The microbiology of IAI reflects a summary of transient or persistent normal gastrointestinal flora with potentially pathogenic micro-organisms. As such, the aetiology of IAI includes Gram-positive, Gram-negative and anaerobe bacteria as well as fungal species. However, about 20–25% of cultures in secondary peritonitis prove negative.[
11] In another quarter of the patients, the infection is monomicrobial and, in half of the patients peritonitis is polymicrobial.[
12] However, the precise mix of pathogens involved is highly variable depending on several factors that lead to the risk profile of the patient.
Table III gives an overview of the most common causative pathogens in IAI. Interpretation of the extent of micro-organisms involved is hampered by the limited microbiological workup of anaerobe cultures in general practice. In principle, IAI are polymicrobial with an aerobic component (inducing local/systemic inflammation) and an anaerobic component (responsible for abscess formation).
The value of microbiological identification is a matter of debate, as cultures often reveal mixed flora in which it is difficult to distinguish contaminants from true pathogens. In cases of community-acquired IAI with secondary peritonitis, cultures rarely influence the management of the patient as the encountered flora is generally susceptible to the standard regimens.[
15] In cases of healthcare-associated IAI, peri-operative culturing is routinely indicated, as responsible pathogens are less predictable and more likely resistant to first-line or standard empirical regimens (amoxicillin/clavulanic acid or combination of cefuroxim plus a 5-nitroimidazole). In addition, peri-operative cultures may reveal other pathogens such as enterococci,
P. aeruginosa or yeast, which are also not routinely covered by first-line antimicrobial regimens. As such, culture results allow for either correction of an initially inappropriate choice or de-escalation. However, irrespective of the isolated pathogens, coverage against Gram-positive, Gram-negative and anaerobic bacteria remains indicated. Besides adapting the empirically initiated antimicrobial therapy, culture results provide insights into local epidemiological patterns, which can be valuable in selecting the proper agents in the empirical phase.
Blood cultures are generally not recommended as they usually do not provide additional valuable information, especially in community-acquired cases. However, in cases with severe sepsis or septic shock, its use is mandatory.[
16] Also, blood cultures positive for anaerobic bacteria, especially
Bacteroides fragilis, in sepsis of unclear origin, are a strong indication of an intra-abdominal focus.[
17] Systemic breakthrough as evidenced by positive blood culture represents fulminant infection, with a particularly grim prognosis.[
18]
Healthcare-Associated vs Community-Acquired IAI
Healthcare-associated infection includes all circumstances in which patients have had close association with either acute or chronic care settings.[
20] The relevance of the issue is that healthcare-associated IAI is inherently associated with an increased risk of resistant pathogen involvement. Based on a 10-year single-centre study of over 2000 cases of complicated IAI, Swenson et al.[
21] found that resistant pathogens were isolated from 79% of healthcare-associated infections. The risk of involvement of resistant pathogens is, however, also the result of exposure to antibacterials, irrespective of having close association with healthcare facilities as mentioned in the definitions.[
20] However, it should be mentioned that, in IAI occurring within the first days of hospitalization, the prevalence of resistant pathogens is not dramatically increased.[
21,
22] In Swenson et al.,[
21] the average time between hospital admission and treatment of resistant pathogens was 10 days (13 days for non-fermenting Gram-negatives, 8 days for resistant staphylococci, 23 days for vancomycin-resistant enterococci [VRE] and 7 days for fungi). Also important, the average time from admission to treatment of non-resistant pathogens was still 5 days. Assuming that in these cases resistant pathogens were also cultured after about 5–7 days, these data indicate a timeframe within which resistant pathogens do not necessarily need to be covered in the empirical regimen. Of course, local epidemiological patterns should always be taken into account (e.g. quinolone resistance in
Escherichia coli in the community). In nosocomial pneumonia, the risk of resistant pathogen involvement is also substantial after a hospitalization of about 5–7 days, leading to the concept of early-onset and late-onset infection.[
23] In IAI, however, such a concept, although probably of value in the context of empirical choices within antimicrobials, is not proposed in the current guidelines, in which no difference is made in risk assessment of healthcare-associated IAI according to length of hospitalization.[
24] In the proposed grid (
table I), a distinction is made between community-acquired or early-onset healthcare-associated IAI (<5–7 days) and late-onset healthcare-associated IAI. Yet, recent exposure to antimicrobial agents must be considered as a risk factor for resistance, irrespective of origin of onset. Therefore, when it comes to selecting an empirical antimicrobial regimen, cases with community-onset IAI with recent antimicrobial exposure (≥2 days of therapy) are categorized together with late-onset healthcare-associated IAI.
Local Epidemiology of Resistance
Empirical choices of antimicrobial agents are based upon the risk assessment of involvement of resistant pathogens. In IAI literature, the concept of resistance goes beyond in vitro multidrug resistance and basically includes all pathogens not essentially covered by first-line empirical regimens. As such, the following micro-organisms are considered ‘resistant’: methicillin-resistant Staphylococcus aureus (MRSA), coagulase-negative staphylococci, VRE, ESBL-producing Enterobacteriaceae, quinolone-resistant E. coli, non-fermenting Gram-negative bacteria (including P. aeruginosa, Stenotrophomonas maltophilia, Acinetobacter baumannii) and yeasts.
Although the ongoing emergence of resistance is a global concern, it is local epidemiology data that should be taken into account as they can substantially differ from the national average.[
25] For example, in a general hospital in Houston (TX, USA), the prevalence of fluoroquinolone-resistant
E. coli was 2- to 4-fold higher than the national averages.[
25] The increase of quinolone-resistant
E. coli in community-acquired infections was mostly a matter of urinary isolates in male outpatients or in sputum samples from inpatients. Therefore, the extent to which this problem weighs on the specific aetiology of IAIs is uncertain. Large multicentre surveillance studies which focus on resistance are scarce in IAI and hence the true prevalence of resistance is hard to estimate. Regardless, it is unjustified to take national averages of ‘clinical specimens’ as a reference for aetiology in IAIs. Data from the European Antimicrobial Resistance Surveillance System indicate that, in France, about 22% of all clinical isolates of
S. aureus tested (n = 4720) were resistant to methicillin.[
26] In contrast, in a prospective study performed in 25 French centres, Montravers et al.[
27] found not a single
S. aureus isolate from nosocomial IAI to be methicillin resistant. So, although obviously present in other types of infection, MRSA is not a common pathogen in secondary peritonitis. An identical observation can be made for VRE. Although enterococci are important pathogens in IAI, as the overall prevalence of VRE is low, the role of VRE is limited to exceptional cases with late disease and to risk groups such as complicated liver transplantation.[
22,
27‐
29] As such, in the majority of IAI, including community-acquired and early secondary peritonitis, enterococci and, in particular circumstances, VRE, do not need to be covered in the empirical regimen.
The need to cover ESBL
Enterobacteriaceae in community-acquired IAI is also strongly dependent on geographic ecology. There are clear signals of epidemiological indications of expansion of ESBL
E. coli and
Klebsiella species in the community. This may be provoked by excessive use of antimicrobials in medical practice but also by the widespread contamination of the food chain with ESBL
Enterobacteriaceae, which is also the case in countries with – until now – no extreme resistance problems, such as Denmark.[
30‐
32] Data from India indicate that more than half of all
E. coli strains and about one-third of
Klebsiella strains isolated from IAI produce ESBL.[
33] Data from Hawser et al.[
33] show important differences in resistance rates for
E. coli and
Klebsiella species among European countries, albeit that the study is unclear about the origin of infection (either community or hospital onset). Resistance rates for
E. coli were highest in Turkey, Greece and Portugal, while fewer resistance problems were noted in France, Lithuania and Estonia. For
Klebsiella, carbapemens appear to be a good choice for all European countries (more than 90% susceptibility), with the exception of Greece, where only 43% of isolates were susceptible for both imipenem and ertapenem. In particular, because of the huge geographic variation in resistance rates, special attention should be given to patients with a recent history of travelling in regions known to have particular resistance problems, such as India, Thailand or Egypt.[
34,
35] However, the extent to which antibacterial resistance in
Enterobacteriaceae results in worse clinical outcomes in cases of community-acquired IAI, especially in circumstances where adequate surgical source control was promptly achieved, remains unknown.
Finally,
P. aeruginosa is also a feared pathogen, the importance of which is frequently overestimated in the context of IAI.
P. aeruginosa is a typically opportunistic pathogen, generally affecting immunocompromized or otherwise severely debilitated patients. Given the relatively poor ability of this micro-organism to invade tissue, the isolation of
P. aeruginosa often represents contamination rather than true pathogen, especially in cases in which surgical intervention was promptly performed and infection, as such, was not established. In the large study by Swenson et al.,[
21] non-fermenting Gram-negative bacteria (
P. aeruginosa, S. maltophilia and
A. baumannii) accounted for fewer than 6% of all IAIs. In a series of secondary peritonitis,
P. aeruginosa accounted for about 9% of infections. Edelsberg et al.[
36] demonstrated that, in both community-acquired and healthcare-associated IAI, coverage of
Pseudomonas did not add to higher rates of clinical success, indicating the limited importance of these pathogens. Yet, in patients with a particular risk profile for opportunistic pathogens (e.g. immunocompromized patients), coverage of
Pseudomonas is warranted in all cases.[
37]
Coverage Against Enterococci
Enterococcal involvement in IAI is generally encountered in patients with a loaded medical history and is associated with a grim prognosis.[
38‐
40] Routine coverage against enterococci is not mandatory in community-acquired IAI as it appears not to give additional benefit.[
41‐
43] Empirical coverage of enterococci is warranted in nosocomial IAI, which, in the majority, represents cases in whom early surgical source control was not achieved or complications such as anastomotic leakage occur. Reasonable indications for enterococcal coverage include septic shock, prolonged treatment with cephalosporins, immunosuppression, presence of prosthetic heart valves and recurrent IAI with associated severe sepsis.[
44‐
47] When enterococcal infections are suspected, coverage should be included in the empirical regimen. Empirical therapy should primarily target
Enterococcus faecalis.[
24] This may require additional drugs such as ampicillin or vancomycin when cephalosporin-based regimens are prescribed. In patients with a very high risk profile (e.g. a liver transplant patient with an IAI originating from the hepatobiliary tree or a patient known to be colonized with VRE) coverage of
Enterococcus faecium is recommended. In this case, tigecycline or linezolid can be prescribed.[
48,
49]
Coverage Against Candida Species
With the exception of patients with a particular risk profile for fungal infections (e.g. neutropenic patients) or patients with anastomotic leakage,
Candida should not be covered by the empirical regimen.[
50,
51] Even when initial culture results show
Candida spp., the clinical relevance of these isolations remains vague and probably not all patients need antifungal coverage.[
52] In a prospective cohort of 62 patients with peritonitis following peptic ulcer perforations,
Candida was isolated from 37% of cases (n = 23).[
53] Antifungal therapy was initiated in eight patients, of whom only three survived. Yet, all patients in whom no antifungal therapy was initiated survived. This illustrates the ambiguous role of
Candida as a pathogen in IAI.
An algorithm to determine which patients may benefit from antifungal therapy has been proposed.[
54] This algorithm is based on (i) presence or absence of established peritonitis; (ii) isolation of
Candida spp.; (iii) presence or absence of severe sepsis or septic shock; and (iv) immunocompromized status. This algorithm, however, has not yet been subject to clinical evaluation. Invasive candidiasis is nearly always preceded by overt
Candida colonization, and although the positive predictive value of prior colonization is rather low, failure to cover
Candida species in the empirical phase is associated with treatment failure and death.[
55,
56] Therefore, in cases of overt
Candida colonization prior to infection onset, or when peri-operative cultures reveal
Candida species, antifungal therapy is recommended. Dupont et al.[
57] developed and validated a simple score to estimate the likelihood of
Candida involvement in peritonitis. This score describes risk levels of
Candida involvement based on the presence or absence of four risk factors: female gender, upper gastrointestinal tract origin of peritonitis, peri-operative cardiovascular failure and previous antimicrobial therapy. The risk of
Candida involvement is minimal when no or only one risk factor is present and maximal when all four are present. In the presence of three risk factors, the score had 84% sensitivity, but the 50% specificity was rather low, thereby limiting its usefulness in clinical practice.
Previous antibacterial therapy may also result in selection of fungi. With regard to
Candida spp., prior exposure to fluconazole may select for non-
albicans Candida spp. with potentially reduced susceptibility to this agent.[
58] In these cases, an echinocandin is recommended.[
59]