Increased procalcitonin indicates antibiotic escalation in ICU patients: an observational study

Background: Empirical antibiotic therapy often fails to cover all pathogens for patients with critical infection without pathogen identification. In these patients, progressive infection can manifest as a “procalcitonin (PCT) alert”. Delayed proper antibiotic escalation could worsen their prognosis. We hypothesized that for these patients, escalating antibiotics after a “PCT alert” would improve their outcomes. Methods: This was a single-center retrospective study including patients with suspected infection who were admitted to Peking Union Medical College Hospital from January 2014 to June 2018. Patients were labelled as “antibiotic escalation” or “nonescalation” according to their antimicrobial use 48 h before and after the "PCT alert". “PCT alert” was defined as PCT ≥1.0 ng/mL that had not decreased by at least 10% from the previous day or from baseline, or a single measurement ≥ 1.0 ng/ml. Indicators that possibly influenced the prognosis were collected. 28-day intensive care unit (ICU)-free days were calculated; ICU stays >20 days and ≤20 days were considered nonprolonged ICU stays (nPISs) and prolonged ICU stays (PISs), respectively. Difference analysis and binary logistic regression were performed to determine the factors that influenced the 28-day ICU-free days . Results: A total of 1109 patients were included, 654 in the PIS group, other 455 in nPIS group. The PIS group had higher rates of pathogen identification (33.94% vs 28.13%, P=0.047) and escalated antibiotic therapy (35.47% vs 20.66%, P<0.001) but a lower proportion of surgical patients (39.45% vs 54.95%, P<0.001) than the nPIS group. Regarding PCT, the values on the 1st day (20.36±43.89 vs 14.89±30.37 ug/L, P=0.014) and on the “alert day” (24.24±46.38 vs 18.75±32.69 ug/L; P=0.021) were higher in the PIS group than nPIS group, but no significant difference in SOFA (PCT, WBC, Tmax, lactate level, dose of NE, SOFA score), st as baseline and alert” advanced monitoring real-time clinical equipment laboratory examination software cooperation

Conclusions Escalating antibiotics in high-risk infection patients whose PCT does not decrease expectedly after administering broad-spectrum antimicrobials may reduce their ICU stay.

Background
The level of procalcitonin (PCT) is regarded as an indicator of bacterial infection [1].
Many studies have confirmed that PCT-guided antibiotic deescalation can decrease the duration and dose of antibiotics without increasing the risk of adverse events [2][3][4][5]. However, most of these studies were randomized controlled trials (RCTs), which are not consistent with real world clinical practices, and focused on deescalation.
For most critical infection patients, especially those with sepsis, broad-spectrum antibiotics are recommended first; deescalation is performed after improvement, stabilization or pathogenic identification [6]. In the intensive care unit (ICU), the proportion of drug-resistant bacteria is much higher than that in general wards [7]. A PCT level that does not decrease as expected in suspected infection patients in the ICU after empirically administering broad-spectrum antibiotics might indicate that the infective microorganisms are not covered by the antibiotics or have some level of drug resistance. Thus, the administered antibiotics should be escalated to higher grade agents, such as vancomycin for gram-positive (G+) bacteria or carbapenems for gram-negative (G-) bacteria.
However, escalating antibiotics on the basis of only increased PCT is not recommended by some guidelines and research [8,9], as it may result in the overuse of antibiotics without outcome improvements. But previous trials were randomized controlled trials (RCTs) and did not conform to the double-blinded principle. Therefore, we conducted this study to explore the significance of antibiotic escalation in critical infection patients with a "PCT alert" in a real-world setting. We excluded patients who were discharged against medical advice, patients who died.

Data collection
The medical records of all the recruited patients were examined to collect information on sex, age, surgery (yes or no), duration of MV, LOS ICU, PCT level, WBC count, maximum daily temperature (Tmax), lactate level, dose of NE (>0ug kg -1 min -1 ), SOFA score, pathogen identification, and antibiotics. For the indicators (PCT, WBC, Tmax, lactate level, dose of NE, SOFA score), we collected data from only two days (1 st day as baseline and the "PCT alert" day). Our department has an advanced monitoring system that records real-time clinical data from bedside equipment and laboratory examination results. This system is maintained by the Donghua software cooperation through the DtHealth system.
The 28-day ICU-free day values were calculated as 28 minus the LOS ICU; the 28day MV-free day values were calculated similarly. The values of the 28-day ICU-/MVfree days in patients who survived or required ventilation longer than 28 days or died within 28 days were regarded as zero. The mean arterial pressure (MAP) was not included in this study because in our department, we correct blood pressure that is lower than normal with vasopressors; this, patients probably had normal blood pressure despite their different illnesses.
In addition to the detailed antibiotic information, data from 48 h before and after the "PCT alert" day were also collected.

Definitions
A "PCT alert" was defined as a PCT level ≥1.0 ng/mL that had not decreased at least 10% from the previous day or from baseline, or a single PCT measurement ≥ 1.0 We referred to a document which was released by Beijing Municipal Health Commission about antibiotic management in Beijing, China. An additional file shows this document in more detail [see Additional file 1]. In that document, every antibiotics was labelled as "no restriction", "restriction" or "only under very special condition" according to their property. "No restriction" means the antimicrobial covers most of the common bacteria without evident sideeffect. We regarded this group as the lowest grade. "Only under very special condition" indicates very strict rules for its super antimicrobial property, which is more likely to induce drug resistance. Besides, theses antibiotics have more serious sideeffects. We considered this group as the highest level. Published research articles [10]were another referential resource. Besides, in our department, a group of physicians, including 4 residents, 4 attending doctors, 2 senior doctors and 2 professors held a meeting to finally come up with the grades of each type of antibiotic ( Figure 1).
With the aim of determining the influence of antibiotic escalation after a "PCT alert", we first defined "escalation" and "nonescalation". The main types of antimicrobials in our hospital were divided into 2 groups according to their antibacterial spectrum: G+ and G-. Information on antibiotic administration 48 h before and after the "PCT alert" was analyzed. "Escalation" was defined as 1) the maximum grades in both groups were escalated; 2) the maximum grade in one group was escalated, while the grade in the other group remained unchanged or was deescalated; or 3) the maximum grades in both groups were unchanged, but the overall types of antibiotics increased. It is worth noting that patients for whom the maximum grades in both groups remained unchanged or were deescalated but the overall types in either group were increased were not included in the escalation group. The patients who were not included in the "escalation group" were included in the "nonescalation" group.

Statistical analysis
Descriptive analyses were performed for the nonescalation and escalation groups.
Continuous variables are expressed as means ± standard deviations, and categorical variables are expressed as absolute values and percentages. For the continuous variables, the data were analyzed using Student's t-test, the Mann-Whitney U test or the Kruskal-Wallis test depending on the data distribution and the number of variables. The categorical variables were analyzed using chi-square or Fisher's exact tests.
Patients who stay longer than 7 days in the ICU are more likely to have complications such as ventilator-associated pneumonia, ICU-acquired weakness, and high long-term mortality [11][12][13]. Therefore, in our study, those with 28-day ICU-free days >20 days were considered the nonprolonged ICU stay (nPIS) group. The parameters between the two groups were compared.
A multivariable binary logistic regression model was built to determine the factors associated with an nPIS. Indicators on the "PCT alert" day, baseline data and demographic factors were included in the model. All comparisons were two-tailed, and P < 0.05 was required to exclude the null hypothesis. The statistical analyses were performed using IBM SPSS Statistics, Version 20.0 (Armonk, NY: IBM Corp.).

Participants
During our study period, 8672 patients were admitted to our department. A total of 7563 patients were excluded for the following reasons: 481 were younger than 18 years old; 1985 were discharged from the ICU within 24 h; 1749 did not receive antibiotics; 2533 did not have a culture performed; 649 did not present a "PCT alert"; and 166 were discharged against medical advice. Finally, 1109 patients were

discussion
The most important finding of our study was that the timely escalation of antibiotics after the PCT reached an "alert" level was beneficial for those in the nPIS group, which contradicts previous conclusions. In addition, this is the first real-world PCT research in China.
PCT has been widely recognized as an indicator of bacterial infection for a long time, and many high-quality studies have demonstrated its significance in guiding antibiotic applications. Most studies focused on when to initiate or stop antimicrobial drugs in patients with infection. However, in the real-world clinical environment, physicians tend to be more cautious, especially regarding critically ill patients or patients at high risk for drug resistance.
Andreas Hohn et al. [5] reported low adherence to a PCT-guided antibiotic treatment protocol in real-world clinical situations involving high-risk sepsis patients if they had insufficiently decreasing or even increasing PCT. This result might reflect the significance of PCT in the real world, which is why we performed this study in a realworld clinical environment, unlike previous studies[8]; we found that the timely escalation of antibiotics when the PCT level does not decrease as expected is a negative factor for PISs (28-day ICU-free days ≤20 days). This important finding leads us to hypothesize that a "PCT alert" might remind medical staff of the high possibility of multidrug resistant germs and that escalating drug grades might improve the outcomes of patients.

Jensen et al.[8] proved that antimicrobial spectrum escalation guided by PCT
prolonged the patients' ICU stays and increased the possibility of organ-related harm. However, there was an important inherent design flaw in this study; the physicians were not double-blinded to the PCT values, which may have led to bias because physicians tend to be "overcautious" if they are alerted to an abnormal PCT level. This means that physicians' decisions regarding transfer or extubation may have been unfairly influenced in the "PCT intervention" group [14].
In our study, PCT tests were performed, and the results were informed, which eliminated bias. To rectify the influence of a patients' baseline condition, we included their baseline information (sex; age; surgery; and the SOFA score, PCT level, Tmax, lactate level, and the P/F ratio on the 1st day after ICU admission) in the binary logistic regression model. The two designs and data analysis methods made the results more reliable.
In the real world, doctors tend to be more scrupulous in adjusting antibacterial drugs, especially for patients in the ICU. It is not only because of the illness but also because of the increased possibility of bacterial drug resistance in ICUs. Peking Union Medical College Hospital is a very renowned hospital in China, with the responsibility of treating complicated and serious diseases, especially those requiring ICU admission. Many of our patients are transferred from other hospitals and have long-term antimicrobial application histories. Therefore, the proportion of multidrug resistant bacteria is increased. Understandably, our physicians tend to escalate antibiotics when patients with infection do not present obvious recovery or present characteristics such as a "PCT alert" after the application of broad-spectrum antibiotics. However, it should be noted that our conclusion might be biased because it is suited to only a certain group of patients. Applying this conclusion practically and scientifically requires further investigation. Despite the bias, a "PCT alert" is still important for informing doctors of uncontrolled infection and the potential necessity of antimicrobial escalation.
In addition to the above findings, we also found that surgical patients tended to have longer ICU stays than patients without surgery, but PIS patients had a lower proportion of surgery than nPIS patients. Our department is a surgical ICU ward, and patients admitted to our department for surgery usually have less severe health conditions than patients who are admitted for nonsurgical reasons. Nonsurgical patients tend to stay longer in the ICU than surgical patients. Surgical site infections (SSIs) account for a large proportion of healthcare-associated infections, which means that surgery could be a risk factor for infection [15][16][17]. Surgical patients who are transferred to the ICU are more likely to have nosocomial infections than surgical patients in non-ICU wards [18]. All of the above factors might influence patients' ICU stays. Therefore, we analyzed the influential variables by building a binary logistic regression model and found that surgery was an independent risk factor for delayed ICU discharge.
Our study is the first to focus on escalating antimicrobials in a real-world environment, and the results are very novel. To determine what to do in the face of a "PCT alert", we analyzed only the indicators on the alert day along with baseline and demographic characteristics. There were some limitations to our study. First, it was a single-center retrospective study, and considering our hospital's specificity, our conclusion on escalating antibiotics according to a "PCT alert" might be restricted to serious patients receiving long-term treatment or hospitals with a high proportion of multi-resistant bacteria. Despite this possible limitation, an unexpected decrease in the PCT level should be regarded as a warning signal for doctors. Second, the patients were very heterogeneous. To reflect real clinical conditions, we did not establish very strict inclusion criteria. This may have led to some deviations and counteracted some of the significance. Therefore, it would be better to further divide the groups into subgroups to analyze the data in depth.
However, because of the limited sample size, the number of patients in some of the subgroups would have been very small, affecting the analysis. We should collect more data to perform subgroup analyses in the future. Third, the definition of nPIS and PIS may have been a too arbitrary. However, for most ICU patients, an ICU stay longer than 7 days increases the possibility of complications and indicates poor outcomes. Therefore, we used "28-day ICU-free days >20" as the cutoff value to distinguish patients with relatively short ICU-free days and thus worse prognoses.
This value might be slightly increased for all ICU patients because this particular group of patients accounted for less than 30% of the patients in our department.
Fourth, despite our large total sample size, as a retrospective study, missing data was inevitable.