BRIEF COMMUNICATION Biomarkers predictive value for early diagnosis of Stroke- Associated Pneumonia Elena Zapata-Arriaza1 , Fernando Mancha1,*, Alejandro Bustamante2, Francisco Moniche1,3, Blanca Pardo-Galiana1,3, Pilar Serrano-Gotarredona4, Silvia Navarro-Herrero4, Esther Pallisa5, Julia Faura2,  Angela Vega-Salvatierra1, Anna Penalba2, Irene Escudero-Martınez1,3, Vıctor Darıo Ramos-Herrero1, Leire Azurmendi6, Jean Charles Sanchez6 & Joan Montaner1 1 Stroke Research Program, Institute of Biomedicine of Seville, IBiS/Hospital Universitario Virgen del Rocıo/CSIC, University of Seville, Seville, Spain 2 Neurovascular Research Laboratory. Vall d’Hebron Institute of Research (VHIR), Universitat Auto noma de Barcelona, Barcelona, Spain 3 Department of Neurology, Hospital Universitario Virgen del Rocıo, Seville, Spain 4 Radiology Department, University Hospital Virgen del Rocio, Seville, Spain 5 Department of Radiology, Vall d’Hebron University Hospital, Barcelona, Spain 6 Department of Human Protein Sciences, University of Geneva, Geneve, Switzerland Correspondence Abstract Alejandro Bustamante, Neurovascular Research Laboratory, Vall d’Hebron Institute To confirm the diagnostic accuracy of candidate biomarkers in stroke-associ- of Research (VHIR), Hospital Universitari Vall ated pneumonia (SAP), we prospectively enrolled ischemic stroke patients with d’Hebron, Passeig Vall d’Hebron 119-129, NIHSS ≥ 10 on admission from March-2016 to August-2017. Blood samples 08035 Barcelona, Spain. Tel: +34 were collected at baseline, 24 and 48 h after stroke onset. Biomarkers (MR- 934894029; Fax: +34 934894015; E-mail: proADM, suPAR, SAA) were determined by immunoassays. Regarding alejandro.bustamante@vhir.org biomarkers, MR-proADM at 24 h (P = 0.04) and both suPAR and SAA at 48 h (P = 0.036 and P = 0.057) were associated with pneumonia. The combination Funding Information of SAA > 25.15 mg/dL and suPAR> 3.14 ng/mL at 48 h had 80% sensitivity This project was partially funded by the ISCIII and 95.8% specificity when both biomarkers were above the cut-off. The evalu- project PI14/00971. The ITRIBiS project ated biomarkers represent promising tools to be evaluated in future large, (Improving Translational Research Potential at prospective studies on SAP. An accurate SAP diagnosis by thorax CT might the Institute of Biomedicine of Seville) has help to reduce variability in such studies. the registration number REGPOT-2013-1. Cooperative Cerebrovascular Disease Research Network (INVICTUS+) (RD16/0019/ 0015). AB is supported by a Juan Rodes research contract (JR16/0008) from Instituto de Salud Carlos III. Received: 25 April 2019; Revised: 24 June 2019; Accepted: 25 June 2019 Annals of Clinical and Translational Neurology 2019; 6(9): 1882–1887 doi: 10.1002/acn3.50849 *Co-First authors. from the Centre for Diseases and Control and Prevention Introduction (CDC)1 by the Pneumonia in Stroke Consensus Group Respiratory tract infections occurring within 7 days of (PISCES).2 However, one of the main limitations in SAP stroke onset have been incorporated under stroke-associ- research is the lack of a gold-standard test for clinical ated pneumonia (SAP) term, after adapted definition diagnosis. Thorax high-resolution computed tomography 1882 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. E. Zapata-Arriaza et al. Correlation with thoracic computed tomography (THRCT) is a useful adjunct to conventional radiography and high spatial frequency double filter. Every THRCT and an accurate study to identify underlying findings in was evaluated by two skilled thorax radiologists, and dis- stroke patients, which could serve as a guide in SAP diag- crepancies reviewed by another expert thorax radiologist, nosis. In fact, we have recently reported that this tool rep- all blinded to the clinical data. Consensus was reached in resents an accurate imaging test for the evaluation of cases with disagreement. clinical criteria.3 In this study, SAP was diagnosed by treating physician Stroke alters systemic immunity, predisposing patients according to diagnostic categories established by the to immunosuppression and infections.4 Therefore, the use PISCES criteria.2 PISCES Probable SAP was defined by of blood biomarkers associated with the systemic inflam- the presence of all CDC criteria1 but in the absence of matory response and the immune changes would be an diagnostic changes on initial CXR and repeated CXR (or interesting starting point for SAP prediction. Some sepsis where CXR not undertaken), and no alternative explana- biomarkers, recently identified as promising predictors of tion or diagnosis. Likewise, PISCES definite SAP was stroke-associated infections,5,6 represents interesting can- diagnosed when all CDC criteria met including diagnostic didates for SAP. Soluble urokinase-type plasminogen acti- CXR changes on at least one CXR. vator receptor (suPAR), participates in immunologic Blood samples for biomarkers analysis were obtained at functions.7 Pro-adrenomedullin (MR-proADM), seems to baseline, 24 and 48 h after stroke onset. Then, samples correlate with the severity of pneumonia.8 And serum were centrifuged at 1500g for 15 min at 4°C and serum amyloid A protein (SAA) with an important role in and plasma were frozen at 80°C until biomarker mea- attracting leukocytes and immune cells to the sites of surement. damage.9 We aimed to confirm the diagnostic accuracy of We measured a 4-blood biomarkers panel including mentioned candidate biomarkers related to the features SAA, MR-proADM, suPAR and CRP. MR-proADM on THRCT.3 (BRAHMS MR-proADM KRYPTORTM- ThermoFisherScientific, MA) and CRP (CardioPhase HS CRP, Siemens Healthineers Spain)were measured in Methods plasma EDTA samples and suPAR (Quantiquine ELISA A cohort of 45 consecutive stroke patients was included kit, R&D Systems, MA) was measured in serum. SAA1 in a monocentric, prospective, observational study from was determined in serum with the Meso Scale Discovery March/2016-August/2017. The study was approved by our (MSD) Vascular Injury Panel-I ECL assay, as per manu- institutional review committee (ID 0103-M1-14). The facturer’s instructions (MSD, Gaithersburg, MD). All patients eligible for the study were those with ischemic biomarkers except MR-proADM were assessed in dupli- stroke confirmed by CT or magnetic resonance imaging cate and the mean intra-assay coefficient of variation (MRI) and NIHSS ≥ 10 on patient admission. Stroke (CV) was <20%. Inter-assay variation was determined onset to inclusion was ≤24 h, and informed consent testing two times in each plate with a commercial internal signed by the patient or a relative was needed at inclu- control (Human serum type AB, male, from clotted, sion. Mechanical ventilation pneumonia of intensive care unit transferred patients and antibiotic therapies within Table 1. Baseline characteristics of the stroke patients that were last 24 h previous to hospital admission were exclusion include in the study and received THRCT. criteria. Every patient underwent a comprehensive inter- view, neurological and general examination on admission Variable n = 41 % and every day after inclusion until the seventh day after Male 21 51.2 admission by the same neurologist blinded to the patient Age (years old) (mean  range) 75 [67–82] study participation. Dysphagia was evaluated within the NIHSS on admission 20 [15–22] first 24 h with the Acute Stroke Dysphagia Screen test.10 Time to probable SAP (PISCES) (d) (Mean [range]) 1.5 [1–2] Chest X-ray (CXR) and laboratory test were run fol- Time to confirmed SAP on THRCT (d) (Mean [range]) 2 [1.5–3.5] Smoker 12 29.3 lowing neurologist criteria during admission. A THRCT Chronic obstructive pulmonary disease 8 19.5 with volumetric technique acquisition was performed, Dysphagia 31 75.6 without contrast, between 5th and 7th day of admission Urinary tract infection 5 12.1 in all included patients, independently of the suspicion of Phlebitis 1 2.4 respiratory infection. CT was performed in a 64 detectors Clinical respiratory infection (PISCES Probable) 8 19.5 rows equipment. Dosage Modulation: 120–178 Kv. Colli- Stroke associated pneumonia on THRCT 5 12.2 mation: 0.5–1 mm, according the equipment. Reconstruc- (Bronchopneumonia) Stroke associated chest infection on THRCT 8 19.5 tion interval: 1 mm. PITCH could be between 0.8 and 1. (Bronchopneumonia + Bronchitis) Acquired data reconstruction was done with soft parts ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1883 Correlation with thoracic computed tomography E. Zapata-Arriaza et al. Sigma-Aldrich), and it was <20%. MR-proADM, as Whitney or Kruskal–Wallis tests. Where relevant, instructed, was assayed by simple. Intra-assay and inter- biomarkers were dichotomized by the cut-off with the assay CV for this method are <10% and <20%, respec- highest predictive accuracy for infections, which was cal- tively. Biomarker measurement was performed blinded to culated in receptor operating characteristic (ROC) curves. clinical information. Sensitivity and specificity for every biomarker cut-off were Statistical analyses were conducted with Statistical evaluated. A P value <0.05 was considered statistically sig- Packages for Social Sciences (SPSS), version 22. Inter- nificant. Given the small sample size and the preliminary group comparisons were performed using Chi-squared nature of the study, P values under 0.1 were considered test for categorical variables and Mann–Whitney U test as trends. for continuous variables. Biomarkers were tested for nor- mality with the Kolgomorov-Smirnoff test. As biomarkers Results were not normally distributed, comparisons on biomarker levels between groups were performed with non-paramet- Of the forty-five included patients, 3 died and one pre- ric tests. Intergroup differences were assessed with Mann– sented with psychomotor agitation that precluded Figure 1. Biomarker level according to clinical diagnosis of SAP following PISCES criteria. Boxplots represent median and interquartile range in both no SAP and probable SAP groups. Just P values <0.1 are represented. CRP, C-reactive protein; SAA, serum amyloid-A protein; MR-proADM, mid-regional proadrenomedullin; suPAR, soluble urokinase-type plasminogen activator receptor. 1884 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. E. Zapata-Arriaza et al. Correlation with thoracic computed tomography THRCT performance. PISCES probable SAP rate was cut-off, and 60% sensitivity and 94.1% specificity when 17.8% (N = 8). There was no patient with PISCES defi- one of them was above the cut-off (Fig. 2). nite SAP according to treating physician criteria. More- over, chest radiography low quality was insufficient to a Discussion correct interpretation of some cases to be considered definitive SAP. Among those mentioned 8 patients, 5 Our results provide various key findings, about the most (62.5%) were identified as bronchopneumonia, 2 (25%) common complication after stroke. Employment of as infectious bronchitis and 1 showed no pathology on THRCT in SAP, could reduce diagnostic discrepancies in CT. Baseline data is shown in (Table 1). According to patients with clinical suspicion and low quality chest PISCES criteria, patients with probable SAP trended to radiography. From the evaluated biomarkers, MR- had higher levels MR-proADM at 24 (P = 0.083) and proADM seems to be the earliest predictor of SAP, when higher levels of suPAR at 48 h (P = 0.095). There were respiratory infection is yet not suspected. The combina- no significant differences for CRP and SAA among any tion of SAA/suPAR improves the predictive value, but at visit (Fig. 1). later time-points. However, regarding THRCT findings, MR-proADM at The poor predictive value of previously SAP studied 24 h (P = 0.040) and both suPAR and SAA at 48 h biomarkers11 and the limited specificity of validated (P = 0.037 and P = 0.057) were associated with pneumo- pneumonia predictive scores12 induce uncertainty in SAP nia. At 24 h, a cut-off of MR-proADM > 1.023 nmol/L prediction and diagnosis. This leads to a late start of the had 80% sensitivity and 72.2% specificity for pneumonia. antibiotic treatment, with the unfortunate consequences At 48 h, both SAA > 25.15 mg/dL (60% sensitivity, associated, or an overuse of antibiotics in patients without 85.7% specificity) and suPAR > 3.14 ng/mL (80% sensi- a clear diagnosis, in order to avoid these fatal outcomes, tivity, 57.1% specificity) were associated with infections. increasing antibiotic resistance in cases without real SAP. The combination of SAA > 25.15 mg/dL and Therefore, the identification of an early reliable biomar- suPAR > 3.14 ng/mL at 48 h had 80% sensitivity and ker, could guide respiratory infection management in 95.8% specificity when both biomarkers were above the stroke patient. Figure 2. Biomarker level according to radiological diagnosis of SAP using thoracic HRCT findings. Boxplots represent median and interquartile range in normal, bronchitis and pneumonia groups. Just P values < 0.1 are represented. CRP, C-reactive protein; SAA, serum amyloid-A protein; MR-proADM, mid-regional proadrenomedullin; suPAR, soluble urokinase-type plasminogen activator receptor. ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1885 Correlation with thoracic computed tomography E. Zapata-Arriaza et al. Our previous study showed for the first time THRCT are compared for the first time with THRCT findings, radiological findings of stroke associated respiratory infec- this adds important strength to our study. tions,3 allowing us to conduct this confirmatory study to In conclusion, among explored biomarkers, MR- test the diagnostic accuracy of candidate biomarkers proADM would have an important role in SAP predic- related to both clinical and radiological diagnosis. In our tion, due to its early elevation in stroke patients suffering cohort, regarding clinical criteria, patients with probable respiratory infections later on. In addition, the combina- SAP had higher levels of all evaluated biomarkers across tion of SAA/suPAR improves the predictive value, espe- the time-points, but without statistical significance. How- cially adding specificity, but at later time-points. Further ever, when SAP was diagnosed according to THRCT find- prospective studies should be developed to assess the pre- ings, we found that MR-proADM at 24 h and both dictive value of these results. An accurate SAP diagnosis suPAR and SAA at 48 h were significantly associated with by thorax CT might help to reduce diagnostic uncertainty pneumonia. These results may suggest a correlation in such studies. between the acute-phase response and development of an infection, which would expand biomarkers scope of use Acknowledgments because of their predictive value in SAP patients, con- firmed by THRCT. To the Pneumonia in Stroke Consensus (PISCES) Group According to our results, it is interesting to highlight and specially to Craig Smith and Andreas Meisel for the insufficient association of tested biomarkers with SAP interesting conversations on the use of chest CT and when they are explored regarding clinical criteria. Such other biomarkers for the diagnosis of post stroke pneu- clinical criteria have shown high sensitivity rates (87.5%) monia. when respiratory infection suspicion is confirmed with THTCT.3 However, high false positive rate of clinical cri- Conflict of Interest teria leads to discrepancies in biomarkers-SAP relation- ship that clearly improves when compared with THRCT None. findings. In fact, with the use of THRCT we would have been able to give a diagnosis of definitive SAP instead of References probable SAP in seven out of the eight patients with clini- 1. Horan T, Andrus M, Dudeck M. CDC/NHSN surveillance cal suspicion. Such results are consistent with low accu- definition of health care-associated infection criteria for racy of clinical criteria; showing reliable results due to specific types infections in acute care settings. Am J Infect THRCT performance in this study. Control 2008;36:309–332. Finally, for post-stroke infections prediction, to define 2. Smith CJ, Kishore AK, Vail A, et al. Diagnosis of stroke- biomarker´s cut-points in patients with SAP would help associated pneumonia: recommendations from the in preventive strategies, especially in early stages. In our pneumonia in stroke consensus group. Stroke cohort, the combination of SAA > 25.15 mg/dL and 2015;46:2335–462340. suPAR > 3.14 ng/mL at 48 h showed 80% sensitivity and 3. Zapata-Arriaza E, Serrano-Gotarredona P, Navarro- 95.8% specificity. Given most SAP cases appears within Herrero S, et al. Chest CT findings and validation of first seven days after stroke onset,2 the availability of clinical criteria of stroke associated pneumonia. J Stroke described biomarkers, even after 48 h, could help in early 2019;21. (Epub ahead of print). detection of respiratory infection in stroke patients, with 4. Meisel C, Schwab JM, Prass K, et al. Central nervous low rates of false negative results. system injury-induced immune deficiency syndrome. Nat Therefore, these new cut-offs should now be tested in Rev Neurosci 2005;6:775–786. larger cohorts, since prospectively validating a cut-off is a 5. Bustamante A, Garcıa-Berrocoso T, Penalba A, et al. Sepsis necessary and relevant approach in the biomarkers biomarkers reprofiling to predict stroke-associated research field that we do not usually do systematically. infections. J Neuroimmunol 2017;15:19–23. The main limitation of the study is that our results are 6. Azurmendi L, Lapierre-Fetaud V, Schneider J, et al. based on a single cohort and our findings need to be vali- Proteomic discovery and verification of serum amyloid dated in an independent and larger cohort before an A as a predictor marker of patients at risk of post- intervention based on these results may be recommended. stroke infection: a pilot study. Clin Proteomics In addition, the small sample size and the low rate of 2017;14:27. events in both clinical and radiological diagnosis pre- 7. Thunø M, Macho B, Eugen-Olsen J. SuPAR: the molecular vented us to perform logistic regression analysis to adjust crystal ball. Dis Markers 2009;27:157–172. by potential confounders, such as stroke severity or dys- 8. Schuetz P, Christ-Crain M, Muller B. Biomarkers to phagia. However, we consider that given laboratory test improve diagnostic and prognostic accuracy in 1886 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. E. Zapata-Arriaza et al. Correlation with thoracic computed tomography systemic infections. Curr Opin Crit Care 2007;13:578– 11. Fluri F, Morgenthaler NG, Mueller B, et al. Copeptin, 585. procalcitonin and routine inflammatory markers- 9. Gabay C, Kushner I. Acute-phase proteins and other predictors of infection after stroke. PLoS ONE 2012;7: systemic responses to inflammation. N Engl J Med e48309. 1999;340:448–454. 12. Zapata-Arriaza E, Moniche F, Blanca PG, et al. External 10. Edmiaston J, Connor LT, Loehr L, Nassief A. Validation validation of the ISAN, A2DS2, and AIS-APS scores for of a dysphagia screening tool in acute stroke patients. Am predicting stroke-associated pneumonia. J Stroke J Crit Care 2010;19:357–364. Cerebrovasc Dis 2018;27:673–676. ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1887