Procalcitonin (PCT)-guided antibiotic stewardship: an international experts consensus on optimized clinical use

Consensus process

The expert group reviewed the current evidence from interventional trials on PCT-guided ABS and discussed the different approaches and algorithms that were used, including those that did not lead to reduced antibiotic exposure. The experts also discussed the clinical evidence for different patient populations such as primary care [15], ED [16], ICU [7], [17] or geriatrics and exchanged practical experience from own experience in routine clinical practice. In addition, experience on the process and the impact of clinical implementation was shared and discussed. Based on the discussions, three modified PCT algorithms were proposed for patients with mild, moderate and severe illness.

The controversial issues were openly discussed, debated and the algorithms were further edited during several feedback rounds by incorporating adjustments until consensus was found. All delegates who attended the meeting then voted to: 1) agree, 2) disagree or 3) abstain, on each algorithm on the same day. For the voting a modified Delphi process was used [18].

PCT algorithms used in clinical trials on ABS in respiratory infections

The concept of PCT-guided ABS was first tested in ED patients with infection of the lower respiratory tract [19]. Considering that PCT remains low (undetected) in viral infection and increases in bacterial infection, the study algorithm recommended very strongly or strongly against the use of antibiotics if PCT levels were <0.1 μg/L or <0.25 μg/L, respectively. The algorithm also included some overruling criteria, so patients at very high risk would still be treated empirically despite low PCT levels. Accordingly, a two-level recommendation was given also for the initiation of antibiotic treatment for patients with PCT higher than 0.25 μg/L (antibiotic recommended) and PCT higher than 0.5 μg/L (antibiotic strongly recommended). The study demonstrated significant reduction in antibiotic prescription rates, particularly in patients with bronchitis and COPD exacerbation.

Later studies not only investigated PCT for the initiation of empirical antibiotic therapy, but also used PCT to monitor the response to therapy and to decide on discontinuation of antibiotic therapy on an individual basis [19], [20]. The drop of PCT below 0.25 μg/L or by at least >80%–90% from the peak was used as stopping rule thresholds. This approach further decreased antibiotic exposure by shortening the duration of therapy, particularly in patients with CAP. Subsequently a large Swiss-wide, randomized, non-inferiority trial found this approach to be highly effective in reducing antibiotic exposure by more than 3 days with no increase in the risk for adverse outcome [16], [21].

Today, several similar trials have been conducted in different countries and different clinical settings (i.e. from primary care to emergency departments and medical wards, to intensive care). A recent meta-analysis including individual patient data from 6708 patients with different types and severities of respiratory infections from 26 randomized-controlled trials performed in 12 different countries investigated the effects of PCT-guided antibiotic decision making in the context of respiratory infections [22], [23] compared to aggregate data meta-analysis, patient-level data meta-analysis permitted standardization of outcome definitions and subgroup analyses by type of infection and clinical setting. The study showed that PCT use in the setting of respiratory infection reduces antibiotic exposure (initiation of antibiotics from 86% to 72% and a reduction in overall exposure from 8.1 days to 5.7 days), side effects from antibiotics (decreased from 22.1% to 16.3%) and significantly reduced mortality by 14% (from 10% to 8.6%). Results were consistent for the different clinical settings (i.e. primary care, ED or critical care) and types of infections (pneumonia, bronchitis, COPD exacerbation).

The secondary analysis of PCT cut-offs used in the trials revealed that the main cut-offs used for initiation of antibiotics were adapted to the severity of the patient. Thus, for patients with mild or moderate disease treated in primary care or the ED, a PCT cut-off of 0.25 μg/L was used, whereas for patients with severe disease (e.g. those requiring ICU admission) cut-offs of 0.5 μg/L were mainly used [24]. Similarly, for antibiotic discontinuation a PCT decrease below 0.25 μg/L or 0.5 μg/L were used for mild/moderate or severe illness, respectively, or when PCT showed a drop by ≥80% from the peak [8].

Trials and algorithms for PCT-guided ABS in sepsis

The mentioned concept of using PCT to individualize antibiotic therapy is also appealing in patients with sepsis in the ICU. Yet, initially, there were concerns about safety of this approach in the sickest patients as too short courses of antibiotics could possibly result in the recurrence of infection and higher mortality. The first proof-of-concept study looking at ICU patients with sepsis [14] found a reduction of antibiotics exposure without obvious negative impact on outcome. The trial used the higher PCT cut-offs for sepsis patients (0.5–1 μg/L) and also used the PCT kinetics (decline by >90% from the peak level) to recommend cessation of antibiotic. This concept was later validated in the multicenter, non-inferiority PRORATA trial [17], where PCT use was efficient and safe when used in sepsis patients. Because sepsis patients in ICU have an a priori high risk and time to treatment is crucial, PCT was used mainly for early discontinuation of treatment, but not for guiding empiric treatment [25]. This concept was later also tested in the Stop Antibiotics on guidance of Procalcitonin (SAPS) study where antibiotics were started in all patients with clinical suspicion of sepsis, but were recommended to be discontinued when either the PCT level declined by at least 80% from the peak level and/or when the PCT fell below 0.5 μg/L. Compared to standard of care, the use of PCT resulted in lower antibiotic exposure (antibiotic duration reduced from 7 days to 5 days) and in improved survival (6.1% better survival at 1 year) [7].

A recent patient-level meta-analysis of 11 trials on PCT-guided ABS in critically ill patients with sepsis [26] also found that mortality was significantly lower in the 2252 PCT-guided patients compared with the 2230 control group patients (21.1% vs. 23.7%). This finding was consistent in subgroup analyses stratified by type of infection (respiratory, urinary tract, abdominal, skin or central nervous system), sepsis-3 definition or severity of sepsis (defined by either the sequential organ failure assessment [SOFA] score, the presence of septic shock or renal failure, or the need for vasopressor or ventilatory support). The study also revealed that in sepsis patients, PCT use results in earlier discontinuation of antibiotics with a reduction in treatment duration from 10.4 to 9.3 days, with stronger effects seen in patients with less severe sepsis and those with respiratory infections. A further meta-analysis also revealed, that survival benefit was associated with a PCT algorithm for cessation of antibiotics only, whereas for ICU patients a PCT algorithm for antibiotic initiation or a mixed approach did not reduce mortality [27]. Finally, a meta-analysis focusing on bacteremic patients with positive cultures also found a significant effect of PCT use on antibiotic usage with no evidence for adverse clinical outcomes despite shorter antibiotic durations [28].

Learnings from ‘negative’ studies

Beside the studies mentioned already which showed efficacy and safety of PCT guided ABS, there have also been some studies with negative results.

First, the PASS trial was an interventional study with the rational to improve survival by early antibiotic intervention and escalation any time when PCT is rising. However, although diagnostic and therapeutic measures were escalated in the PCT-guided intervention group, there was no outcome benefit. The lack of effect may be explained by an outbalance of the benefits of earlier treatment with side effects due to escalated antibiotic treatment [29], [30].

Second, the recently published multicenter ProACT trial, which studied the effects of PCT in lower respiratory tract infection (LRTI) in US hospitals, [31] failed to demonstrate any benefit of PCT guidance despite the fact that a similar algorithm was applied as already successfully used in European studies [24]. Several factors may have contributed to this result. Inclusion of patients with very low severity of disease and low likelihood of bacterial infection (extremely low median PCT values, primarily mild CAP, high number of asthma and bronchitis patients), poor adherence to algorithms and study protocols, both for patients admitted and those discharged to primary care, lack of physician inexperience with PCT, and very short overall antibiotic use in both the control and PCT arms [24]. The low experience to PCT use of investigators and a low inclusion rate for the intervention arm (26–111 patients in 2.5 year/center) provided the clinicians with little opportunity to get practical experience with the PCT concept and develop trust in the approach.

Third, the recently published HiTEMP study, investigated the value of PCT to guide antibiotic therapy in ED patients with fever in regard to rates of antibiotic prescription, clinical outcomes and costs [32]. Investigators did not find any added benefit of PCT, demonstrating that while PCT-guided therapy was non-inferior in terms of safety, it did not reduce the prescription of antibiotics. There are several important lessons learned also from this study, which may help to better understand the use of PCT in clinical practice and should be considered when designing future trials [33]. The trial included unselected patients with fever including patients with no diagnostic uncertainty (e.g. patients with skin and soft tissue, intra-abdominal and urinary tract infections). In such patients, PCT has little potential to add value to clinical judgement, as the decision to initiate antibiotics has already been made. The investigators used only a single PCT measurement upon ED admission with a high PCT cut-off (>0.5 μg/L). As the default position in patients with a high pre-test probability for bacterial infection is generally to prescribe antibiotics, a high negative predictive value is mandatory, and thus lower PCT levels <0.25 μg/L or <0.1 μg/L may have been preferable to allow safe withholding of antibiotics. While most previous trials relied on repeated PCT measurements for early discontinuation of therapy, this trial only used one single measurement [24]. Adherence to the study PCT protocol was low with half of patients receiving antibiotics despite low PCT levels illustrating again the need for provider education and feedback when implementing PCT-guided care.

Fourth, the BPCTrea study investigated a PCT algorithm to guide antibiotic treatment for COPD patients admitted to the ICU for ventilator support [34]. Although the 28-day mortality was similar for both the PCT and standard of care group, a higher 3-month mortality was observed for the PCT group. The increase in mortality was most prominent in patients who did not immediately get antibiotic treatment based on the protocol. For patients who received antibiotics immediately the outcome of the PCT group was non-inferior to the standard of care group. These data support the recommendations that in high-risk patients with suspected bacterial infection antibiotic treatment should be started immediately to improve safety of PCT use.

Real world data on clinical integration into ABS programs

It has been shown that the successful implementation of ABS programs requires a holistic approach across the hospital, with education of all stakeholders as a key element. This educational process ideally includes regular auditing, measuring of the success based on clear outcome parameters and feedback to all involved parties to increase trust, confidence and eventually adherence, and achieve the desired effect to reduce antibiotic use and thus antibiotic resistance [35].

This holds true also for an integration of PCT to guide the judicious use of antibiotics in patients suspected of bacterial infection. The low adherence rates observed in some interventional studies such as proACT [31], reflect missing experience with PCT and its interpretation in the clinical context as well as lack of trust in the efficacy and safety of this approach. In contrast, successful, education-based clinical integration of PCT-guided ABS led to reduced use of antibiotics and reduced resistance rates and was associated with improved clinical outcomes like lower re-admission rates, less Clostridium difficile infections and shorter length of stay, without any negative impact on survival [35]. Thus, education may result in increased physician confidence and, when combined with better communication with patients on the risks associated with inappropriate antibiotic therapy, lead to improved algorithm adherence and decreased antibiotic overuse [10], [36].

The results of both, the clinical trials as well as the implementation studies, support the view that the PCT cut-offs and algorithm need to be adapted by the severity and clinical risk of the patient to secure effective and safe use and that education, auditing and continuous feedback are required to develop trust and increase adherence to PCT algorithms.

Derivation of a consensus algorithm for PCT use

Based on the analysis of trials, the group agreed that for optimal use, the PCT levels should be put into the context of the clinical assessment in regard to severity of illness and probability of bacterial infection to make reasonable recommendations. We thus derived three different algorithms based on severity of illness (i.e. mild, moderate and severe illness) and stratified patients according to the probability of bacterial infection (uncertainty vs. bacterial infection highly suspected) (Figures 1–3). PCT should then be added to the assessment of patients with PCT cut-offs of <0.25 μg/L in non-ICU patients and <0.5 μg/L in ICU patients indicating a low likelihood of bacterial infection. While in patients with mild disease and low probability of bacterial infection a low PCT level should advise physicians against the use of antibiotic, for patients with moderate or high severity, empiric therapy may still be used with retesting of PCT after 6–24 h to re-evaluate the need for antibiotic therapy. Further, for patients where empiric antibiotic therapy was started, serial testing of PCT levels is recommended to monitor the response to antibiotic therapy and control of infection. A drop in PCT from the peak by >80% and/or fall below the cut-off was taken as a strong indicator for resolution of illness and earlier discontinuation of antibiotics is recommended when a patient is clinically stable.

Figure 1 shows the proposed algorithm for patients with mild disease (e.g. patients seen in primary care or patients in the ED with bronchitis). The initial step is to evaluate the pre-test probability for a bacterial infection based on the clinical assessment, radiographic assessment and, if indicated, microbiological work-up. In patients with mild disease and diagnostic uncertainty regarding bacterial infection, a low PCT level <0.25 μg/L effectively rules out bacterial infection and there is no benefit from antibiotic treatment. Yet, additional diagnostic tests may be warranted to establish the final (non-infectious) diagnosis. If there is concern that the initial value was negative due to early stages of infection, a second PCT test can be considered within 6–24 h before discharging the patient. If PCT is elevated >0.25 μg/L, the presence of bacterial infection becomes more likely and, if appropriate in the overall clinical context, antibiotic treatment should be initiated. PCT should be re-measured every 24–48 h to discontinue antibiotics when the level falls below 0.25 μg/L or declines by at least 80% versus peak.

In patients with mild disease and highly suspected bacterial infection based on the clinical, radiological and microbiological assessment, a PCT <0.25 μg/L still argues against bacterial infection, but antibiotics may be started based on clinical judgement. Again, the indication for antibiotic use should be re-evaluated when after 24 h with a second PCT test and results of additional microbiological tests. If PCT >0.25 μg/L a bacterial infection is very likely and empiric antibiotics should be started immediately. The PCT level should be re-tested every 24–48 h to assess response to therapy and control of infectious focus. With PCT decline by >80% vs. peak value or fall <0.25 μg/L, the antibiotics can be discontinued if the patient is clinically stable.

Figure 2 shows the proposed algorithm for moderate severity of disease as assessed by clinical scores. The use of PCT in this situation is similar as for mild disease, yet empiric antibiotic treatment in patients with diagnostic uncertainty may still be advised despite low PCT levels to increase safety and adherence of physicians.

Figure 3 shows the proposed algorithm for high severity patients in the ICU with a recommendation for antibiotic treatment in all patients including those with low PCT. Yet, for patients with PCT <0.5 μg/L, further diagnostic testing should be initiated to look for non-bacterial causes of the clinical symptoms or fungal infections. Repeated testing every 24–48 h is recommended to decide on antibiotic discontinuation when PCT levels drop to <0.5 μg/L or by at least 80% from the peak.

For all proposed algorithms, caution was recommended for patients with immuno-suppression (including HIV), autoimmune diseases, cystic fibrosis (CF), pancreatitis, trauma, pregnancy, high volume transfusion and malaria. Also, the algorithm should be used in acute infections, but not in patients with chronic infections (e.g. abscess, osteomyelitis, endocarditis). If patients are pretreated with antibiotics [37], PCT levels may also be low leading to a potential underestimation of infectious complications. For measurements of PCT it should be secured that only high-sensitive PCT assays with sufficient precision in the relevant cut-off ranges are used for optimal safety. Most trials today have used the highly sensitive KRYPTOR immunoassay (BRAHMS, Hennigsdorf, Germany) which has a functional sensitivity of 0.06 μg/L [38], [39]. Thus, the analytical sensitivity, as expressed by the functional sensitivity should be in this range to assure the accuracy of measurement at low PCT levels. In addition, the validation of the assay should assure that the accuracy (bias and imprecision) should be appropriate [34], [35].