Measurement of lactate in pleural fluid rapidly identify infection and guide therapy

Abstract Background Measurements of pleural fluid biomarkers for rapid identification of complicated parapneumonic effusion (CPPE) are crucial for optimal management. Previous studies for biomarker evaluation were however based on pleura culture, not modern DNA technique. Lactate has not been thoroughly studied earlier as a potential biomarker in this regard. Objectives To evaluate whether the routine biomarkers pH, glucose, lactate dehydrogenase (LDH) measured in pleural fluid in a microbiological well characterised cohort could differentiate simple parapneumonic effusion (SPPE) from CPPE and if pleural fluid lactate could be of additional use in this discrimination. Methods Pleural fluid prospectively collected from adult patients (n = 112) with PPE admitted to the Departments of Infectious Diseases (DIDs) at four Stockholm County hospitals were characterised microbiologically with bacterial culture and 16S rDNA sequencing, and biochemically with pH, glucose, LDH and lactate. Results Forty and seventy two patients were categorised as SPPE/CPPE. The median values between SPPE/CPPE differed significantly for all biomarkers with varying overlap. Receiver operating characteristics (ROC) curves showed the area under the curve (AUC) for pH 0.905 (CI 0.847–0.963), glucose 0.861 (CI 0.79–0.932), LDH 0.917 (CI 0.860–0.974) and lactate 0.927 (CI 0.877–0.977), corresponding to best cut-off levels and sensitivity/specificity for pH of 7.255, 0.819/0.9, glucose 5.35 mmol/L, 0.847/0.775, LDH 9.8 µcat/L, 0.905/0.825 and lactate 4.9 mmol/L, 0.875/0.85. Conclusions To distinguish between SPPE/CPPE, pH and LDH performed well, but optimal cut-off values differed from earlier established recommendations. Pleura lactate had the largest AUC of the investigated biomarkers and may be used in the analyses of PPE-staging.


Introduction
Up to 40% of patients with a community-acquired pneumonia (CAP) will develop a parapneumonic effusion (PPE) [1]. In the majority of the cases no bacteria can be identified in the exudate and the infection can resolve on antibiotic treatment, classified as uncomplicated or a simple PPE (SPPE) [2]. Bacteria may though translocate into the pleural space and progression into a fibrinopurulent stage will occur classified as a complicated PPE (CPPE) with risk of development into an empyema, a complication with high morbidity [3] and a mortality rate reported up to 15% [4,5]. Early diagnosis of a pleural space infection and rapid initiation of effective treatment represent therefore a keystone principle for managing patients with PPE [6].
The ability to identify the causing pathogen in pleural fluid through culture is however low due to a combination of prior antibiotic treatment, low rate of viable bacteria, preanalytical error and technical difficulties [7]. DNA sequencing protocols have been developed sensitive enough to detect low titres of microorganisms in otherwise sterile fluids including pleura fluid [8][9][10]. In a prior study on the patient cohort described in this study including adults with CAP and PPE, the 16S rDNA sequencing method was superior in detecting and classifying bacteria species in the pleural space compared to commonly used culture [11].
Microbiological tests are though time consuming and therefore limited as a decision tool in acute or subacute disease courses. Evaluation of alternative simple biomarkers analysed in pleura fluid has shown pH, lactate dehydrogenase (LDH) and glucose to best discriminate between SPPE and CPPE [2,12] where pH have been presented as the biomarker with the highest diagnostic accuracy [6]. However, these conclusions were based on studies where culture was the only bacterial characterisation method, entailing a potential risk of misclassification regarding PPE-stages which might raise questions regarding the reliability of these results. Further, due to preanalytical factors that affect the result pleura fluid pH is complicated to sample [13].
Pleural fluid lactate hasin contrast to pHduring the last decades become increasingly easier to analyse. It has prior been suggested to be a useful biomarker when studying PPE [14][15][16][17][18][19]. However, these studies are decades old and focussed mainly on lactate as a marker to differentiate between empyema and other pleural diseases. There are very few recent publications [15], and we are not aware of any modern study with the aim to assess the diagnostic value of pleural fluid lactate to discriminate between SPPE and CPPE.
The aims of this study were as follows; (1) in a microbiological well-characterised cohortusing 16S rDNA sequencing methodsevaluate the ability of pleural fluid concentrations of pH, glucose and LDH to differentiate SPPE from CPPE and (2) study if lactatein comparison with other biomarkerswould be useful in the characterisation of different PPE-stages.

Patients' characteristics and ethical approval
This prospective study was conducted at the Departments of Infectious Diseases (DIDs) at four hospitals in Stockholm County: the Karolinska University Hospital Solna, the Karolinska University Hospital Huddinge, and the community teaching hospitals Danderyd Hospital and S€ odersjukhuset. All admitted patients at the DIDs that underwent thoracentesis for pleura culture during 2011-2014 were evaluated consecutive for inclusion in the study by two of the study investigators (NJ, JH). A group of 197 adult patients with PPE due to CAP were finally included in the study [11]. All enrolled patients or nearest relative provided written informed consent. The study was approved by the regional ethics committee in Stockholm, Sweden.
The inserted chest tube for pleura fluid drainage enabled sampling for microbiological and biochemical testing i.e. microbiologically characterised with both bacterial culturing and 16S rDNA sequencing and biochemically analysed for pH, glucose, LDH and lactate. Only patients with complete testing (n ¼ 112) were included in this part of the study. For additional information about the cohort see Figure 1.
Patients were categorised as CPPE or SPPE. CPPE was defined as fulfilment of one of the following criteria: (1) findings of bacteria in culture and/or 16S rDNA sequencing in pleural fluid, (2) frank pus in pleural fluid and (3) thorax surgery with decortication [11,20]. SPPE was defined as lack of all these criteria.

Clinical data
Medical outcome and relevant clinical data were collected using medical records available at the time of admission (Table 1).

Specimen collection and analytical procedure
Pleural fluid was collected at the time of thoracentesis and relevant information regarding procedure, macroscopic appearance of the fluid, etc., was noted in the medical records. The study profile is shown in Figure 1.

Microbiological evaluation
Microbiological samples were collected in culturing bottles and sterile test tubes for traditional culture and 16S rDNA sequencing, respectively. The culture procedure has been described earlier [11].
16S rDNA sequencing analysis A modified protocol for 16S rDNA sequencing analysis was used [21]. In brief a $460 bp region of 16S rRNA gene for the bacteria was amplified by real-time PCR followed by a chemical purification protocol (ExoProStar) and standard dideoxy nucleotide sequencing based on  were forward 5 0 >CGGCCCAGACT-CCTACGGGAGGCA GCA <3' and reverse: 5 0 >GCGTGGACTACCAGGGTATCTAAT CC <3 0 . For bacterial identification and typing the obtained DNAsequences were analysed by standard BLAST database search and RipSeq 16S sequence analysis [22].

Chemistry evaluation
Biochemical samples were collected as follows: glucose and lactate in collection containers coated with sodiumfluoride/potassium-oxalate (367922, BD Vacutainer), pH anaerobically in 50 mL adhesive dispenser syringes and LDH in lithium heparin coated collection containers (367962, BD Vacutainer). Collected samples were analysed within 2 h at the routine biochemical laboratory at the hospital where the patient was admitted. The analyses were conducted on the following platforms: (1) pH (ABL800, Radiometer A/S), (2) glucose, lactate and LDH (Cobas8000, Roche Diagnostics, Basel, Switzerland).

Statistical analyses
Wilcoxon's test was used in the comparison of median values for the four investigated biomarkers between the groups SPPE and CPPE, respectively. The analyses were performed with the JMP statistical computer program (SAS Institute, Cary, NC). A p value of <0.05 was judged as significant.
Each of the four investigated biochemical variables was presented with mean, median and range subdivided for SPPE/CPPE groups. Distributions of these variables were further compared between the two PPE-groups with Wilcoxon Sum-Rank Test. To analyse the pairwise magnitude of association between variables Spearman Rank Correlation was applied. To illustrate and understand how well the four biochemical variables are to predict CPPE from SPPE ROC analysis were performed. The area under curve (AUC) was presented for each method along with the optimal cut-point value. Criterion for the point was defined as the point on the PROC curve closest to the point (0,1), i.e. upper left corner of the unit square. Sensitivity and specificity for the optimal criterion cut-point value were computed.
Analyses were performed in R version 3.5.3 (R Core Team (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.) and p ROC (1.14.0) [23] package was used to perform ROC analysis.
Fisher's exact test was used for comparison of nonparametric categorical values.

Results
In this analysis, 112 patients were included as described above. The cohort included 74 males and 38 females with a mean age of 64 years. A total of 40 patients were categorised to the SPPE group while 72 cases were identified as CPPE. More than 2/3 of the study population had underlying diseases. No significant differences in comorbidity were seen between the groups SPPE and CPPE (Table 1). In total 88 bacterial pathogens were found in pleural samples by culture and/or with 16S rDNA sequencing ( Table 2). The most common findings were viridans streptococci, Fusobacterium spp. and Streptococcus pneumoniae.

Biochemistry markers and different stages of PPE
In the cohort median values within the groups SPPE/CPPE were: pH 7.42/7.01 (p < 0.0001), glucose 6.2/1.25 (p < 0.001), LDH 4.4/43.7 (p < 0.0001) and lactate 2.3/10.9 (p < 0.0001) and the individual results distributed as graphically shown in Figure 2. Calculated p values corresponds to the probability for a separation between the two groups for the biomarkers, respectively.
The covariance for pH, glucose, LDH and lactate was also studied showing a high correlation between the biomarkers for SPPE and CPPE groups, respectively, but the variability between individual patients was large (Supplemental material).

ROC curve analysis and optimal cut-points
To estimate optimal cut-off values and the sensitivity/specificity for the discrimination between SPPE and CPPE, receiver operating characteristic (ROC) curves were performed for all four biomarkers as shown in Figure 3.
The When comparing AUC for lactate with other biomarkers a significant difference was seen in the case of lactate vs. glucose; 0.927 vs. 0.861, p ¼ 0.008, but not for the other parameters.
In the total cohort (n ¼ 197), 160 patients had complete microbiological characterisation, but incomplete sampling for biochemistry. When comparing the biomarkers between the SPPE/CPPE groups in this cohort the results were similar to the findings in the cohort of 112 patients with complete sampling for both microbiology and biochemistry (supplemental material).

Comparison of SPPE/CPPE and ROC analysis with optimal cut-pointspatients with empyema excluded
In the studied population nine patients had findings of frank pus when thoracenteses were performed. After having excluded these nine patients a subgroup analysis was done for the remaining population (n ¼ 103). In this cohort, the median values within the groups SPPE/CPPE were for pH 7.425 (range 7.78-9.91)/7.00 (range 6.31-9.97) (p < 0.0001), glucose 6.

Discussion
Development of PPE from a more benign stage is a dynamic process where the initially sterile pleural fluid become invaded by bacteria translocated from the lung tissue and consequently open up for a confined infection to advance into a more serious condition. A delayed correct diagnose can within a few days change the course of disease to develop into a more complicated path resulting in a more advanced stage of pneumonia, implicating a higher risk for morbidity and mortality [3,24] and in need of more invasive actions. An important challenge is therefore to find biomarkers specific as well as sensitive enough to momentarily identify patients with a non-purulent but nevertheless infected pleura fluid so that adequate interventional procedures and pharmacological treatment prompt can be initiated. The four biomarkers pH, glucose, LDH and lactate studied in this cohort were selected as they mirror the additional metabolism in the pleural fluid due to microbial presence. In this study, we showed a high covariance between the biomarkers in the correlation analyses, but they were not totally exchangeable.
For the established biomarkers pH, glucose and LDH we found a significant difference between the SPPE and CPPE groups in accordance with previous studies [25][26][27], but there was an overlap between the groups resulting in a grey area where straight decision-making is difficult to implement. Similar data have previously been described [26,28,29].
The American College of Chest Physicians (ACCP) as well as the British Thoracic Society have in their guidelines stated a clinical decision limit for pH at 7.20 as an optimal value to discriminate SPPE from CPPE [2,30,31]. An implementation of this cut-off value for pH to our cohort will give a high specificity but in contrast a low sensitivity (76.4%) leading to the consequence that every fourth patient with a CPPE would have been unidentified. We observed similar trends for the established clinical decision limits for glucose at 3.4 mmol/L and LDH at 1000 IU/mL (16.67 mcat/L) with even lower sensitivity compared to pH. This implicates those previous investigations [2,6] had put the threshold too low for pH and glucose and too high for LDH compared with our results. One possible reason for this could be due to misclassification when differentiate disease stages if only having access to culturing and not 16S rDNA sequencing. Further microbiological well-characterised studies are needed for investigation of optimal cut-off levels regarding these biomarkers.
A metanalysis [6] showed that pH, compared to glucose and LDH had the highest diagnostic value for staging patients with PPE. Yet, instability of the sample gives a potential risk of affected results due to preanalytical error. So if clinical decisions are to be based upon pleural fluid pH, the pleural fluid has to be withdrawn and contained without exposure to air, preferable into a syringe, kept in ice slurry and measured by a blood gas analyser within 2 h [13]. Glucose samples must also be chilled if containers without glycolysis inhibitors are used. Regarding LDH, samples must be centrifuged within 4 h and are then stable at room temperature for up to 7 d [32]. The turnaround time for LDH and glucose are long due to transportation, preparation, time on analytical instruments and result delivery making them less applicable in urgent situations [33].
The pathophysiology for the development of an acidotic environment and the formation of lactate are well connected, triggered by an anaerobic metabolism [34,35]. Besides, most anaerobic bacteria produce lactate, D-or L-isomers or both. Lactate has also been identified to be a robust and quick biochemical marker to analyse in pleura [19,36,37]. However, only a few studies have been published regarding pleura fluid lactate and PPE which may be explained by tradition, costs, lack of resources to perform the analysis and availability to an essay that only measures L-lactate. Today essays both on routine automation and blood gas machines measure the sum of D-and L-lactate and the availability of the test has increased dramatically as well as the price per test has dropped. Despite that lactate has become an important prognostic marker in emergencyand intensive care [38,39], the validity of using lactate as a biomarker in PPE has not been thoroughly investigated in earlier studies and a cut-off level regarding lactate to separate SPPE from CPPE has never been established. Santotoribio et al. found that measurement of the lactate concentration in patients with pleural effusions could be useful to differentiate between SPPE and CPPE [15]. In that study though, the central objective was to measure the accuracy of pleural fluid lactate concentration for diagnosis of PPE vs. pleuritis of other origin and there were few patients with SPPE (n ¼ 10) and CPPE (n ¼ 20).
In the present investigation we found, in accordance with the other biomarkers studied, a significant difference in lactate median values between the patients with SPPE and CPPE, respectively. As measured by the area under the ROC curve pleural fluid lactate had the same high diagnostic accuracy in discrimination between SPPE and CPPE as compared with pH and LD, and significantly higher in comparison with glucose.
As stated above, the most important function for a pleural biomarker is to identify a patient with CPPE in the event of clear pleural fluid to minimise the delay for intervention. To check the robustness of the cut-off values an analysis of the subpopulation without empyema was done for all studied biomarkers showing an overall slightly lower AUC compared to the whole study population. Notably, in this subpopulation pleural lactate still had the largest AUC for separating SPPE from CPPE.
Previous investigations have often found high specificity but lower sensitivity when using isolated biomarkers [12,40], an aggravating factor considering the poor prognosis among patients with CPPE and empyema [3,24]. In this study, a peak AUC in the ROC analysis for pH also correlated to high specificity but not with sufficient sensitivity, with almost one fifth of all CPPE-patients undetected. For LDH the results were contrariwise with a high sensitivity but with an inferior specificity, findings similar to what Jimenez Castro et al. found [29]. A peak AUC for glucose corresponded to a lower sensitivity compared to both LDH and lactate but foremost with a markedly low specificity. For a pleural lactate value of 4.9 mmol/L the sensitivity was highonly slightly inferior to LDHand still with a high specificity. According to the results in this study, together with the robust and quick performance when analysed in pleura, lactate might be superior in comparison to the other three biomarkers studied to optimise CPPE diagnosis.
In the present cohort of 72 patients with CPPE pleura culture identified a pathogen in only 17 cases. In 54 patients bacteria were only detected with 16S rDNA sequencing. Using pleura lactate and the cut off value of 4.9 mmol/L would result in that more than 4/5 of these cases (45/54) could rapidly be identified having a CPPE. Many hospitals do not have access to DNA sequencing methods in a clinical setting but only pleura culture, entailing a risk of delay in PPE-staging, and by extension a poorer clinical development. In these settings, pleura lactate could be decisive for rapid diagnosis of a CPPE condition with potentially major impact of the clinical management such as broader antimicrobial therapy directed against bacteria common in CPPE with a longer duration as well as optimal drainage of the pleural space at an early stage.
A limitation of this study is that there may be a theoretical risk of misclassification of patients between SPPE and CPPE. However, the use of novel molecular techniques such as 16S rDNA sequencing to detect and classify bacteria in the pleural fluid in this study would help to diminish the risk for misclassification compared to earlier similar studies. Further, sample collection was not complete in all included patients which is a recurrent problem in most prospective studies especially when conducted in an acute setting. However, comparison of data in the group with complete sampling versus the whole study population showed similar results. Besides, none of the patients or nearest relatives denied participation in the study. Selection bias in the collection of samples is therefore unlikely. Finally, due to the lack of a golden standard the clinical relevance of bacterial DNA findings is difficult to interpret in all clinical sequencing trials, also in those where the samples are collected from a sterile location. However, the fact that the concordance between positive culture and DNAfindings in PPE was high and that the 16S rDNA-analyses were negative in the control-group [11], would support the opinion that most of the DNA-findings had a clinical significance.
Strengths of this study include the prospective design conducted at four different academic hospitals in a microbiological well-characterised cohort using 16S rDNA sequencing methods. The concept of measuring lactate levels collected from pleural fluid in patients with pneumonia is available in most hospital clinical laboratories, easy to perform and appears to translate into a clinically useful tool.
In conclusion, optimal cut-off points for pH, LDH and glucose to separate SPPE from CPPE, differed substantially from those stated in earlier established recommendations. Glucose also performed insufficient as a biomarker from this aspect. Pleural lactate showed high sensitivity and specificity in the ability of SPPE/CPPE-discrimination and performed equal or better compared with the other investigated biomarkers. As pleural lactate is easy and robust to measure compared to pH, we suggest that lactate should be included in the analyses regarding the evaluation of PPE-staging.