; this is significantly higher than that of other measurements (CSF glucose level, CSF/blood glucose ratio, protein level, and total pleocytosis). This suggests that lactate concentration may be used as a simple marker for diagnosis of BM (although the article also reports that it would be advisable to measure the remaining parameters for maximum reliability). The second meta-analysis (Sakushma et al.22), including 33 studies and 1885 patients, establishes an optimal CP for CSF lactate concentration of 35±1mg/dL, which obtains excellent results: a sensitivity of 0.93 (95% CI, 0.89-0.96), a specificity of 0.96 (95% CI, 0.93-0.98), a positive likelihood ratio of 22.9 (95% CI, 12.6-41.9), a negative likelihood ratio of 0.07 (95% CI, 0.05-0.12), and an OR of 313 (95% CI, 141-698). As stated previously, the meta-analysis showed reduced diagnostic power of lactate in patients who had previously received antimicrobial drugs, with sensitivity decreasing to 0.49 (95% CI, 0.23-0.75).22

As with serum PCT levels, researchers have studied CSF PCT values to distinguish between BM and VM. A recent publication compares the diagnostic power of serum and CSF PCT concentrations, concluding that blood PCT measurement is much more useful than the CSF value.71 Although one study found significant differences between CSF PCT concentration in BM vs VM (4.71±1.59ng/mL vs 0.13±0.03ng/mL), it only included 19 and 11 patients in each group.72 In any case, although additional studies have assessed the usefulness of CSF PCT levels, their diagnostic power is very limited and the results obtained are inferior to those of the remaining analysed variables.30–33 Therefore, CSF PCT determination is currently not recommended.73–75