Community-acquired pneumonia: identification and evaluation of nonresponders

Abstract

Community acquired pneumonia (CAP) is a relevant public health problem, constituting an important cause of morbidity and mortality. It accounts for a significant number of adult hospital admissions and a large number of those patients ultimately die, especially the population who needed mechanical ventilation or vasopressor support.

Thus, early identification of CAP patients and its rapid and appropriate treatment are important features with impact on hospital resource consumption and overall mortality. Although CAP diagnosis may sometimes be straightforward, the diagnostic criteria commonly used are highly sensitive but largely unspecific. Biomarkers and microbiological documentation may be useful but have important limitations.

Evaluation of clinical response is also critical especially to identify patients who fail to respond to initial treatment since these patients have a high risk of in-hospital death. However, the criteria of definition of non-response in CAP are largely empirical and frequently markedly diverse between different studies. In this review, we aim to identify criteria defining nonresponse in CAP and the pitfalls associated with this diagnosis. We also aim to overview the main causes of treatment failure especially in severe CAP and the possible strategies to identify and reassess non-responders trying to change the dismal prognosis associated with this condition.

Introduction

Community-acquired pneumonia (CAP) is a relevant public health problem, constituting an important cause of morbidity and mortality [Garibaldi, 1985; Pimentel and McPherson, 2003; Arnold et al. 2007]. In Portugal, between 1998 and 2000, CAP accounted for about 3% of the total adult hospital admissions [Froes, 2003] and 71.6% of those patients were over 65 years. The overall hospital mortality was 17.3%, increasing with age to 21.5% and 24.8% in those over 65 and 75 years respectively [Froes, 2003]. Overall the population who needed mechanical ventilation or vasopressor support usually had a high mortality rate, which reached almost 50% [Froes, 2003; Rello, 2008].

Thus, early identification of patients with CAP and its rapid and appropriate treatment are important features in the approach, and will potentially have an impact on hospital resource consumption and mortality [Houck et al. 2004]. Although CAP diagnosis may sometimes be straightforward, the diagnostic criteria commonly used, namely physical examination and chest X-ray, are highly sensitive but largely unspecific. Furthermore, biomarkers per se are not specific for CAP [Póvoa, 2008], microbiological documentation may not be feasible in 30–50% of patients [Lim et al. 2009], particularly for viral infections, and its results are usually unavailable before at least 48 h of clinical presentation.

Evaluation of clinical response is also critical, especially to identify patients whose condition fails to respond to initial treatment since these patients have a high risk of in-hospital death. However, the criteria of definition of nonresponse in CAP are largely empirical and frequently markedly diverse between different studies. According to the Infectious Diseases Society of America (IDSA) guidelines [Mandell et al. 2007], treatment failure in hospitalized patients with CAP should only be considered after at least 72 h of initial treatment, which is the standard time required to achieve clinical stability and to reduce bacterial load [Halm and Teirstein, 2002]. This may be problematic because the evaluation of response relies mostly on normalization of the same unspecific clinical and radiological criteria, previously used for CAP diagnosis. Moreover, patient recovery may take longer than bacterial killing itself and the use of cultures to prove bacteria eradication may be misleading and is not recommended [Lim et al. 2009]. Therefore, nonresponse may easily be mistakenly stated.

Biomarkers may also help to identify patients with CAP with a complicated course. Recent data have linked the pattern of C-reactive protein (CRP) kinetics with treatment failure, namely persistence of elevated concentration after 3–4 days of antibiotic therapy [Coelho et al. 2007].

Nonresponse to antibiotics may simply unveil a wrong diagnosis. However, in the presence of true CAP, different factors related to the host, the bacteria and the antibiotic therapy itself may lead to treatment failure. Among them, antibiotic prescription delay or inadequacy, and failure to provide early and adequate resuscitation of severely ill patients are probably the most important modifiable risk factors [Vincent and Marshall, 2008].

In this review, we aim to identify criteria defining nonresponse in CAP and the pitfalls associated with this diagnosis. We also aim to overview the main causes of treatment failure, especially in severe CAP, and the possible strategies to identify and reassess nonresponders trying to change the dismal prognosis associated with this condition.

Assessing nonresponders

Prevalence

Several criteria, mostly subjective, have been used to classify nonresponse in CAP (Table 1).

Table 1.

Nonresponse according to the prespecified definitions.

	N	Definition		Setting	Risk factors or causes of failure	Prevalence	Mortality
		Early failure	Late failure
Menendez et al. [2008]	453	MV, shock, or death <72 h	Persistence of fever, radiographic progression, impairment of respiratory failure, MV or shock >72 h	In hospital	Elevated IL-6, PCT or CRP at day 3; PSI class = 5	Early failure 8%; late failure 10%	Early failure 24%; late failure 43.7%; no failure 0.5%
Roson et al. [2004]	1383	Clinical and/or radiologic deterioration at 48–72 h, changes in antibiotic and/or invasive procedures, including MV and chest drainage		Ward	Younger age; PSI score >90; legionella or Gram-negative infection; multilobar infiltrates; inadequate antibiotics	Early failure 6%	Early failure 27%; no failure 4%
Arancibia et al. [2000]	444	Persisting fever 38°C and/or clinical symptoms after <72 h	MV and/or septic shock >72 h	In hospital	Infection related (38 patients); noninfection related (9 patients); other (8 patients)	Early failure 6.8%; late failure 4.3%	Failure 42.9%; no failure NA
Genne et al. [2006]	228	Clinical deterioration and fever >3 days or need to change antibiotics, severe side effect or death >48 h		In hospital	Neoplasia; neurological disease; aspiration pneumonia	Failure 24%	Failure 25.9%; no failure NA
Cilloniz et al. [2012]	568	Hemodynamic instability, impairment of respiratory failure, radiographic progression, or new metastatic infectious foci		Outpatients evaluated in an emergency room	NA	Failure 2.3%	Failure 23.1%; no failure NA
Ott et al. [2012]	1236	Need to switch antibiotics >72 h or broadening spectrum		In hospital	CURB65 class >1; antibiotics other than moxifloxacin or β-lactam + macrolide	Failure 15.9%	Failure 17.3%; no failure 5.2%
Minogue et al. [1998]	944	Hospitalization within 30 days		Outpatients	Pneumonia related (40 patients); comorbidities (26 patients); refused initial admission (5 patients)	Failure 7.5%	Failure 4.2%; no failure 0.3%

CRP, C-reactive protein; CURB, confusion, urea, respiratory rate and blood pressure score; IL-6, interleukin 6; MV, mechanical ventilation; NA, not available; PCT, procalcitonin; PSI, pneumonia severity index.

In the absence of a consensus definition, the published nonresponse rate is related to both the selected criteria and the definition. Consequently, in distinct in-hospital populations, the frequency of nonresponsiveness ranges between 5% and 30% [Mandell et al. 2007]. Nevertheless, independently of the used criteria, the reported CAP mortality rate for nonresponders is always in the higher range, usually near 20%, and much higher than that of responders (Table 1).

Among outpatients, treatment failure is usually defined as the need for hospitalization or for a change in the antibiotic therapy whilst treatment failure among hospitalized patients is mainly related to clinical deterioration and classified as early (within 72 h) or late, although slightly different definitions are used (Table 1). Early treatment failure is usually defined as the need to change antibiotics or clinical deterioration, namely septic shock or the requirement for mechanical ventilation. Late treatment failure is characterized by persistent or recurrent fever associated with respiratory symptoms, the need for mechanical ventilation or septic shock requiring intensive care unit (ICU) admission [Corrêa et al. 2009].

Accordingly, in a study assessing patients with CAP admitted to wards, early failure was identified in 83 (6%) patients and was associated with a fourfold increase in mortality (27% versus 4%) [Roson et al. 2004], despite the fact that resistant bacteria were isolated in only 31% of those patients. Similar findings were noted in another study including 453 patients with CAP, in which early nonresponse was identified in 8.4%, and 10.2% had late failure [Menendez et al. 2008]. Both groups had a high mortality rate, 24% and 43.7%, respectively.

A broader definition, including the need to enlarge antibiotic spectrum, was applied to another CAP cohort who presented a nonresponse rate of 15.9% and an in-hospital mortality of 17.3% (compared with 5.2% of responders) [Ott et al. 2012]. The nonresponse rate was lower when moxifloxacin [odds ratio (OR) 0.43] or a combination of a macrolide and a β-lactam (OR 0.68) was selected as first-line therapy [Ott et al. 2012].

Reasons for nonresponse were systematically evaluated in 62 patients with CAP [Genne et al. 2006]. A host-related cause (including suppurative complications) was the most frequent, accounting for 61% of cases, 18% had a drug-related cause and resistant bacteria was identified in only 16%. Overall, bacteria resistance to antibiotics seems to be seldom found in patients with CAP with nonresponse [Roson et al. 2004; Genne et al. 2006].

In another cohort of 444 hospitalized patients with CAP, an 11% nonresponse rate was found, with an even higher mortality rate, 42.7% [Arancibia et al. 2000]. In 39 of these patients, nonresponse was due to an infection, either CAP progression or another hospital-acquired infection. In both studies, a small number of patients had a confounding noninfectious diagnosis, three and nine patients, respectively [Arancibia et al. 2000; Genne et al. 2006].

In the outpatient setting nonresponse was much lower, ranging from 2.3% [Cilloniz et al. 2012] to 7.5% [Fantin et al. 2001] or 8% [Minogue et al. 1998]. In all these studies, mortality rate was below 5% and appeared to be mainly related to the patient’s comorbidities and not to antibiotic selection or to pneumonia itself [Minogue et al. 1998; Fantin et al. 2001; Cilloniz et al. 2012].

The role of biomarkers

To account for the diagnostic and prognostic uncertainty in CAP, the use of biomarkers has been proposed as a complementary tool. Several studies found an association between the early concentration of some biomarkers in CAP and patient outcomes [Hogarth et al. 1997; Hausfater et al. 2007; Seligman et al. 2008].

Moreover, biomarkers, mostly procalcitonin (PCT) and CRP, have also been shown to be useful in the diagnosis and monitoring of response in patients with CAP [Póvoa, 2008] and may improve diagnostic accuracy [Bafadhel et al. 2011], including the etiology itself [Garcia Vazquez et al. 2003]. However, they should be used with caution and not as a substitute for clinical outcomes [Wunderink, 2010].

The value of biomarkers (PCT and CRP) in the identification of nonresponders was addressed in a cohort of 394 in-hospital patients [Menendez et al. 2009]. Clinical stability at 72 h was found in 220 patients and predicted an uneventful CAP course during the whole in-hospital stay [area under the receiver operator characteristic curve (AUC ROC) of 0.77]. Both CRP (4.2 mg/dl versus 7 mg/dl) and PCT (0.33 ng/ml versus 0.48 ng/ml) presented lower levels in this subgroup. Moreover, CRP concentration changes over time improved the ability of clinical evaluation to predict a good outcome (AUC ROC 0.84) but the same was not true for PCT (AUC ROC 0.77) [Menendez et al. 2009].

The available data suggest that CRP is the marker with the best performance to monitor treatment response in CAP [Póvoa, 2008]. Coelho and colleagues studied the patterns of CRP kinetics in 191 critically ill patients with CAP to assess therapeutic response [Coelho et al. 2012]. Using daily determinations of serum CRP concentration, the authors calculated the ratio between the CRP serum concentration after 5 days of therapy and its initial concentration. An association between a high CRP ratio (>0.5) and increased mortality, with a sensitivity of 0.81 and a specificity of 0.58 [Coelho et al. 2007], was found. Patients with a nonresponse pattern, defined as a CRP ratio >0.8 by day 7, had a high mortality rate of 36.4%, while in patients with a fast response pattern (CRP ratio ≤0.4 by day 5) only 4.8% died [Coelho et al. 2012]. The main finding of this study was the association between CRP ratio kinetics and clinical course, providing a useful and simple tool to identify patients who do not respond to their initial therapy. These findings were also noted to be independent of the presence of neutropenia [Póvoa et al. 2011] and of corticosteroid therapy [Salluh et al. 2011].

The PCT concentration and its relationship with clinical resolution has also been used to identify patients who may safely stop their antibiotic therapy [Christ-Crain et al. 2004; Bouadma et al. 2010]. However, these results have been disputed mostly due to the reported high exclusion rates, the high rate of protocol overruling and the long duration of antibiotic therapy usually employed in the control arms [Póvoa and Salluh 2012a]. Moreover, contrary to CRP, PCT is removed by hemodialysis [Dahaba et al. 2003].

Diagnosis

Exclusion of a confounding diagnosis

Treatment failure should be approached in an early, practical and systematized manner and its potential causes should be investigated before changes in the therapeutic regimen are made (Figure 1). A full diagnostic workup, including a detailed clinical history, physical examination, and reevaluation of laboratory and radiological exams should be routinely performed.

Figure 1.

Proposed algorithm for the evaluation of nonresponse CAP at day 3–4. Persistence of signs of infection, elevated biomarkers or clinical deterioration in CAP should prompt consideration of nonresponse. Microbiologic evaluation should be performed to identify the cause of infection and help to select appropriate antibiotic therapy and dosage (accounting for the pharmacokinetic changes). Any suppurative complication should be properly addressed to ensure prompt source control. Appropriate testing to exclude alternative diagnoses should be performed according to clinical findings. CAP, community-acquired pneumonia; CHF, congestive heart failure; CRP, C-reactive protein; PCT, procalcitonin; PE, pulmonary embolism.

Failure to respond to antibiotics may merely unveil an alternative diagnosis. Indeed, several conditions may mimic the clinical presentation of CAP (Table 2), including congestive heart failure, lung cancer, pneumonitis, atelectasis, pleural effusion, fibrosis [File, 2001], increasing the diagnostic uncertainty. Accordingly, several studies assessing patients with CAP whose condition failed to respond to antibiotic therapy found a small group of patients, approximately 1–2%, who ultimately proved to have a noninfectious condition [Arancibia et al. 2000; Genne et al. 2006]. Cough, dyspnea and chest pain are unspecific and may be found in several confounding clinical syndromes, not necessarily of infectious origin.

Table 2.

Clinical conditions that may mimic community acquired pneumonia.

Infectious diseases	Endocarditis, meningitis, arthritis, pericarditis, cholecystitis
Noninfectious	Congestive heart failure, myocardial infarction, pulmonary embolism, pulmonary infarction, lung cancer, pulmonary contusion, atelectasis, postenotic bronchial obstruction, foreign body, pulmonary effusions, pancreatitis.
Inflammatory diseases	Cryptogenic organizing pneumonia, hypersensitivity pneumonitis, eosinophilic pneumonia, sarcoidosis, Wegener’s granulomatosis, acute interstitial pneumonitis, collagenosis, alveolar hemorrhage
Drug reaction	Drug fever
Acute respiratory distress syndrome

An initial suspicion of CAP may be challenged by the clinical course. Moreover a persistently low CRP [Póvoa et al. 2005] or PCT [Balc et al. 2003] may help to exclude the diagnosis of CAP. Additional tests should be performed according to the clinical setting, including chest ultrasound or computed tomography scan, to rule out pulmonary emboli, pleural effusions, lung abscess, lung cancer and airway obstruction. The radiological pattern may also suggest an alternative noninfectious disease, such as cryptogenic organizing pneumonia.

Microbiological testing

In patients with CAP not responding to antibiotic therapy, a collection of samples for microbiological examination (if not already performed), including urinary antigen tests for Streptococcus pneumoniae and Legionella pneumophila, should be performed, particularly in patients with severe CAP or in those whose condition is deteriorating. Furthermore, microbiological culture data and antibiotic sensitivity pattern, not available at admission, may unveil the cause of treatment failure. In addition, a further history of any risk factors for infection for unusual pathogens, including viruses and fungus, may lead to the selection of appropriate testing.

The interpretation of respiratory sample cultures collected after prescription of antibiotic treatment is problematic because early colonization with resistant organisms is not uncommon [Visscher et al. 2008]. Nevertheless, negative cultures for multidrug-resistant bacteria, such as methicillin-resistant Staphylococcus aureus or Pseudomonas aeruginosa, provide strong evidence against their presence [Mandell et al. 2007].

The role of molecular testing in the diagnosis of nonresponding CAP is not yet defined, although it can reduce time and improve the rate of microbiological identification [Mancini et al. 2010]. However, this method still has limitations in providing information on resistance.

The use of invasive techniques, mainly bronchoscopy, may be useful to identify unusual resistant pathogens. In a study by Ortqvist and colleagues, bronchoscopy unveiled a microbiological etiology of nonresponsiveness in four out of the 18 patients with nonresponding CAP [Ortqvist et al. 1990]. However, the use of invasive versus noninvasive diagnostic strategies in a nonresponsive CAP population has never been properly evaluated.

The use of invasive microbiological sampling may be particularly advisable in the presence of risk factors for resistance, particularly recent hospitalization, nursing home residency [Shorr et al. 2008] and recent antibiotic therapy [El Solh et al. 2004]. Kikuchi and colleagues documented silent aspiration during sleep in 10 out of 14 older patients with CAP whose condition failed to respond to initial antibiotic therapy compared with 10% of the age-matched control subjects (p < 0.02). In that study, an invasive diagnostic strategy was able to identify a large number of resistant microorganisms (methicillin-resistant S. aureus in 33%, P. aeruginosa in 14% and other Gram-negative bacteria in 23%) [Kikuchi et al. 1994].

Thoracocentesis should also be performed for evaluation and drainage of parapneumonic effusions, especially to rule out empyema. Transbronchial and open lung biopsies can also be useful in the appropriate setting.

Due to the high mortality rate of this population, and while test results are pending, temporary empirical broadening of the antibiotic regimen and transfer to a higher level of care is probably appropriate [Mandell et al. 2007].

Treatment failure in community-acquired pneumonia

In the presence of true CAP, several factors may contribute to treatment failure, which may be related to the bacteria, the antibiotic or the host, and their interactions.

Bacteria factors

Resistance to antibiotics seldom represents a cause of antibiotic failure in CAP, except in patients at high risk of multidrug-resistant microorganisms [El-Solh et al. 2002; Shorr et al. 2008]. In fact, in a cohort of 444 patients, bacteria resistant to antibiotics were found in only six of the 23 patients with persistent CAP [Arancibia et al. 2000].

Bodi and colleagues evaluated the impact of adherence to 2000 IDSA guidelines for CAP on the outcome of a cohort of 529 ICU patients with severe CAP [Bodi et al. 2005]. The proportion of patients who were not treated according to the guidelines (42.2%) had a significantly higher ICU mortality rate (33.2% versus 24.2%, OR 1.7). In the same cohort the isolation of P. aeruginosa was the main risk factor for initial antibiotic inadequacy (15 out of 20 cases). P. aeruginosa pneumonia was associated with concomitant chronic obstructive pulmonary disease (OR 17.9), malignancy (OR 11.0), previous antibiotic exposure (OR 6.2) and radiographical findings demonstrating rapid spread of disease (OR 3.9).

Polymicrobial and viral CAP may also contribute to nonresponse. In a study of 196 patients with microbiologically documented CAP admitted to an ICU, polymicrobial pneumonia was found in 39 (19.8%), of whom 15 presented both viral and bacterial etiology [Cilloniz et al. 2011]. This was associated with more severe disease and proved to be a risk factor for antibiotic inadequacy and response failure, which in turn independently predicted hospital mortality (39% versus 10%, p < 0.001). Similar findings were seen in another cohort of patients with CAP admitted to the wards. The 35 patients presenting with both bacteria and viral pneumonia had a more severe CAP than those with monomicrobial bacterial infection (OR 4.98) [Johansson et al. 2011], resulting in a significantly longer hospital stay. Also community-acquired methicillin-resistant S. aureus, although an infrequent cause of CAP, has been associated with severe necrotizing pneumonia and a high mortality rate [Micek et al. 2005; Webster et al. 2007; Nazareth et al. 2011]. In addition, different S. pneumoniae serotypes may express diverse virulence factors, including pneumolysin, causing a severe sepsis syndrome with slow resolution, which may mimic nonresponse [Dockrell et al. 2012].

Bacterial inoculum may also influence CAP outcomes. This was documented in a cohort of 93 patients with CAP caused by S. pneumoniae (36.5% of whom had positive blood cultures), using real-time polymerase chain reaction (rt-PCR) performed in whole blood samples [Rello et al. 2009]. A positive rt-PCR assay was associated with higher mortality (OR 7.08), shock (OR 6.29) and the need for mechanical ventilation (OR 7.96). Within this subgroup, 29% of the patients with the highest bacterial load, more than 10³ copies/ml, had a statistically significant higher risk for septic shock (OR 8.00), need for mechanical ventilation (OR 10.50) and hospital mortality (OR 5.43).

Similar results were reported in a prospective observational study of patients with severe community-acquired sepsis, 60.6% of whom had CAP. The 162 patients with positive blood cultures at admission (39.6% with CAP) had higher ICU mortality (42.4% versus 27.2%, p = 0.016) [Gonçalves-Pereira et al. 2012].

The worst outcome associated with large bacteria inoculum may contribute to the higher mortality observed with the late initiation of antibiotic therapy [Houck et al. 2004], which allows time for increased bacterial proliferation.

Antibiotic factors

Nonadherence with the prescribed antibiotic therapy, especially in an outpatient setting, is a cause of nonresponse and should always be excluded before assuming treatment failure.

Moreover, antibiotic efficacy is not easily assessed. The median time to achieve clinical stability is around 3 days for all patients, but some, usually those with severe CAP, may take over 6 days to stabilize [Halm et al. 1998]. Radiological improvement takes even longer and initial increase in infiltrates is not uncommon [American Thoracic Society, 1996]. Conversely, after clinical stability has been reached, subsequent deterioration and transfer to the ICU is uncommon.

Therefore diagnostic testing should be tempered before at least 72 h of antibiotics. Therapeutic changes within this period should only be considered for patients with marked deterioration or whenever new culture or epidemiologic data suggest alternative microbiological etiologies.

Even in patients receiving appropriate therapy, active against the causative microorganism, clinical resolution may not be achieved because of inadequate antibiotic concentrations. In fact, in severely infected patients, an increase in the volume of distribution [Gonçalves-Pereira and Póvoa, 2011] of some antibiotics (especially β-lactams) is always to be expected, in particular in the presence of organ failure [Ulldemolins et al. 2011], due both to host response to sepsis and to therapeutic procedures. Furthermore, creatinine clearance, which generally correlates with clearance of several antibiotics [Lipman et al. 2003], may also be increased, but commonly used formulas to compute it are useless [Baptista et al. 2011]. Consequently, PK parameters measured in the healthy population may not correctly predict antibiotic concentrations in critically ill patients with sepsis, particularly at the early stages of severe pneumonia [Hansen et al. 2001; Roberts and Lipman, 2006] and higher dosages may be needed [Pea and Viale, 2009]. Moreover, in critically ill patients with sepsis it has been suggested that larger antibiotic concentrations and time of exposure are necessary to achieve infection control [McKinnon et al. 2008].

Conventional dosage may be misleading in patients with severe infection. In fact, in one study, ceftazidime was found to be in the therapeutic range in less than half of the patients (36.9% had low concentration and 27.2% had toxic levels) [Aubert et al. 2010]. The same was noted in a larger population, mostly receiving β-lactam antibiotics by continuous infusion, in whom around half of the patients had a steady-state concentration below the intended target of four times the minimum inhibitory concentration [Roberts et al. 2010]. Furthermore, even patients with renal failure receiving modern renal replacement therapy may experience lower than expected antibiotic concentrations [Fish et al. 2005]. Also the epithelial lining fluid concentration of several antibiotics, mainly β-lactams, in infected patients [Boselli et al. 2003, 2004a, 2004b, 2005a, 2005b, 2006, 2008] was lower than expected.

Systematic therapeutic drug monitoring has been proposed to address the PK of critically ill patients with sepsis to avoid both underdosage and accumulation of the antibiotics [Sime et al. 2012]. However, target concentration guidance criteria are poorly defined and a clear-cut relationship between antibiotic serum or lung concentrations and outcomes has not been well demonstrated.

Another controversial issue is the usage of combination antibiotic therapy. In a large multicenter CAP study the prevalence of atypical microorganisms has been found to range from 20% to 28% worldwide. In this study the use of antibiotics with atypical coverage was associated with better outcomes, including lower mortality (7% versus 11.1%) [Arnold et al. 2007].

In patients with CAP admitted to an ICU, combination antibiotic therapy was associated with increased survival, but only in patients with septic shock (hazard ratio 1.69, p = 0.01) [Rodriguez et al. 2007]. Again the benefit of combination therapy in CAP may be attributed to the concomitant use of macrolides [Giamarellos-Bourboulis et al. 2008; Martin-Loeches et al. 2009], although this remains controversial [Eliakim-Raz et al. 2012]. Recent data have also unveiled a survival benefit of patients receiving dual therapy with a macrolide and a β-lactam, but again only in patients with severe CAP [Rodrigo et al. 2012].

Overall we think that atypical microorganisms should be strongly considered in patients experiencing nonresponse who are not receiving antibiotics with atypical coverage [Mandell et al. 2007]. At that timepoint, the addition of a macrolide to the therapy may be useful, especially if the patient is in shock, although data supporting this approach are sparse.

Host factors

Many host-related factors have been identified as possible causes of inadequate clinical response in severe CAP. Old age, comorbid conditions, alcohol consumption and smoking have all been associated with slow CAP resolution [Menendez et al. 2004] and may be misinterpreted as nonresponse. Initial severe clinical presentation, multilobar pneumonia, empyema and bacteremia, but not chronic obstructive pulmonary disease, were also risk factors for slow CAP resolution [Menendez et al. 2004; Roson et al. 2004], even in immunocompetent patients. Splenectomy [Tajiri et al. 2007], malignancy [Genne et al. 2006], neurological disease [Genne et al. 2006; Cilloniz et al. 2012] and immunosuppression [Japanese Respiratory Society, 2009] have all been suggested to increase CAP severity and treatment failure rate.

In addition, suppurative complications, such as empyema or abscess [File, 2001, Genne et al. 2006], may lead to treatment failure, probably related to low antibiotic concentration in the infection site as well as to their slow resolution. Therefore, thoracic drainage or a surgical approach may be needed to achieve clinical cure.

Genetic polymorphisms may also be implicated in different patterns of host susceptibility and response to infection. These polymorphisms concern several steps of immune response to infection, such as antigen recognition, proinflammatory and anti-inflammatory responses and effector mechanisms. Both basally elevated interleukin 6 (IL-6) and tumor necrosis factor (TNF) were associated with a high risk of developing CAP, at least in older patients [Yende et al. 2005]. In addition, mutation of the toll-like receptor 4 gene is associated with a greater propensity for severe infections. The same is true for some TNF polymorphisms, associated with higher mortality [Menendez and Torres, 2007]. Conversely, patients with the GG genotype of IL-6, which is associated with lower production of this cytokine, have better outcomes in sepsis [Menendez and Torres, 2007]. In fact, an exaggerated inflammatory response seems to contribute to slow or even nonresponse and a poor outcome [Ioanas et al. 2004].

Although antibiotic therapy is of utmost importance, particularly in severe CAP, adequate antibiotic therapy alone, even when optimally used, is probably not sufficient to reduce mortality [Lode, 2009]. Therefore, interest in nonantibiotic, adjunctive therapy has continued to grow [Cohen, 2009].

The single most studied pharmacological intervention is the use of corticosteroids. Several effects in the immune system have been described that could theoretically account for an improved host response in CAP [Prigent et al. 2004]. However, a recent review on the topic [Póvoa and Salluh, 2012b] found that, despite the fact that corticosteroids were largely used to treat patients with CAP, there was no proven benefit of these drugs. Moreover, in most of the studies evaluated, adverse events such as superinfections and hyperglycemia were more frequent in patients receiving corticosteroids [Póvoa and Salluh, 2012b].

Even the use of nonsteroid anti-inflammatory drugs in CAP may be associated with a more invasive disease and more pleuropulmonary complications, such as empyema and lung cavitation even in young, healthy patients [Voiriot et al. 2011].

Yet no single drug has been proved to be beneficial in patients with CAP and use of a single drug in patients with nonresponse cannot be recommended at present.

Conclusion

Nonresponse in CAP is associated with increased morbidity and mortality. Bacteria inoculum and virulence, resistance to antibiotics as well as host immune dysfunction, PK changes and suppurative complications may all be responsible for treatment failure.

Early identification of nonresponse and timely use of diagnostic and therapeutic procedures can help to improve the outcomes of these patients.

A clear definition of nonresponse in CAP and a systematic clinical approach to treating patients whose condition fails to respond to initial antibiotic therapy are needed.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of interest statement

JG-P has received honoraria and served as an advisor for Pfizer, Astra-Zeneca, Gilead, Abbott, Wyeth-Lederle, Janssen-Cilag, Merck Sharp & Dohme, and received an unrestricted research grant from Astra-Zeneca. PP has received honoraria and served as an advisor for Astra Zeneca, Ely-Lilly, Gilead, Janssen-Cilag, Merck Sharp & Dohme, Novartis and Pfizer and received unrestricted research grants from Brahms and Virogates. CC has no competing interests to declare.