C-reactive protein levels in community-acquired pneumonia

C-reactive protein (CRP) is an acute phase protein synthesised by hepatocytes. In response to infection or tissue inflammation, CRP production is rapidly stimulated by cytokines, particularly interleukin (IL)-6, IL-1 and tumour necrosis factor 1–3. Although its exact function in vivo is not known, it probably has a role in opsonisation of infectious agents and damaged cells.

Two different uses of CRP have been investigated. Firstly, as a diagnostic tool to distinguish between noninfectious and infectious conditions and within the latter between viral and bacterial or superficial and deep infections 1, 2. CRP levels are usually lower in viral and superficial bacterial infections than in deep bacterial infections. Secondly, as a prognostic and follow-up test, as serial measurements may be useful to evaluate the response to antibiotic treatment and to detect complications in patients with infections 1, 2, 4, 5.

Several studies have corroborated the role of CRP in the diagnosis of bacterial versus viral meningitis 6 in children and suspected septicaemia in neonates 2, 7. However, its role in the aetiological diagnosis of respiratory infections is not well established. Usefulness of CRP to distinguish between bacterial and viral pneumonia or typical and atypical pneumonia has been previously analysed but data are discordant 8–10 and its value as a first-line method of screening remains inconclusive.

The aim of the present study was to evaluate the diagnostic value of CRP serum levels at admission as an indicator of the aetiology of community-acquired pneumonia (CAP).

Patients and methods

A total 1,222 consecutive patients aged >14 yrs with acute symptoms consistent with CAP were studied from October 1996–October 2000, according to a standard protocol in the Respiratory and Infectious Diseases Services at the Hospital Clínic in Barcelona (Spain), a 800-bed university teaching hospital.

CAP was defined as the presence of a new infiltrate on the chest radiograph along with appropriate clinical history and physical signs of lower respiratory tract infection in a patient not hospitalised within the previous month and in whom no alternative diagnosis emerged during follow-up. Clinical, laboratory and radiological features at presentation as well as other epidemiological data were recorded on a specific questionnaire and entered in a computer database. CAP severity was assessed within the first day of admission using patients outcome research team (PORT) score 11. This study population has been previously described 12.

Patients with neutropenia (<1.0×10⁹·L⁻¹), human immunodeficiency virus (HIV) infection, tuberculosis, fungal infection and those treated with steroids in a prednisone-equivalent dosage of >20 mg·day⁻¹ for ≥2 weeks were excluded.

Results

An aetiological diagnosis was achieved in 534 of 1,222 evaluated patients (45%). One-hundred and thirty-six patients who had two aetiological diagnoses (11%) were excluded from the analysis. Out of the remaining 398 patients with a single diagnosis only 258 had a CRP measurement within the first 24 h of admission and these constituted the study population. These CRP levels were available in 80 dual aetiology cases and in 383 CAP cases of unknown aetiology.

The mean±sd age of the 258 patients was 66±18.6 yrs and 167 were male (65%). Distribution of sex and ages was not different according to aetiology of pneumonia.

In the analysed cohort, the predominant pathogen was S. pneumoniae (80 cases, 31%), followed by L. pneumophila (30 cases, 12%) and Haemophilus influenzae (26 cases, 10%). An atypical bacterial pneumonia due to C. pneumoniae, C. psitacci, M. pneumoniae or C. burnetii was diagnosed in 52 patients (20%) and a viral pneumonia in 35 (14%).

Mean CRP levels according to aetiological diagnosis are shown in table 1⇓. CRP values (mean) were significantly higher in patients with L. pneumophila (25 mg·dL⁻¹) than in those with pyogenic pneumonia (16 mg·dL⁻¹), viral pneumonia (14 mg·dL⁻¹) and atypical pneumonia (13 mg·dL⁻¹) (p=0.0002). When grouping the patients in those with (n=30) and without (n=228) L. pneumophila pneumonia, the mean CRP values in the L. pneumophila group was still significantly higher than in the group with other diagnoses (25 mg·dL⁻¹ versus 15 mg·dL⁻¹, p=0.0003). Mean CRP values were not significantly different among non-L. pneumophila diagnostic groups (p=0.24) or among those 80 patients with more than one pathogen identified or the 383 with CAP of unknown aetiology (p>0.05).

The mean number of days that patients were ill before admission was not significantly different in those with L. pneumophila pneumonia (5 days) and those with non-L. pneumophila pneumonia (5.4 days) (p=0.8).

CRP levels were not correlated with PORT score (r=−0.04; p=0.5) or age (r=0.05; p=0.4) and accordingly, they were not significantly different among PORT groups (19 mg·dL⁻¹ in groups I–II, 16 mg·dL⁻¹ in group III and 16 mg·dL⁻¹ in groups IV–V, p=0.5) or between the elderly and younger patients (16 mg·dL⁻¹ in patients aged <65 yrs (n=95) and 16 mg·dL⁻¹ in patients aged ≥65 yrs (n=163); p=0.85). On the other hand, although age was not significantly different among aetiological categories (p=0.21), mean PORT scores among aetiological groups were different by ANOVA analysis (p=0.02). This was apparently due to the fact that patients with L. pneumophila pneumonia had slightly less severe pneumonia than patients with pyogenic pneumonia (mean PORT score 91 (p<0.05) versus 110 (p>0.1), respectively, by pair-wise analysis with Bonferroni correction).

According to a receiver operator curve, a cut-off point of 25 appeared to be the best CRP value for the diagnosis of L. pneumophila pneumonia. A CRP level >25 mg·dL⁻¹ had a sensibility, specificity, positive predictive value and negative predictive value of 0.6, 0.83, 0.3, and 0.94, respectively.

As expected, logistic regression analysis confirmed that high CRP values were independently associated with L. pneumophila aetiology. After controlling for age and PORT score, the odds that patients with L. pneumophila pneumonia had a CRP level >25 mg·dL⁻¹ were 6.9 times higher than that of patients with non-L. pneumophila pneumonia (95% confidence interval (CI) 3.02–15.8; p<0.0001); 5.7 times higher than those with pyogenic pneumonia (95% CI 2.4–13.6; p<0.0001); 13 times higher than those with atypical pneumonia (95% CI 3.6–47.7; p<0.0001); and 7.8 times higher than those with viral pneumonia (95% CI 2–29.6; p=0.002) (table 2⇓).

Discussion

Early and appropriate treatment of CAP patients is one of the most important factors to reduce morbidity and mortality 13. A causative agent is seldom identified in >50–70% of cases, S. pneumoniae being the most common pathogen. Pneumonia due to C. burnetii, C. psitacci, C. pneumoniae, M. pneumoniae and respiratory viruses is usually mild, though L. pneumophila pneumonia may be severe and its detection can be of epidemiological importance. Regarding the aetiological diagnosis of pneumonia, clinical and chest radiographical findings lack accuracy 14, 15 cultures take at least 24 h to produce a positive result and specific rapid tests based on the detection of soluble antigens of S. pneumoniae or L. pneumophila in body fluids are not always available. Therefore, initial antibiotic therapy is usually chosen on an empirical basis 16.

The role of CRP as a tool in the diagnostic work-up of patients with lung infiltrates of presumed infectious aetiology remains controversial 8–10. Adnet et al. 17 showed that high CRP levels were helpful in the diagnosis of bacterial pneumonia secondary to aspiration in patients with drug-induced coma and pulmonary infiltrates, in whom other parameters such as fever and white blood cell counts were of poor sensitivity and specificity. Pönka and Sarna 18 and Örqvist et al. 19 showed that pneumonias caused by S. pneumoniae, especially when bacteraemic, were associated with a greater host response (higher levels of IL-6 and CRP) than those caused by other pathogens (M. pneumoniae and viruses). Lehtomäki 20 found that higher CRP values distinguished pneumococcal pneumonia from other pneumonias (16 mg·dL⁻¹ in the former versus 5 mg·dL⁻¹ in adenoviral pneumonia, 6 mg·dL⁻¹ in mycoplasmal pneumonia; p<0.001). Some 21, 22 but not all studies 8, 23, pointed out that CRP had among other markers of inflammatory response the best predictive value to distinguish viral from bacterial pneumonia.

In the present series, CRP levels in patients with a viral diagnosis were not significantly different from those of patient with pyogenic or non-L. pneumophila, nonviral atypical pneumonia. The present data referring to patients with viral pneumonia showed that mean CRP concentrations (14 mg·dL⁻¹) were higher than those observed in children (2.6–5.4 mg·dL⁻¹) 24, 25 but of the same range as those reported both in children with viral plus bacterial superinfection (9.5 mg·dL⁻¹) 26 and in hospitalised adults with influenza (12.3 mg·dL⁻¹)27. The authors believe that the most plausible interpretation of this high CRP value in patients with a sole viral diagnosis is that they actually had secondary but nondocumented bacterial infection 28.

The present data show that CRP levels in patients with L. pneumophila pneumonia are higher than in those with CAP of any other aetiology, independently of potential confounders such as age and severity of illness. To the best of the authors' knowledge, this association has not been previously reported and raises the question of whether L. pneumophila triggers more (or different) inflammatory pathways than other atypical microorganisms. The production patterns of cytokines and acute-phase proteins are not similar in different inflammatory conditions. It has been speculated that in lung infections, damage in proximity to pulmonary circulation produces an immunological stimulus to hepatic CRP synthesis 1–7. The greater the lung damage, the higher the CRP levels. The role of the inflammatory response in the pathogenesis of L. pneumophila pneumonia compared to that of pneumonias caused by other atypical agents should be further investigated, since the identification of important pathogenic factors could possibly provide alternative means for intervention in an entity that remains a potential fatal disease. Both duration of infection and antibiotic treatment can influence CRP results. It is known that radiological chest infiltrates in patients with L. pneumophila infection may worsen after the first dose of antibiotics, probably due to a Jarisch-Herxheimer-like reaction 29. It could be that as part of this reaction, CRP levels also increase. Therefore, CRP concentrations previous to antibiotic therapy should be considered in order to accurately assess the relationship between CRP levels and L. pneumophila pneumonia. This was a limitation of the present study, because CRP was measured within the first 24 h of admission and not necessarily before the administration of the first antibiotic dose.

Despite the association of high CRP levels with L. pneumophila pneumonia, a satisfactory cut-off point to fulfil the positive predictive requirements for a useful diagnostic test at the prevalence of disease observed in this study, could not be found. However, with the current sensitivity and specificity of a CRP level >25 mg·dL⁻¹, the positive predictive value could reach 77% at a prevalence of L. pneumophila pneumonia of 50%. Therefore, it is predictable that in an epidemic situation, CRP can be a cheap and readily available test for the presumptive identification of case patients, particularly in settings where other rapid diagnostic tests such as those aimed at the detection of urinary antigen are not available. Conversely, the data show that CRP can be very useful to rule out L. pneumophila pneumonia (negative predictive value of 94%), even at a disease prevalence typical of an endemic situation. It is recognised, however, that the availability of any diagnostic test aimed at the presumptive diagnosis of L. pneumophila pneumonia may be less of a need if, as recommended, a macrolide is systematically included as part of the empirical regimen given to patients with CAP who require hospitalisation or if a new quinolone is chosen as a single therapy. The recommendation of including a macrolide is based not only on the frequency of atypical microorganisms as causal agents of CAP, but also on the evidence that patients with pneumococcal pneumonia do better with a β-lactam-macrolide combination than with a β-lactam alone 13, 30. Nevertheless, given the known antagonism of macrolides on the in vitro antimicrobial activity of β-lactams against S. pneumoniae 31, the putative beneficial effect of a β-lactam-macrolide combination in patients with pneumococcal pneumonia should be confirmed, ideally in the setting of a clinical trial.

To conclude, the data suggest that L. pneumophila leads the inflammatory host response in a different way to other intracellular organisms and that C-reactive protein may aid physicians to rule out L. pneumophila pneumonia.