Is Pyroglutamic Acid a Prognostic Factor Among Patients with Suspected Infection? A Prospective Cohort Study

Pyroglutamic acid (PGA) is a compound that accumulates during oxidative stress and hence, elevated levels may be associated with poor prognosis in patients with infection or sepsis. To examine this hypothesis, patients presenting with acute infection were recruited in the emergency department and prospectively followed for 30 days. Sport urine samples were quantified for PGA. Outcomes were mortality and composite outcome of death or organ failure. Thirty two (32%) patients had qSOFA≥2. Median urine PGA was 22.9 (IQR 17.64, 33.53) µmol/mmol creatinine. Four patients demonstrated PGA values ≥ 63 µmol/mmol creatinine. Univariate analysis showed that PGA concentration ≥ 75th percentile (i.e. 33.53 µmol/mmol creatinine) was associated with higher rates of in-hospital mortality (p = 0.041) with similar trend for PGA ≥ 63 µmol/mmol creatinine (p = 0.04). However, multivariate analysis showed that PGA was not associated with worse outcomes, whereas heart rate was associated with both composite outcomes (HR 1.0, p = 0.008 and HR 1.02, p = 0.001 for composite outcome with 30 days and in-hospital mortality, respectively). Among low risk patients, high PGA levels were consistently associated with worse outcomes. In conclusion, urine PGA concentration was not associated with worse outcomes among septic patients. Nevertheless, future studies should evaluate this association in larger cohorts.

www.nature.com/scientificreports www.nature.com/scientificreports/ clearance 16 . The objective of the current study was to examine whether PGA concentration is associated with organ failure and mortality among patients with suspected infection or sepsis.

Material and Methods
Study Design and cohort definition. We conducted a prospective observational study between November 2018 and April 2019 at the Emergency Department (ED) of a large tertiary hospital in Israel. Subjects ≥ 18 years old arriving to the ED with a clinical presentation suggestive of infection defined as: (1) The presence of fever ≥ 38.0 Celsius in the preceding 24 hours and (2) Clinical (e.g. urinary or respiratory complaints), laboratory (e.g. neutrophil left deviation, pathologic urine analysis) or radiographic evidence for infection. Oncologic patients and patients with neutropenic fever were excluded. The study was approved by the institutional Review Board (Helsinki committee of Sheba Medical Center, approval code: 5291-18-SMC) and an informed consent was obtained from all participants or their legal guardians. All methods were carried out in accordance with relevant guidelines and regulations. Data extraction. Data collected for each patient included demographic characteristics, comorbidities, chronic medications, number of hospitalizations in the month prior to index hospitalization and main diagnosis. ED data included hemodynamic parameters (blood pressure, pulse, respiratory rate), laboratory values (complete blood count, blood chemistry panel, blood gases, calculated anion gap, lactate and urine ketones) and time until first antibiotic therapy. For each patient, qSOFA score was calculated. Follow up was defined as the time from admission to death or 30 days. Cases in which patients were discharge before 30 days, the investigator (IG or YPO) contacted them or their family member via phone call.
Definitions and endpoints. Pyroglutamic acid levels. Spot urine samples taken shortly after arrival to ED were tested for PGA concentration using gas chromatograph mass spectrometry. The analyses were performed on a Hewlett-Packard (PaloAlto, CA) HP5890A gas chromatograph coupled to an HP5970B mass-selective detector and an HP59940A ChemStation. Quantification of the acids was based on the specific ion masses 17 . Increased PGA concentrations were defined as: 1. PGA ≥ 63 µmol/mmol creatinine, based on the values described in the general population 18 . 2. PGA ≥ 75th percentile of the sample population.
Outcomes. The primary outcome was 30-days or in-hospital mortality. Secondary outcomes were defined as composite outcome, analyzed separately for the combination of 30 days mortality or in-hospital mortality together with at least one organ failure defined as: 1. Kidney failure -defined as creatinine rise ≥ 150% of baseline. 2. Invasive ventilation. 3. Intravenous amines treatment.
Additionally, an adjusted qSOFA score was calculated by adding 1 point for patients with low qSOFA score (i.e. ≤1) and elevated PGA levels as defined above.

Data analysis.
Comparisons between groups were conducted using paired t-tests or Mann-Whitney U test for parametric and non-parametric analysis, respectively. Categorical variables were compared using Chi-square or Fisher's exact test. Kaplan-Meier analysis was employed to examine the association between PGA levels and mortality. Composite outcomes were analyzed using logistic regression models. The latter included confounders that were demonstrated to be significant in the univariate analysis. Additional sub-analysis was conducted for patients with high and low mortality risk (qSOFA above and below 2, respectively). Finally, qSOFA and adjusted qSOFA area under the Receiver Operating Characteristic (AUROC) curve and its associated terms were calculated and their 95% CI compared. Missing values were omitted from the analysis. All analyses were 2-tailed and P ≤ 0.5 was considered significant. All statistical analyses were performed by using SPSS software (version 25 IBM, SPSS Inc).

Discussion
The present study demonstrates that elevated PGA levels among patients with probable infection and high mortality risk has no major role in predicting worse outcomes. While unadjusted analysis demonstrated significant association between PGA levels, in-hospital mortality and composite outcomes, multivariate analysis revealed that cardiovascular state (i.e. heart rate and plasma albumin) are more important predictors for infection-induced complications. Interestingly, patients with elevated PGA levels were characterized by lower albumin and a trend towards higher urea concentrations. These may reflect their catabolic state and probably disease severity, which is further supported by higher rates of in-hospital mortality. However, rates of patients with qSOFA ≥ 2 did not www.nature.com/scientificreports www.nature.com/scientificreports/ differ between PGA groups. Additionally, no significant association was noted for composite outcome, either with or without mortality.
Normal values of PGA are commonly defined as ≤ 63 µmol/mmol creatinine. However, previous analytical studies among healthy adults reported upper limits as low as 32.6 µmol/mmol creatinine and 54 µmol/mmol creatinine 18 . Since only 4/100 patients in our cohort demonstrated elevated levels, we also evaluated a lower cutoff according to the 75 th percentile (i.e. 33.5 µmol/mmol creatinine). However, the latter did not demonstrated significant association with either outcome.
Our cohort reflected well qSOFA ≥ 2 sensitivity and specificity for in-hospital mortality, reported to range from 56 to 75% with a positive likelihood ratio of 1.6 and 2.3 6 . Interestingly, sub-analysis showed that among patients with acute infection that had not demonstrated organ dysregulation (i.e. qSOFA ≤ 1), higher PGA levels were significantly associated with mortality and worse outcomes, both in the univariate and multivariate analysis. This observation, concluded from the sub-analysis only, may reflect metabolic changes associated with oxidative stress that are yet to be observed clinically and may assist in identifying patients in higher mortality risk. Neverthless, as these observations are preliminary, we believe that its prognostic role cannot be definitively determined. In light of the latter, together with the significant association in the univariate analysis between PGA and in-hospital mortality, we further examined the impact of high PGA values on low qSOFA score and ROC curve terms. This demonstrated to increase the sensitivity to 100% and PLR to 2.7. However, AUROC curve demonstrated significant 95% CI overlap, and by that excluding the additive value of such an adjustment.
Previous studies have shown that increased plasma concentration of oxidative stress markers (e.g. microRNA-25, nitrotyrosine) among septic patients are associated with poor outcomes, further strengthened by evidence for greater antioxidant potential among survivors 9,19,20 . Another in vitro study suggested a pivotal role of glutathione depletion in sepsis generated ROS and endothelial cell damage, with reduction in the latter with NAC pretreatment 16 . Accordingly, several studies have attempted to enhance antioxidant capacity by NAC administration, but reported conflicting results. This may be explained by the fact that NAC was given to all patients regardless of objective evaluation of oxidative stress, and although oxidative stress markers have been already recognized, they are mostly impractical for clinical use. Hence, given sepsis induced glutathione depletion, and the observed PGA accumulation in the glutathione depletion states, we sought to study its association with sepsis complications, with potential future benefit from NAC administration. However, in these patients we found no such clinically significant association. This lack of association may be explained by several limitations in our study: first, patients in our cohort were heterogeneous with respect to severity, with only a few demonstrating hemodynamic compromise. This is reflected by the higher than expected pH, low rate of MAP ≤ 65 mmHg and qSOFA ≥ 2. However, when focusing on patients with a priori low mortality risk, PGA was consistently associated with poor outcomes. This observation, though part of a sub-analysis only, may suggest a prognostic role. Additionally, while our study included a single spot urine sample for PGA, a previous study reported high within-individual PGA variability among healthy individuals, also affected by previous weeks of metabolic stress 21 . However, rates of previous hospitalization due to infection did not differ between groups and hence could partly account for this potential bias. Data on BMI and malnutrition was available only in 27 patients and hence we could not fully adjust the results for this variable. Nevertheless, none of patients had signs of severe malnutrition. Cancer-related cachexia was not a factor as patients with malignant disease were excluded from the study. Hence, we believe that malnutrition did not significantly biased out results. Lastly, as data about PGA concentrations among septic patient is unavailable, our sample could only be defined by the period during which patients were recruited; this may have underpowered our results in order to detect significant association between PGA and worse outcomes. However, in order to overcome this problem several analyses were conducted for different PGA cutoffs, with no significant change on the final result. Of note is the fact that current methods to measure pyroglutamic acid concentration is commonly conducted by GC/MS and cannot be routinely, and quickly, used for evaluation in the emergency room setting. If future studies will also demonstrate the prognostic value of high PGA concentration in patients with acute infection, easy and fast methods for PGA quantification must be evaluated.
In conclusion, our study demonstrates that urine pyroglutamic acid concentration is not associated with worse outcomes among patients with suspected infection and high mortality risk. Nevertheless, as our sample might have been too small, and in the light of the finding among low risk patients, future studies should consider examining this association among larger cohorts.