Background

Acute respiratory tract infections (ARIs) are the most common reason for antibiotic therapy in primary care [1, 2]. Yet, about 70 % of ARIs are treated with antibiotics, despite their mainly viral etiology [3]. As a consequence, antibiotic overuse results in increased antibiotic resistance, antibiotic-induced adverse events, and costs [4, 5]. A novel approach for more individualized and targeted antibiotic therapy is the use of blood biomarkers, such as C-reactive protein (CRP) or Procalcitonin (PCT) [6–8]. Several trials in primary care have shown that the use of one of these markers reduces antibiotic exposure, although head to head comparison trials are currently lacking [9–11]. A recent systematic search and meta-analysis found a total of four observational studies from primary care looking at CRP and PCT for their ability to predict pneumonic infiltrate, need for hospitalization, and diagnosis of group A streptococcus pharyngitis [9]. Also, previous observational studies compared the diagnostic performance of CRP and PCT in the hospital setting [12, 13] and in septic patients in the ICU setting [14, 15]. However, a comprehensive comparison analysis between these two markers in the primary care setting is currently lacking. Importantly, so far PCT protocols have been used mainly for management of hospitalized and intensive care patients due to the lack of high sensitive point of care tests (POCT). Recently, a sensitive POCT has been developed and may be increasingly used in primary care [16].

Herein, using data from a previous trial, our aim was to study how well CRP and PCT correlate, and their predictive value for outcome discrimination in a large-scale well-defined ARI cohort [17, 18].

Methods

Study design and setting

This predefined ancillary project included all patients with upper and lower ARI enrolled between December 2004 and April 2006 into the PARTI trial (Procalcitonin-guided antibiotic use versus a standard approach for acute respiratory tract infections in primary care) [17]. Study details have been reported elsewhere [18]. In brief, 53 primary care physicians in northwest of Switzerland consecutively screened adults with symptoms of an ARI and, in their physician’s opinion in need of antibiotics, were randomized to either a PCT-guided approach to antibiotic therapy or to a standard approach where physicians were asked to adhere to current guidelines. The primary non-inferiority outcome was the number of days, during the first 14 days after baseline, where a patient’s activities were restricted by an ARI.

Selection of participants

All patients with the diagnosis of upper or lower RTI and the physician’s intention to prescribe antibiotics, based on evidence guidelines, were consecutively included. Exclusion criteria were antibiotic use within the previous 28 days, psychiatric disorders or inability to give written informed consent, not being available for follow-up, not being fluent in German, severe immunosuppression, cystic fibrosis, active tuberculosis, and the need for immediate hospitalization.

Classification of patients was standardized according to previously defined guidelines for the correct diagnosis of patients with ARI [18]. For these subgroup analyses we divided patients into three groups based on the initial evaluation by a primary care physician: upper ARI comprised patients with rhinosinusitis, pharyngitis/tonsillitis, and otitis media; lower non-pneumonic ARI comprised patients with common cold, tracheo-bronchitis, influenza, acute exacerbation of asthma or chronic obstructive pulmonary disease, and lower pneumonic ARI comprised patients with community-acquired pneumonia (CAP).

Biomarker measurement

Blood samples were collected from all recruited patients on admission and day seven in ethylenediaminetetraacetic acid (EDTA) tubes. PCT determinations were made using a centralized time-resolved amplified cryptate emission technology-based assay (Kryptor® PCT, Thermo Scientific Biomarkers [B · R · A · H · M · S AG], Hennigsdorf, Germany) with a 0.06 μg/L functional sensitivity [19]. CRP concentrations were determined by an enzyme immunoassay having a detection limit <5 mg/dL (EMIT; Merck Diagnostica, Zurich, Switzerland). For the purpose of comparing cut-off ranges, PCT and CRP were stratified in four corresponding groups each, namely for PCT <0.1 μg/L, 0.1-0.25 μg/L, >0.25-0.5 μg/L and >0.5 μg/L, and for CRP ≤20 mg/dL, 21–50 mg/dL, >51-99 mg/dL, and ≥100 mg/dL, respectively, based on the use of these markers in previous antibiotic stewardship trials in the low acuity setting [2, 20, 21].

Outcomes

In line with the initial study protocol [18], the predefined primary endpoint was the number of days, within the first 14 days after baseline, during which a patient’s daily activities (work or recreation) were restricted by a respiratory tract infection. The secondary endpoint was persistence of discomfort from infection at day 14. Both endpoints were assessed by seven medical students, blinded to the goal and design of the study, by conducting standardized follow-up interviews by telephone at 14 and 28 days after baseline.

Statistical analysis

We used descriptive statistics including mean with standard deviation, median with interquartile range (IQR), and frequencies to describe the populations, as appropriate.

Assessment of correlation between biomarkers was performed by Spearman’s rank correlation analysis and visualized with scatter-plots. We looked at correlations between continuous biomarker levels and across cut-off ranges as defined above. For biomarker kinetics, we defined the change from baseline to day seven and classified the relative change into 4 groups (decrease of <25 %, 25–50 %, 50-75 % and >75 %).

To investigate associations between initial biomarker levels and endpoints, we used linear and logistic regression analysis. We adjusted for sociodemographic characteristics (age, gender, existence of comorbidities and education level) and in a second model also for disease severity (e.g. days spent with restricted activities before randomization and antibiotic prescription). Analyses were performed in the overall ARI patient population and in subgroups of different ARI types (upper ARI, non-pneumonic lower ARI, and pneumonic lower ARI). We calculated sensitivity and specificity using different clinically established cut-offs as defined above.

Statistical t-Tests were carried out at a 5 % significance level. All statistical analyses were performed with STATA Version 12.1 (Stata Corp, College Station, TX, USA).

Results

Patient population

This study included a total of 458 patients with a median age of 45 years (60 % females). Forty percent (184 patients) had upper ARI, particularly acute rhinosinusitis and acute pharyngitis/tonsillitis. Lower non-pneumonic ARI was diagnosed in 45 % of patients with acute bronchitis being the pre-dominant diagnosis (28 %) and 15 % had pneumonic lower RTI, namely CAP. Chest X-ray was performed in 121 patients, in 23 % of non-pneumonic ARI and in 97 % of CAP patients with non-pneumonic and pneumonic LRTI. The median reported degree of discomfort on recruitment was 6 – 7 (on a scale of 0 – 10 with 0 meaning no discomfort at all). Additional baseline characteristics of the study population are summarized in Table 1.

Table 1 Patient characteristics
Full size table

Correlation between PCT and CRP in different ARIs

Figures 1 and 2 show correlations of CRP and PCT in the overall population, and stratified by subgroups at baseline and on day seven. At baseline, we found significant correlations with however a low correlation coefficient between the two markers in the overall study population (r2 = 0.16), as well as in all subgroups. Highest correlations were found in the pneumonic lower ARI group (r2 = 0.34). On day seven, there were again weak correlations between both biomarkers in the overall population as well as in the subgroups.

Fig. 1
figure 1

a-d: Correlation of Procalcitonin and CRP levels in lower and upper respiratory tract infections at baseline

Full size image
Fig. 2
figure 2

a-c: Correlation of Procalcitonin and CRP levels in lower and upper respiratory tract infections on day 7

Full size image

We also investigated correlation of markers across well-established cut-off ranges (Tables 2 and 3). At baseline, 42 % of patients were in similar cut-off ranges with 9 % of patients being in higher PCT ranges and 49 % in higher CRP ranges. For day seven, 76 % of patients were in similar cut-off ranges with 13 % of patients being in higher PCT ranges and 11 % in higher CRP ranges. Looking at kinetics found 11 % of patients had a similar decrease in biomarker level with 75 % of patients having a stronger CRP decrease and 9 % a stronger PCT decrease. An increase in biomarker level was observed in 5 % of patients.

Table 2 Classification of biomarkers at baseline and day seven
Full size table
Table 3 Classification of biomarkers at kinetics
Full size table

Association of admission biomarker levels and adverse outcome

Table 4 summarizes the results of regression analyses investigating the association of initial biomarker levels and adverse outcomes, namely days with restricted activities and persistence of ongoing discomfort at day 14. Analyses were performed according to baseline and day seven biomarker levels. For CRP, we found levels >100 mg/dL to be significantly associated with days of restricted activities (coefficient: 1.6 (95 % CI: 0.5, 2.6), p = 0.005). For day seven, results were similar with again levels >100 mg/dL being significantly associated in sociodemographically adjusted analysis (coefficient: 3.3 (95 % CI: 0.5, 6.2), p = 0.022) and a trend in fully adjusted analysis. For PCT, only initial levels >0.5ug/L were significantly associated with days with restricted activities in sociodemographically adjusted analysis (coefficient 2.0, 95 % CI 0.0 to 4.0) with a trend after full adjustment (p = 0.089).

Table 4 Biomarker at baseline and day seven as predictor for days with restricted activities or persistence of discomfort after 14 days
Full size table

We also repeated the same analysis in the different subgroups, i.e. in patients with upper ARI and in lower pneumonic and non-pneumonic ARI. While for the upper ARI no significant associations were found, best results were seen in the non-pneumonic lower ARI group (see Additional files 1, 2, 3, 4 and 5).

Finally, we also investigated associations of biomarkers with persistent discomfort after 14 days. For this endpoint, no significant associations were found for both markers at baseline. For the day seven analysis, significant results were found for CRP, but not PCT.

Discussion

The findings of this analysis including a large primary care population with different types of ARI are threefold. First, correlations between CRP and PCT levels were low in the overall population and particularly in the upper and non-pneumonic lower ARI. Second, high levels of both markers in the highest cut-off range were associated with days with restricted activity but not with ongoing discomfort. Third, kinetics did not significantly improve outcome prediction.

To date, ARI still represent an important public health issue affecting millions of patients around the world. Incorrect or delayed diagnosis may directly lead to severe complications and increases morbidity and mortality, especially in the inpatient setting [3]. Sputum and blood cultures are commonly seen as the gold standard. Still, these techniques have low sensitivity and lack practicability (e.g. in case of non-productive cough) [22, 23]. Because differentiation of mild viral infection from more severe infections is challenging, clinicians often start empiric antibiotic treatment which increases resistance of common bacteria and causes drug-related side effects [24, 25]. In this context, objectively measurable blood biomarkers may help to assess severity and need for more intensive treatment. There is still uncertainty about optimal use of PCT or CRP in the primary care setting for the management of patients. Herein, our analysis including a large and well defined cohort of ARI patients sheds new light on this important topic.

Compared to higher acuity settings, patients with low acuity ARI had low PCT levels with minimal changes over seven days. Yet, CRP levels showed a higher variability. This is also consistent with findings from other primary care studies [26], suggesting that PCT is more specific to bacterial infections and low in patients with mainly viral infections, while CRP increases independent of infection type as a “inflammatory marker”. Interestingly, as shown in our correlation analyses, patients displayed differences in biomarker profiles with low correlation between markers and between cut-off ranges. Because both markers had suboptimal prognostic accuracy for outcome prediction, and because no goldstandard for ARI in need of antibiotics exist, only a head to head trial comparing the performance of both markers in regard to antibiotic management may help to clarify the question which marker is more helpful in the management of ARI patients.

Several limitations should be considered when interpreting our results. First, this is a secondary analysis of previous randomized study with a different study question, i.e., whether antibiotic therapy guided by PCT reduces the use of antibiotics without increasing the restrictions experienced by patients. Second, no external validation of ARI diagnoses was done. Instead, physicians were encouraged to use updated guidelines presented in an interactive seminar. Also, chest X-ray was not mandatory for diagnosis of CAP and some patients may have been misclassified. In addition, we classified patients with common cold, influenza, and acute laryngitis-bronchitis as lower ARI because they frequently presented with cough as a main symptom and pathological findings in chest auscultation. Third, we focused on adverse outcomes as specified in the original trials but did not look into other adverse outcomes such as unplanned readmission to the GP, hospital admission, and respiratory failure among others. Further, PCT was not measured on site, due to lack of sensitive POC at that time. Future studies comparing these markers should also look into POC system for both markers in primary care.

Conclusion

This is the first large and comprehensive study investigating correlations between CRP and PCT and their predictive value in the primary care setting in patients with different types of ARI. The low correlations between the two biomarkers and the only moderate prognostic accuracy calls for a head-to-head trial comparing the ability of both markers to manage primary care patients with ARI to answer the question which marker is superior.

Ethics approval and consent to participate

The ethics committee of the University Hospital Basel, Switzerland, approved the trial protocol. The trial was supervised by an independent monitoring board consisting of a general internist in primary care, an infectious disease specialist, and a pneumologist. All participating physicians and patients gave written informed consent.

Consent for publication

Not applicable.

Availability of data and materials

Due to our local Institutional review board policy, we do not have permission to make the data sets on which the conclusions of the paper rely publicly available. A truncated data set (removing all potentially identifying features) may be made available on an individual request basis.