Community-acquired pneumonia (CAP) is a frequent and severe infection, and is considered the primary cause of death from infection, and the sixth most common cause of overall mortality in Western countries 12. Consequently, CAP represents one of the leading causes of infectious admissions to the intensive care unit (ICU) 3. Indeed, the latest studies have reported that up to 10 % of all patients hospitalised with CAP require ICU management 4. In this specific subgroup of severely ill patients, the overall mortality rate remains unacceptably high despite improvement in critical care management 5. Furthermore, the medical burden of CAP is very high in terms of direct costs, associated morbidity and long-term disability 67.

Streptococcus pneumoniae (S. pneumoniae) is the principal causative agent of CAP requiring hospital or ICU admission 8. Paediatric and adult literature about non-severe pneumococcal pneumonia is abundant, but specific data on patients requiring ICU admission are scarce. In the two studies focusing on the epidemiology of pneumococcal pneumonia among patients admitted to ICU, co-morbidities negatively influenced patient outcomes, but were over-weighted by the severity of the clinical features 910. Recent therapeutic researchers have pointed out that early and adequate antibiotherapy is of greatest importance during sepsis, whereas adjuvant therapies like steroids or activated protein C may reduce the fatality rate 11. However, the roles of these anti-inflammatory agents, as well as the association with macrolides during severe pneumococcal pneumonia are a matter of debate 12. Thus, increased knowledge of severe S. pneumoniae pneumonia that may improve early detection and treatment of this particular subgroup carries high interest for ICU physicians.

The aim of this present study was to provide recent epidemiological data through a large cohort of adult patients admitted to ICU for severe pneumococcal CAP. In addition to analysis of microbiological features, we assessed the respective influence of co-morbidity and organ failure on mortality. We also investigated the potential impact of adjuvant therapies on outcome.

Of the 3,658 patients enrolled in the two cohorts, 423 had positive microbiological samples for S. pneumoniae. Among them, 133 had evidence of aspiration or bronchitis and 290 were subsequently screened. After the exclusion of 68 patients with concomitant meningitis, 222 patients were finally included in this study.

We report here the most important cohort of pneumococcal pneumonia requiring ICU management. The strength of our study is that we studied a homogenous population, including patients with only microbiologically documented S. pneumoniae CAP, with exclusive pulmonary infection and reviewing of the patients' files by two independent investigators. Moreover, we chose to eliminate aspiration pneumonia or chronic obstructive pulmonary disease (COPD) exacerbated by S. pneumoniae colonisation. We used a multicentre cohort, reflecting both university (tertiary referral) and non-university hospitals with a wide scope of hospital sizes, locations and practice environments.

It is striking that underlying co-morbidities have no impact on outcome, even if some of them have been demonstrated to increase the incidence of invasive pneumococcal disease 21. If tobacco and alcohol abuse were more frequent in the pneumococcal CAP than in the general population, they did not influence the course of the infection. Similarly, respiratory failure and septic shock were the main admission motives, but they were not independent prognostic variables. In contrast, male gender, age and need for renal replacement therapy were overwhelmingly independent risk factors for mortality. In a set of 43 patients with severe pneumococcal pneumonia, death was driven by organ failure rather than co-morbidity 10. In another group of 137 ICU patients, Georges et al. showed similar results, except for renal replacement therapy 9. These authors reported that leucopenia, nosocomial infections and inadequacy of initial antimicrobial therapy were other risk factors for death. In our study, none of these factors were associated with adverse outcomes. Astonishingly, male gender had a major impact on outcome. This is in line with the higher mortality rate of sepsis in male patients despite lower incidence than in female patients 22. Underlying hormonal-, social- or disease-associated mechanisms remain to be explored to explain these findings 23.

Pneumococcal pneumonia requiring ICU admission is associated with a high rate of fatality, with more than one fourth of the patients dying in hospital. The highest proportion of death occurs early in the course of the disease, despite an excellent proportion of initial adequate antibiotherapy 24. This is considered to be related to the early inflammatory process, which is overwhelming host defences. In a recent study, Garcia-Vidal et al. investigated independent factors associated with early deaths in CAP and demonstrated that age, altered mental status, multilobar pneumonia, shock, bacteraemia and inadequate empiric antibiotic therapy were predictors of death within 48 hours 25. The small number of patient dying in the first two days in our cohort did not allow us to perform similar analysis.

The present study confirmed that S. pneumoniae pneumonia-related bacteraemia is not associated with a worse outcome. Clinical features and organ failures prevent identification of bacteraemic and non-bacteraemic patients (Table 2). This observation is similar to previous reports 2627 including the study from Bordon et al. who investigated a cohort of 1,847 patients, including 125 patients with bacteraemia, and found that prognosis was not influenced by S. pneumoniae bloodstream infection 27. Our findings are in line with prior investigations, suggesting that presence of pneumococcal bacteraemia might not be a contraindication for de-escalation and short duration of antibiotherapy in clinically stable patients.

As recommended by international guidelines 28, initial antibiotherapy in the setting of severe CAP must include a β-lactamin active against S. pneumoniae. As a result, the rate of adequate antibiotherapy is usually high, as reported in our study. Consequently, improvement of antibiotherapy delivery is unlikely to improve the outcome of pneumococcal pneumonia. Thus, adjuvant therapies might have a determinant role to reduce mortality rate 92429. In our study, we failed to demonstrate that use of activated protein C was associated with a better outcome, despite the suggestion by Laterre et al. that its use was of particular interest during S. pneumoniae pneumonia in the Prowess study 30. However, the recent Prowess Shock study did not replicate these results and led to drug retrieval 31. Similarly, the Captivate study was the largest clinical trial conducted on severe CAP and examined the recombinant tissue factor pathway inhibitor. With 694 proven cases of pneumococcal pneumonia, the study failed to demonstrate any benefit on survival in either the whole population or the pneumococcal subgroup 32.

Low-dose steroids represent another controversial therapy in severe sepsis. Some studies suggest this treatment improves survival in critically ill patients with severe CAP 33. In contrast, others, such as Snijders et al., who compared the administration of corticosteroids plus antibiotics, vs. antibiotics alone in severe sepsis, reported no benefit with steroids 34. The low number of treated patients, the observational characteristics of our study and the variability of steroid use between the participating centres preclude drawing any conclusions. However, in the present study, low doses of steroids were associated with increased mortality in univariate analysis. With the findings of the Corticus study, low-dose steroids should be restricted to patients with refractory septic shock 35.

Association of β-lactamin and either a macrolide or a fluoroquinolone has been linked with a lower mortality in severe CAP 836. However, the clinical impact of such a combination is seriously debated, and may vary among pathogens. For S. pneumoniae, combined antibiotic therapy might be beneficial to treat undiagnosed co-infections, or through anti-inflammatory properties of macrolides. However, these data emerged essentially from retrospective or small prospective studies 836373839404142, and are challenged by several other reports 434445464748 as with our study, and will have to be confirmed in a multicentre randomized study. To date, the recommendation of dual antibiotherapy for S. pneumoniae CAP has not been endorsed by scientific societies.

Pneumococcal pneumonia occurred preferentially during autumn or winter. This phenomenon has already been described in non-critically ill patients, with a clear link to external temperature 49 and concomitant respiratory viral infections 50. Here we report for the first time this seasonality in ICU pneumococcal CAP. Interestingly, the severity of pneumonia assessed by the mortality rate was not influenced by season.

We found that S. pneumoniae strains with diminished susceptibility to penicillin were involved in almost 40% of cases. This is consistent with the recent report from the French national reference centre for S. pneumoniae, which collects isolates from invasive pneumococcal diseases 51: the overall proportion of strains that are non-susceptible to penicillin decreased from 45% to 35% in adults between 2001 and 2006. Interestingly, we confirmed in this severe ICU population that penicillin resistance did not influence outcome, as previously described in less severe populations 5253.

A few limitations deserve careful consideration. First, despite prospective data acquisition, there were some missing data, as illustrated by the lack of information about influenza or pneumococcal vaccination or HIV status. Moreover, we did not have access to the time lag between symptom onset and antibiotherapy initiation, which has been proposed as a major outcome determinant 54. Second, due to the multicentre design, therapeutic strategies (that is, antimicrobial therapy, ventilation protocols and adjunctive therapies) were not standardised. Third, we used the 2001 American Thoracic Society guidelines for severe CAP definition, which included three minor criteria, rather than the 2007 ATS/IDSA definition that includes nine minor criteria 28. Phua et al. investigated the prognostic value of minor criteria for ICU admission, according to the 2007 guidelines, and found that mortality rose from 0.9% with no risk factors to 35.2% for patients with at least three minor criteria 55. This underlines the large improvement of determination of prognosis provided with the 2007 criteria, as compared with the initial one that we used. Unfortunately, several of the new minor criteria were not present in our databases. Fourth, we could not study the serotypes of S. pneumoniae in our analysis, and strain specificities seem to be a major element of severity 56. This requires specific investigations, which were not routinely performed. Similarly, biomarkers such as C-reactive protein or procalcitonin were not available in the databases, whereas routine use of these biomarkers has been shown to improve the triage score for CAP patients in the emergency department 57. As all our patients were hospitalised in the ICU, the interest of such predictor tools remains to be demonstrated in this population.

In summary, we conducted a multicentre observational study of the epidemiology and outcome of pneumococcal CAP in a large homogeneous cohort of critically ill patients. We highlighted that S. pneumoniae pneumonia was still associated with a poor outcome, with high early mortality, despite adequate antimicrobial therapy. In the future, determination of factors related to the host, that is, genetic susceptibility, and to the bacteria, that is, virulence, is likely to be crucial for a better understanding of the pathophysiology of invasive pneumococcal pneumonia 58.

• Despite a high proportion of adequate antibiotherapy, mortality in patients with severe pneumococcal CAP admitted to the ICU reaches 29 %.

• Microbiological specificities (penicillin susceptibility and concomitant bacteraemia) have no significant impact on organ failure and mortality.

• Underlying co-morbidities do not impair survival in the ICU, whereas age, male sex and need for renal replacement therapy are associated with an unfavourable outcome.

• Adjunctive therapies do not seem to influence the impact of severe pneumococcal pneumonia.

ARDS: acute respiratory distress syndrome; CFU: colony forming unit; COPD: chronic obstructive pulmonary disease; HIV: human immunodeficiency virus; ICU: Intensive Care Unit; LODS: Logistic Organ Dysfunction System; MIC: minimal inhibitory concentration; NYHA: New York Heart Association; SAPS: Simplified Acute Physiology Score; S.pneumoniae: Streptococcus pneumonia.

JC, AC and JPM are consultants for Lilly. The authors declare that they have no other competing interests relevant to the field of the manuscript.

NM, AM, JPB and JPM designed the study. NM, AM and AB extracted the data. FP performed the statistical analysis. NM, AM, AB, FP, VL, JC, AC, JDC, JPB and JPM contributed to the conduct of study and data analysis. NM and JPM wrote the manuscript. All the authors read and approved the final version of the manuscript.