Urinary tract infection (UTI) is one of the most common infectious diseases. Fever in UTI typically represents the presence of acute pyelonephritis but it may also reflect prostatitis and/or the urosepsis syndrome [1,2]. Patients with febrile UTI generally present with mild illness in primary care but may rapidly develop a life-threatening condition, progressing into septic shock and multiple organ failure. The overall mortality rate of pyelonephritis is approximately 0.3%, but in bacteremic patients it can be as high as 7.5% to 30% [3,4]. In addition, bacteremia in UTI is associated with prolonged hospitalization and higher complication rates [5-7]. Given this spectrum of disease, clinicians are vigilant to identify bacteremia at a patient's presentation.

The incidence of bacteremia in patients with acute pyelonephritis has been reported to be roughly 20% [8-10]. Several studies have been conducted to identify predictive characteristics of bacteremia in patients with UTI [6,7,11,12]. However, no single clinical model has been used in practice because of its poor value in predicting bacteremia. The gold standard for detection of bacteremia remains the performance of at least two blood cultures to achieve sufficient sensitivity [13]. There are, however, practical limitations. First of all, it takes at least 24 to 48 hours to attain the culture result. Secondly, there may be a false positive result as contamination rates of up to 7% have been reported [14]. Furthermore, the implementation of the surviving sepsis campaign, which recommends the immediate initiation of broad-spectrum antibiotic therapy once septicemia is suspected, leads to an increase in the performance of blood cultures with lower yield, likely reflecting the obtainment of additional cultures after initiation of antibiotics [15,16]. Therefore, there is a need for strategies that guide clinicians and help reduce avoidable blood cultures and, by consequence, medical costs.

The biomarker procalcitonin (PCT) is a marker of systemic inflammation and thus it may help to predict bacteremia [17,18]. The aim of this study was to assess clinical characteristics and the PCT value to predict bacteremia in patients with febrile UTI.

In this study, we evaluated the ability of clinical and laboratory characteristics to predict bacteremia in adults presenting with febrile UTI. We found that PCT dichotomized around 0.25 μg/l, is a robust surrogate marker for bacteremia, whereas the actual PCT value reflects bacterial load in the blood stream. PCT might be applied to help guide and limit the use of blood culture resources.

We used a PCT cutoff value of ≤0.25 μg/l after having tested different standard cutoff values as has been advocated by the manufacturer's instructions to indicate absence or presence of sepsis or even absence or presence of bacterial infection as has previously been demonstrated in lower respiratory tract infections [23]. Compared to studies regarding PCT and bacteremia in infections other than febrile UTI, our diagnostic threshold was lower resulting in a higher sensitivity and lower specificity [17,18,24,25]. A recent study with similar design in patients presenting with community acquired pneumonia demonstrated highly similar findings [26]. In that study, a PCT value ≤0.25 μg/l would allow reducing blood cultures by 37% while still identifying 96% of bacteremias [26].

Using a PCT value ≤0.25 μg/l, we demonstrate a 40% reduction of blood cultures in our study population while still identifying 97% of bacteremias. Using PCT as a decision rule to guide taking blood cultures in febrile UTI would thus likely to be cost-effective. Moreover, it might prevent false-positive blood cultures and costs of associated medical consultations. However, other laboratory values that might routinely be measured in patients presenting with febrile UTI such as C-reactive protein (CRP) and the erythrocyte sedimentation rate (ESR) could also be indicative for the presence of bacteremia. In this study, CRP and ESR were measured in a subset of ED patients when indicated by the attending physician. Both were significantly associated with bacteremia but had very limited diagnostic ability compared to PCT (see Additional file 1). This is like other studies that did not recommend the use of CRP and ESR for diagnosing bacteremia [24,25].

The clinical characteristics associated with the presence of bacteremia comprise two categories. One comprises clinical signs which are a result of the host's response to bacterial components and cytokines elicited by the local infection and possible systemic expansion (that is, chills, confusion, temperature >38.6°C, heart rate >100/minute) and the other category includes host-related risk factors for a complicated clinical course of disease such as older age and diabetes. All these clinical factors were found to be associated with bacteremia in previous studies in patients with UTI [6,7,11,12]. Similar to previous reports on smaller cohorts, we were not able to accurately predict the presence of bacteremia based on clinical characteristics only. Likely, this can be explained in part by the relatively old study population (median age 66 years) as various related coexisting illnesses might result in heterogeneous symptoms of bacteremia [27].

A relationship between PCT and TTP of the blood cultures has indirectly been suggested in the setting of discriminating blood contamination from bloodstream infection due to coagulase-negative staphylococci [28]. However, to our knowledge a direct relationship between PCT and the TTP of the blood culture in gram negative bacteremia has not been addressed previously. As the majority of bacteremic UTI is caused by gram negative microorganisms, we hypothesized that the bacterial load likely reflects the level of free lipopolysaccharide and thus the level of endotoxemia, which is correlated with the PCT value [18]. The TTP of the blood culture that depends on the rate of carbon dioxide production by the microorganisms can be used as a surrogate for systemic bacterial load, and we, thus, analyzed its correlation with PCT [29]. Because the TTP depends on the microorganism and the logistics around blood culture obtainment, we decided to analyze this for E. coli bacteremias of one center only [19]. We found a significant loglinear relationship between PCT value and TTP that supports biological plausibility between PCT value and the bacterial load of infection. Probably a similar phenomenon is indirectly illustrated by studies in lower respiratory tract infection that demonstrated that a low PCT value reflects a self-limiting disease that does not require antibiotic treatment while higher PCT values are associated with complicated outcome [30,31]. However, it should be emphasized that in this study low PCT levels were not indicative of absence of urinary tract infection. Hence, all patients included in this study received antimicrobial treatment. Therefore, additional studies are needed as to whether PCT might be of value in guiding antibiotic treatment of UTI and decision upon hospitalization as non-bacteremic patients are likely to be good candidates for outpatient treatment. In this respect, the results of a recent study are not promising as they do not support the use of PCT in helping guide physicians in deciding about hospitalization in patients with acute pyelonephritis [32]. This is in accordance with a smaller study demonstrating that PCT was not correlated with adverse outcome of acute pyelonephritis [33]. Interestingly, this latter study also showed significant higher PCT levels in bacteremic patients compared to nonbacteremic patients.

Our study has several strengths. First of all, we prospectively included consecutive patients with febrile UTI at multiple sites at primary care and ED setting. Thus, our study population reflects the broad population of routine clinical practice. Secondly, we were able to achieve blood culture and PCT results in over 90% of the study population. Furthermore, the rate of bacteremia was 23% indicating that many patients suffered the urosepsis syndrome [2]. Recommended by sepsis guidelines, all such patients require blood cultures before the initiation of antibiotic treatment [16]. Yet, using PCT ≤ 0.25 μg/l as a decision rule would have resulted in a 40% reduction of blood culture utilization, with 3% loss of detection of bacteremia. The relation between PCT and TTP supports previous suggestions in other infections that PCT may serve as a predictive biomarker for degree and severity of bacterial invasion.

There may, however, also be some limitations. Almost 30% of the patients did use antibiotics at the time of presentation as fever apparently developed during treatment of a nonfebrile UTI, for example, cystitis. This may have led to false negative blood cultures and could contribute to a relative low specificity of PCT in diagnosing bacteremia. However, antibiotic pretreatment for cystitis in The Netherlands usually concerns nitrofurantoin, a drug that is unlikely to affect bacteremia in UTI. Consistent herewith, pretreatment was associated with a higher chance of bacteremia and this suggests that antibiotic pretreatment did not skew our results towards negative blood cultures. Nevertheless, this still does not exclude the possibility that the rate of bacteremia may reflect an underestimate. Another limitation might be the measurement of PCT values that was done afterwards. Though the frozen storage of blood sample does not influence its PCT value, the measurement of PCT in routine clinical practice might be different [34]. Furthermore, when used to limit the use of blood cultures, a quick result of PCT, preferably by a readily available point-of-care assay, is mandatory for practical reasons.

This study might have consequences for the current practices on EDs as implementation of a PCT strategy likely is a cost-effective way to avoid taking blood cultures with a very low chance of yielding a positive culture. Moreover, besides in febrile UTI, this also seems to hold for patients presenting with community acquired pneumonia [26]. Taken together, these studies suggest that in the majority of patients presenting with febrile illnesses at ED, being either respiratory or urinary tract infections, medical diagnostic costs can be reduced. However, it should be highlighted that additional validation studies are needed, as the in- and exclusion criteria applied in this study might limit its generalizability to other settings and special patients groups. Furthermore, implementation studies addressing its cost-effectiveness are needed before the widespread use of PCT guidance on doing blood cultures in routine clinical practice can be recommended.

We conclude that PCT accurately predicts the presence of bacteremia and its bacterial load in adults with febrile UTI. A PCT value ≤0.25 μg/l sufficiently rules out bacteremia in febrile UTI and may be used to help guide efficient use of blood culture resources.

• According to sepsis guidelines, blood cultures should be drawn to help diagnose bacteremia in case of febrile UTI, but the usefulness and cost-effectiveness of this practice have been questioned.

• This study confirms that bacteremia in febrile UTI can neither be predicted nor ruled out by bedside available clinical parameters.

• A low value (≤0.25 μg/l) of the biomarker procalcitonin (PCT) sufficiently rules out bacteremia in febrile UTI.

• Implementation of PCT into clinical practice with the aim to limit avoidable blood cultures is likely to be cost effective.

• In case of bacteremia the level of PCT appeared to be a marker of the bacterial load. Whether this might have implications for the dosage and length of antibiotic treatment awaits further studies.

AUC: area under curve; CFU: colony forming unit; CI: confidence interval; CRP: C-reactive protein; ED: emergency department; ESR: erythrocyte sedimentation rate; LR: likelihood ratio; NPV: negative predictive value; OR: odds ratio; PCT: procalcitonin; PPV: positive predictive value; ROC: receiver operating characteristic; TTP: time to positivity; UTI: urinary tract infection.

The authors declare that they have no competing interests.

JWW, CN and JTD were responsible for the original design. CN, JWW and JTD were the guarantors. CN and TNB were responsible for data management, carried out the statistical analysis and wrote the initial draft supervised by JTD and JWW. CN, TNB, JWW, GHG, TK, GHWL, NMD, HCA and EMS were involved in patient recruitment and data collection. JWW, EJK, MJB, GHG, TK, GHWL, NMD, HCA and EMS critically revised the manuscript. All authors contributed to and approved the final version of the manuscript.