Prognostic impact of plasma lipids in patients with lower respiratory tract infections – an observational study

Principles: A decrease in plasma lipids occurs during severe sepsis and has prognostic implications in critical illness.Whether lipids have prognostic implications or could help to differentiate community-acquired pneumonia from other lower respiratory tract infections remains unknown. Methods:We analysed data from patients with lower respiratory tract infections enrolled in four prospective trials. We studied the time courses of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C) and triglycerides (TG) and compared them with the underlying diagnosis and medical outcomes. Results: Of 572 patients included, 372 had community-acquired pneumonia and 200 acute and exacerbations of chronic obstructive bronchitis. We found significantly lower concentrations of TC, LDL-C and HDL-C in all patients on admission as compared to hospital discharge, particularly in community-acquired pneumonia. A multivariate logistic regression analysis including HDL-C, CRP, age and diabetes showed that HDL-C (OR: 0.18 [95%CI 0.11–0.3]) and CRP (OR: 1.01 [95%CI 1.01–1.02]) were independent predictors of community-acquired pneumonia. TC levels were significantly lower in non-survivors than in survivors (3.26 mmol/L [95%CI 2.58–3.96] vs 3.78 mmol/L [95%CI 3.01–4.65]). The prognostic accuracy, defined as the area under the receiver operator characteristic curve of TC to predict mortality, was 0.63 (95%CI 0.53– 0.72) in all patients and increased to 0.94 (95%CI 0.86–1.00) in patients with bacteraemic community-acquired pneumonia. Conclusions: In conclusion, low lipid levels, particularly low HDL-C, pointed to bacterial infection and low TC was predictive of adverse outcomes in patients with lower respiratory tract infections. Reflecting the severity of disease, plasma lipid levels may be a complementary tool in the diagnostic and prognostic workup of patients with lower respiratory tract infections.


Introduction
Funding research support: All authorsdeclare that they have no competing interest.
bility to bacterial superinfection.Bacterial CAP is the most important precursor of sepsis and the major infection-related cause of death [7].In the absence of chest x-rays, clinical signs and symptoms of CAP and routine laboratory tests are not specific and sometimes misleading [8,9].Diagnostic markers discriminating bronchitis from CAP facilitate the diagnostic workup of patients and provide guidance for antimicrobial treatment [10][11][12][13][14].In addition, assessment of disease sever-ity and prediction of outcome in patients with LRTI is crucial for decision-making on patient management, such as the need for hospital or intensive care admission and suitability for hospital discharge [14,15].
In this study we compare plasma concentrations of plasma lipids in 572 patients with LRTI and study the diagnostic and prognostic impact of TC, LDL-C, HDL-C and TG.

Setting and studypopulation
Data from four prospective studies enrolling patients between December 2002 and April 2007 with diagnosed LRTI, who presented to the emergency department of a 950-bed tertiary care centre in Basel, Switzerland, were pooled and analysed [10,[12][13][14].The design of the trials was similar,a nd ac omplete detailed description can be found elsewhere [10,[12][13][14].In brief, in all trials consecutive patients with LRTIs including acute and exacerbation of chronic obstructive bronchitis (COPD) and asthma, and CAP,w ere randomly assigned to procalcitonin (PCT)-guided antibiotic therapy or to standard treatment according to guidelines.The aim of all trials was to study the impact of procalcitonin-guided antibiotic treatment.The first three trials studied antibiotic exposure in different LRTI entities as the primary endpoint, while the fourth study was powered to demonstrate noninferiority for adverse medical outcomes in amulticentre design.For this last trial only patients from University Hospital Basel were included.For all studies, CAP was defined as the presence of anew infiltrate on the chest radiograph accompanied by one or more acquired acute respiratory symptoms and signs such as cough, sputum production, dyspnoea, fever above 38.0ºC, auscultatory findings of abnormal breath sounds and rales, leucocytes higher than 1x10 10 cells/L, or leukopenia below 4x10 9 cells/L [15].Exacerbation of COPD was defined by postbronchodilator spirometric criteria according to the GOLD guidelines as aFEV1/FVC ratio below 70%, and severity categorised according to GOLD criteria [16,17].Acute bronchitis was defined as LRTI in the absence of underlying lung disease or focal chest signs or infiltrates on chest x-ray respectively [18].Patients with cystic fibrosis, active pulmonary tuberculosis, hospital-acquired pneumonia and severe immunosuppression were not eligible for inclusion.
Patients were examined on admission to the emergency department by am edical resident supervised by a board-certified specialist in internal medicine.Baseline assessment included collection of clinical data and vital signs, comorbid conditions and routine blood tests.The laboratory workup for patients with CAP included blood cultures and collection of urine samples for detection of Legionella pneumophila [12].
All trials were approved by the local ethical committees and registered in the Current Controlled Tr ials Database.All patients gave written informed consent.
The intention of this post hoc analysis was to study the value of plasma lipids in assessing prognosis, severity and differentiation of CAP from other etiologies in awell defined cohort of patients with diagnosed LRTI.For the purpose of this analysis we used available clinical and laboratory data of asubset of 572 out of atotal of 853 patients (67%) in whom plasma lipids including TC, HDL-C, LDL-C and TG concentrations were measured on admission and during follow up as part of the routine laboratory workup.Records of all patients included were reviewed to retrieve complete laboratory data on admission, on day 3-5 and on hospital discharge, and additional information on comorbidities and comedications which potentially influence plasma lipid levels, such as the presence of diabetes or statin and corticosteroid use.To compare lipid levels with other well known markers of inflammation, we used C-reactive protein (CRP) and albumin concentrations from the routine laboratory assessment.

Outcome
Adverse medical outcome for this analysis was defined as an eed for intensive care unit (ICU) admission during index hospitalisation, and death from any cause until follow-up.Patients who survived until follow-up were counted as survivors, and patients who died within the follow-up period as non-survivors.

Statistical analysis
The manuscript was prepared according to the STROBE statement [19].Discrete variables are expressed as counts (percentage) and continuous variables as medians and interquartile ranges (IQR) unless stated otherwise.Frequency comparison was performed by chisquare tests.Tw o-group comparison of normally distributed data was performed by Students t-test.For multigroup comparisons, one-way analysis of variance with least square difference for posthoc comparison was applied.For data not normally distributed, the Mann-Whitney-U test was used if only two groups were compared, and Kruskal-Wallis one-way analysis of variance if more than two groups were compared.Logistic regression models and receiver-operating-characteristics (ROC) with sensitivities and specificities for different cutoff values were calculated using STATA9 .2(Stata Corp, College Station, Te x).Thereby,outcomes were either survival until follow-up or adverse medical outcome including need for ICU admission and death until follow-up respectively.C orrelation analyses were performed using Spearman rank correlations.All testing was two-tailed and p values below 0.05 were considered to indicate statistical significance.
As ystemic inflammatory response syndrome (SIRS) as defined by the American College of Chest Physicians and the Society of Critical Care Medicine [20] was found in 70% of the patients (4, 3and 2criteria in 56, 149 and 195 patients respectively).Bacteraemic CAP was diagnosed in 40 patients (7.0%),where growth of bacteria in blood cultures was documented (Streptococcus

Outcomes
Adverse medical outcomes were observed in 97 patients, including 71 admissions to the ICU and/or 38 deaths.All patients in this analysis were treated as inpatients with am edian hospital stay of 13 days (IQR 9-18).

Lipid levels on admission and followup
In all patients with LRTI, significantly lower levels of TC (p <0.0001), LDL-C (p <0.0001) and HDL-C (p <0.0001) were found on admission as compared to hospital discharge.Changes of plasma lipids were more pronounced in patients with CAP and particularly in bacteraemic CAP (fig.1).These changes were also observed for acute phase proteins such as CRP and less pronounced for albumin.

Diagnostic value of lipid levels on admission
To study whether plasma lipids could differentiate viral LRTI from CAP,patients were separated into CAP (n =3 32), blood culture positive (bacteraemic) CAP (n =4 0) and other LRTI (n = 200) including exacerbation of acute and chronic Correlation analysis showed asignificant negative correlation of the acute phase protein CRP with HDL-C (r 2 =-0.25) and ap ositive correlation with albumin(r 2 =0.24),andonly weak corre-lations with LDL-C, TC and TG.In addition, there was aw eak correlation between HDL-C and the CURB-65 score (r =0.17, p=0.001) and the pneumonia severity index (r =0 .11,p=0 .04)[21,22].

CAPN on-CAP
To investigate whether plasma lipids were able independently to discriminate CAP from other LRTI, and bacteraemic CAP from non-bacteraemic CAP respectively,w ec alculated multivariate regression logistic models entering each single lipid value, CRP,t he presence of diabetes and statin use as covariates.HDL-C was the best discriminator of CAP from other LRTI (OR: 0.18, p<0.0001) and the best predictor of bacteraemic CAP (OR: 0.36, p=0 .04).Ta ble 2s hows the respective odds ratios and significance levels of all covariates.
To assess the overall diagnostic accuracy of the different parameters to differentiate CAP from other LRTI, we calculated receiver operating characteristics (ROC) curve analyses.Of all plasma lipids, HDL-C had the highest diagnostic accuracy to diagnose CAP (AUC: 0.82 [95%CI 0.78-0.85])and to predict bacteraemia (AUC: 0.74 [95%CI 0.65-0.83]),which was within the range of CRP (AUC: 0.85 [95%CI 0.82-0.88],p= 0.14 and AUC: 0.73 [95%CI 0.65-0.81],p=0.77) and significantly higher as compared to leucocytes.Details of the ROCs tatistics are presented in table 3.At an HDL-C cutoff of 0.6 mmol/L, the sensitivity (specificity) to diagnose CAP and positive blood cultures was 80% (67%) and 91% (36%), with ap ositive likelihood ratio of 0.11 and 0.29.Conversely,the specificity (sensitivity) at ac utoff of 2.0 mmol/L to exclude CAP and positive blood culture was 80% (62%) and 58% (56%) negative likelihood ratio of 9.8 and 3.4 respectively.R OC curve analysis of plasma lipids and CRP to predict positive blood cultures in the subgroup of patients with CAP revealed similar results, particularly non-significant differences between HDL-C and CRP.

Lipid levels on admission as prognostic indicatorsfor outcome
We assessed the predictive value of plasma lipids on admission to predict death and adverse medical outcome (death and/or ICU admission) respectively.P atients with not surviving LRTI had similar HDL-C and TG levels on admission to survivors, but significantly lower TC (3.Likewise, patients with adverse medical outcome had lower TC (p =0 .03)and LDL-C (p =0 .04)and similar HDL-C (p =0.18) and TG (p =0.14) concentrations.Am ultivariate logistic regression analysis including the presence of diabetes, statin use and CRP as covariates showed that TC and LDL-C, but not CRP,w ere each independent predictors for mortality (OR: 0.63 [95%CI: 0.45-0.90],p=0.01 and OR: 0.56 [95%CI: 0.35-0.90]p=0 .02).As imilar analysis showed that lipid concentrations were not predictive of adverse medical outcome.AR OC curve analysis revealed AUCs of 0.63 (95%CI: 0.53-0.72)and 0.60 (95%CI: 0.50-0.71)for TC and LDL-C to predict death and AUCs of 0.57 (95%CI 0.50-0.63)and 0.56 (95%CI 0.50-0.63)for TC and LDL-C to predict death and need for ICU admission.
In addition, we repeated the same analysis including only patients with CAP with and without bacteraemia respectively.T he prognostic accuracy in the subgroup of CAP patients was similar to all LRTI patients.In bacteraemic patients, however,t he AUCo fL DL-C and TC increased to 0.94 (95%CI: 0.82-1.00)and 0.96 (95%CI: 0.89-1.00).

Discussion
This study investigated changes in plasma lipid levels in patients with LRTI, and particularly whether plasma lipids were different in bacterial and non-bacterial LRTI, and whether they have prognostic implications.We found markedly lower levels of TC, LDL-C and HDL-C on admission, with recovery until discharge in patients with LRTI.These changes were most pronounced in patients with CAP,e specially those with bacteraemia.In am ultivariate regression analysis HDL-C showed high diagnostic accuracy in differentiating CAP from other LRTI, and predicted subsequent bacteraemia in patients with CAP.I na ddition, low TC and LDL-C, but not low HDL-C concentrations, were independent predictors of short-term mortality in LRTI, with high prognostic accuracy in bacteraemic patients.
In accordance with these experimental findings the present study demonstrates that in unselected patients with less severe systemic infections cholesterol values on admission are significantly altered depending on disease severity.R emarkably,t his clinical study demonstrates that plasma lipids correlate with the likelihood of bacterial LRTI and growth of bacteria in blood cultures.If confirmed in further studies and combined with other prognostic clinical and laboratory markers, measurement of plasma lipids may allow clinicians to target patients with pneumonia in whom blood cultures are most likely to yield ap athogen [32][33][34][35][36][37].The variation during the hospital stay and the correlation of HDL-C with albumin and C-reactive protein suggests that HDL-C is ad ynamic surrogate marker of systemic infections [38].Keeping this in mind, physicians should be reminded that measurement of cholesterol values in patients with LRTI and systemic infections should not be used for cardiovascular risk prediction, since circulating levels of TC, LDL-C and HDL-C may be false-low and levels of TG may be false-increased respectively.
In critically ill patients, changes in cholesterol levels, and particularly ad ecrease in HDL-C, have been put forward as apredictor of severity of illness and adverse medical outcome by some investigators [1, 2, 6, 23], while other studies could not confirm this association [4,39].In our study of patients with less severe disease, we found a modest difference in TC and LDL-C, but not in HDL-C and TG, between survivors and non-survivors.Subgroup analysis revealed that these differences were only present in patients with CAP, and especially in bacteraemic patients, where they displayed high prognostic accuracy in predicting short-term mortality.Importantly,inmultivariate regression analysis these changes were independent of underlying diabetes and statin use, two conditions known to affect lipid metabolism.Accurate assessment of disease severity and prediction of outcome are prerequisites for safe decision-making on in-and outpatient management in patients with CAP.I nt he ICU setting, measurement of cholesterol values has been shown to improve risk prediction, and inclusion of lipid values in clinical risk assessment scores of critically ill patients has been advocated [6].Based on the results of this study,m easurements of TC and LDL-C may improve risk prediction in bacteraemic patients with CAP.Some limitations should be considered in the interpretation of the present results.First, at rue gold standard for the differentiation of viral and bacterial LRTI is lacking, as infiltrates on chest radiographs may be unspecific, and isolation of causative microorganisms in blood cultures is possible only in am inority of patients [40].We therefore separated patients into three different groups, namely CAP,bacteraemic CAP and non-CAP LRTI.Second, it may be argued that the diagnostic accuracy of plasma lipids should be compared to serum procalcitonin rather than CRP.However,a st his analysis takes advantage of patients included in intervention studies stratifying patients on procalcitonin values, we believe this analysis would be biased in an on-conservative way.Third, the number of outcome events in this study to assess the prognostic accuracy of lipid levels in regression models was rather low,t hus limiting the generalisabilty of the results.Fourth, we included only patients who had at least two plasma lipid measurements during the hospital stay,and hence this cohort is only representative of hospitalised patients with LRTI.Fifth, the study presented is ap ost-hoc analysis of patients included in prospective studies at asingle institution.So we interpret our results rather as hypothesis-generating than definitive.
Nevertheless, to our knowledge this is the first report investigating diagnostic and prognostic implications of plasma lipids in aw ell defined cohort of patients with LRTI, and is thus noteworthy.
In conclusion, low levels of plasma lipids, particularly low HDL-C, pointed to ab acterial infection and low TC was predictive of adverse outcomes in patients with LRTI.In addition to cardiovascular risk prediction, plasma lipid levels may be ahelpful tool in the diagnostic and prognostic workup of patients with respiratory tract infections.

Figure 1
Figure 1Mean levels of plasma lipids, albumin and C-reactive protein (CRP) on admission, after 3-5 days and on hospital discharge in patients with communityacquired pneumonia (CAP), blood culture positive CAP and other respiratory tract infections (nonCAP).error barsrepresent standard error of the mean (SeM).dashed lines represent normal values.
Baseline characteristics of the 572 patients with lower respiratory tract infections separated into patients with community-acquired pneumonia (CAP) and patients with other lower respiratory tract infection than community-acquired pneumonia (non-CAP lRTi).COPd: Chronic Obstructive Pulmonary disease, iCU: intensive Care Unit.and asthma.On admission patients with CAP had significantly lower TC (p <0.0001), LDL-C (p <0.0001) and HDL-C (p <0.0001), and ah igher TG concentration (p <0.0001) as compared to non-CAP patients.These changes were more pronounced in patients with blood culture positive CAP.C onversely,o nh ospital discharge TC, LDL-C, HDL-C and TG levels were similar in CAP and other LRTI patients.