Our data show that clinical signs and symptoms routinely used and recommended to screen and to establish the diagnosis of CAP are of limited value. Both, PCT and hsCRP improve the diagnostic value of the clinical assessment. On admission, PCT has the highest diagnostic accuracy in differentiating radiologically confirmed CAP from other differential diagnoses, to predict the later finding of bacteremia and to assess the severity of CAP.
According to most guidelines, an infiltrate in chest radiograph is mandatory for the diagnosis of CAP. However, in the setting of primary care, radiography may often not be performed . Instead, based on the clinical suspicion and in view of the severe consequences of delayed antibiotic therapy, physicians have a low threshold to initiate antibiotic therapy in presumed CAP [29, 30]. Obviously, the timely clinical evaluation of patients with symptoms suggestive of CAP is important to estimate the pretest probability for the disease, to assess the severity of illness, and to start appropriate antimicrobial therapy. However, as confirmed by our data, most routinely used clinical parameters, alone or in combination, have poor diagnostic accuracies to predict CAP. Similar clinical signs and symptoms are caused by acute bronchitis, AECOPD, asthma exacerbations and non-infectious diagnoses such as congestive heart failure or atelectasis. In addition, the interpretation of the clinical assessment lacks standardization and validation and is, therefore, prone to interobserver variability [31, 32]. We are aware that the lack of diagnostic accuracy of history and clinical examination has been reported [5, 31]. However, despite their known limitations, clinical signs are daily used for decision making in clinical routine. Herein, we propose the use of biomarkers to complement and improve clinical assessment.
As shown by our data, together with a careful clinical assessment and radiology, biomarkers like PCT and hsCRP can significantly increase diagnostic accuracy for the diagnosis of CAP. One might argue that the diagnosis of bronchitis, CAP and AECOPD should be clinically evident as should the indication for antibiotic therapy. However, despite the presence of excellent and recent guidelines, the implementation into clinical routine namely for these "evident diseases" is insufficient. Thus in our opinion, as a biomarker PCT becomes especially valuable as a powerful tool to better complement and implement these guidelines, as shown in our intervention studies.
In the setting of an emergency department of a hospital, most patients presenting with symptoms of respiratory tract infection will receive a chest radiograph. In the presence of a new infiltrate, a bacterial etiology is usually assumed. However, bacterial causes must be differentiated from other, non-infectious or viral etiologies, which can be challenging. We demonstrate that the routinely used clinical parameters in radiographically defined CAP miss the diagnostic accuracy to differentiate between bacterial and non-bacterial CAP. PCT, better than hsCRP, improves diagnostic accuracy to distinguish bacterial CAP from non-infectious or non-bacterial causes, respectively. Thus, if a patient shows an infiltrate on chest radiograph in the presence of acute respiratory symptoms and very low PCT levels (< 0.1 μg/L), clinicians should actively seek for an alternative diagnosis to bacterial pneumonia.
Positive blood cultures in CAP patients correlate with adverse outcome and, thus, a rapid initiation of antimicrobial treatment is pivotal [33–35]. However, results from microbiological cultures of body fluids are only available after 24 to 48 hours, which can be problematic for clinical practice. A PCT cut-off of 0.25 μg/L had a 98% sensitivity to detect bacteremia. PCT may thus provide valuable and faster information about severity of disease long before blood culture results become available. Accordingly, PCT better mirrored the severity of CAP classified by the PSI, as compared to hsCRP levels, leukocyte counts or the visual analogue scale, which were unable to distinguish mild from more severe pneumonia.
Several limitations of our study merit consideration. First, results of two studies were combined introducing the possibility of a selection bias, although in both studies the primary endpoint was similar. Second, since antibiotics were withheld based on PCT levels, this may have introduced a bias to the favor of PCT. Conversely, cure of CAP under antibiotic therapy may falsely have been considered as proof of bacterial etiology in a considerable proportion of patients who indeed had a non-bacterial etiology. Third, our results may not apply for immunosuppressed patients and other clinical settings or sites of infection, especially localized or fungal pulmonary infections such as empyema and aspergillosis. However, these are unlike conditions in outpatients presenting with lower respiratory tract infections. The diagnostic accuracy of PCT in patients with immunosuppression or hospital-acquired pneumonia has to be evaluated in future studies. These patients were excluded for safety reasons. Forth, interobserver variation in the clinical evaluation of patients with CAP has not been examined. Other studies have revealed considerable interobserver variability in the recording and evaluation of symptoms. Fifth, we only assessed total leukocyte count and not band forms. However, in recent studies, the superiority of hsCRP and PCT as compared to leukocyte count and band counts has been shown [25, 36, 37]. Finally, the rate of microbiologically documented CAP in our study population was rather low, limiting information about the diagnostic accuracy of PCT for the etiological diagnoses. We did not routinely perform serology or PCR or culture in blood and respiratory secretions for Mycoplasma pneumoniae and Chlamydia pneumoniae. Moreover, search for Streptococcus pneumoniae antigen in urine was not routinely done. However, using representative respiratory secretions and blood cultures the rate of documented bacterial CAPs in our study was very similar to the one in different recent studies  or .
Strengths of our study are first that the study population included a relatively diverse group of patients with lower respiratory tract infections. Second, we did not include clinically unrealistic control patients without suspected infection, but only patients with a high pretest probability of CAP, covering the spectrum that is likely to be encountered in the future use of these tests . Our study is thus based on a real-life patient sample to closely resemble clinical practice in a emergency room setting. As compared to primary care setting, our study cohort might have a higher pretest probability for pneumonia which might bias our question about the diagnostic accuracy of PCT for diagnosing CAP solely based on history, clinical examination and laboratory parameters without radiography, an approach occasionally done in primary care.
Finally, the rational in our trials was the concept that diagnosis is not the principle outcome measure in the traditional sense of diagnostic test evaluation. Instead, these intervention studies looked directly at patient outcomes, assuming that if the patient recovered without antibiotics then there was no serious bacterial illness. Although not being a new "gold standard", this circumvented the problem of the non-existent diagnostic "gold standard" to decide on the presence or absence of a clinically relevant bacterial infection based on traditional criteria.