Antibiotic treatment of severe bacterial pneumonia: a scoping review to characterise the evidence for current treatment options

Protocol

The methods were described in a protocol written before the work commenced (this is available in the data repository for this paper¹⁸).

Searches

An experienced health information specialist (LW) developed a comprehensive search to identify relevant studies. The following databases were searched: MEDLINE (to 12/07/2021), Embase (to 14/07/2021), Cochrane Library (to 14/07/2021), International Health Technology Assessment Database (to 14/07/2021), and the Database of Abstracts of Reviews of Effects (to 14/07/2021). The search strategies are provided in the data repository for this paper.¹⁸ Database searches were limited to 2001 onwards, based on clinical expert opinions that had informed the protocol. All searches were limited to studies published in English.

Screening processes

Two reviewers (JP and LH) independently screened the titles and abstracts of the references identified by the searches against the criteria ( Table 1). Disagreements were resolved through discussion. Full-text articles were retrieved when titles and abstracts met the eligibility criteria or when it was unclear if they met the criteria (e.g., when titles had no abstract or the information available was insufficient to make a judgement). The full-text articles were screened using the same criteria. References identified from the reference lists of systematic reviews that were not among the references identified by our searches were added to the database and screened using the same processes.

Inclusion and exclusion criteria

The inclusion and exclusion criteria specified in the protocol were refined by pilot testing before the final criteria listed in Table 1 were applied (the refinements are indicated in the footnotes in Table 1).

Data charting process

Data on the patient population, type of pneumonia, treatments, pathogens, and outcomes from each included study were extracted into a Microsoft Access database. The data extraction fields were piloted independently by two reviewers (LH and JP) on a subset of the included studies. Where possible, we used the terminology present in each study. However, for consistency and to avoid very few entries for some categories, some adaptations were necessary following the pilot phase. For example, health-care-associated pneumonia (HCAP) was categorized as other, and populations including HCAP plus hospital-acquired pneumonia (HAP) and/or ventilator-associated pneumonia (VAP) were categorised as nosocomial. The data from the remaining studies were extracted by a single reviewer (LH or JP). When data extraction was complete, the Microsoft Access database was interrogated to produce a descriptive map of the evidence for the antibiotic treatment of pneumonia.

Data items

The data extraction fields and options are shown in Table 2.

Screening and selection of evidence

Our searches identified 5736 unique records, and 52 references were identified and added from the reference lists of systematic reviews. After screening titles and abstracts (if available), we sought to retrieve 864 full-text papers. However, there were 33 reports that we did not retrieve either because they were not freely available electronically from our institution’s library (n = 31), or because the reference could not be found (n = 2). Thus, 831 full texts were assessed for eligibility and 371 were eligible for inclusion in the map. Some included studies were reported in multiple publications; therefore, the total number of studies included in the map was 327. The full list of 327 studies and the data exported from the database for each study to inform this paper are provided in the data repository for this paper.¹⁸ The remaining 460 full texts were excluded from analysis. A summary of the screening process is shown in Figure 1, and lists of the 33 reports that could not be retrieved and 460 excluded studies are available from the data repository.¹⁸

Figure 1. Prisma flow chart.

Data charting

Study locations

Of the 327 studies that met our inclusion criteria, the geographical location of more than half was in Europe (55.0%). The remainder were conducted in the USA and/or Canada (28.4%), Australia and/or New Zealand (1.5%), a mixture of countries that met our inclusion criteria (2.1%), or a mixture of countries that included some countries that did not meet our inclusion criteria (12.8%).

Participants

All the studies included adults, with 14 (4%) focusing on a subgroup of older adults (defined as aged 65 years or older in 12 studies).

Pneumonia type and severity and setting at baseline

To report the results of this study, the pneumonia categories HAP, HAP & VAP, and nosocomial were combined under the heading nosocomial. The pneumonia categories mixed, other, and unclear were combined under the heading mixed/other (all these categories remain separate in our Microsoft Access database). Community-acquired pneumonia (CAP) was the focus of 39% of studies, nosocomial pneumonia 19%, VAP 25% and the remainder (17%) were designated as ‘Mixed/Other’ ( Table 3). In the majority of the studies (86%), all patients had pneumonia. A fifth of the studies focusing on CAP only included patients with severe pneumonia. It was more common for CAP studies to include a mix of patients with severe and non-severe pneumonia, and in some CAP studies, the severity of the pneumonia was not stated or was unclear. Studies focusing on nosocomial pneumonia or VAP rarely stated that the pneumonia was severe, but this may be because it was assumed that nosocomial pneumonia and VAP would be classified as severe pneumonia. The setting in which the study took place was often not stated or unclear (29% of studies) and, in line with our focus on severe pneumonia where the setting was stated, this was often the intensive care unit (ICU) (40% of studies).

Pathogens

The majority of the studies that included patients with nosocomial pneumonia or VAP reported the treatment of a specific species of bacteria or a specific named group of bacteria ( Table 4). Among these studies, the proportion in which the pathogen was not stated or unclear was very low (3% or less). In contrast, in the majority of CAP studies, pneumonia was caused by a variety of different bacteria, with fewer studies reporting the treatment of a specific pathogen or specific group (19%) or with the pathogen(s) not stated or unclear (17%).

Among the 144 studies that focused on a specific single or specific group of bacteria, the most commonly studied were gram-negative pathogens, with Acinetobacter baumanii and Pseudomonas aeruginosa being the most common specific single bacterial species studied (both are gram-negative pathogens) ( Table 5). Studies of patients with nosocomial pneumonia or VAP were more likely to focus on a specific single or specific group of bacteria than studies of patients with CAP or pneumonia characterised as ‘Mixed/Other’.

Study designs and sample sizes

Sixty-seven (20%) of the 327 included studies used an RCT design and a similar proportion (23%) were prospective cohort studies ( Table 6). The remainder were retrospective cohort studies (56%). Almost half of the studies had sample sizes between 100 and 999.

Treatment

The treatment provided for pneumonia was the most difficult study feature to categorise; therefore, to provide flexibility, we used three different approaches: how the treatment was specified, which antibiotics were used in the study, and whether the study focused on any of the antibiotics on the WHO reserve drug list¹⁹ (because the use of these antibiotics could be an indication that earlier treatment options had failed). Additionally, we categorised any studies that included inhaled adjuvant therapy because this will be the mode of delivery for FLAMOD.

The study treatment was specified as ‘Individual’ when a named, single antibiotic was used or two single antibiotics (or different doses of one antibiotic) were compared with each other and 154 of the 327 studies were categorised in this way ( Table 7). Most (51%) of these 154 studies were retrospective cohort studies, 35% were RCTs, and 15% were prospective cohort studies. In some studies, the treatment was a particular class of antibiotics e.g. macrolides, third generation cephalosporins or fluoroquinolones, so the treatment was specified as ‘Class’. Only 18 studies were classified as ‘Class’ (50% retrospective cohort, 39% prospective cohort, and 11% RCTs). Some studies described a ‘Mixed’ treatment that comprised one element categorised as ‘Individual’ but another categorised as ‘Class’. Twenty-two studies were categorised with the ‘Mixed’ treatment type (82% retrospective cohort, 18% prospective cohort). The final two ways in which treatments were specified were ‘Group only’ or ‘Group mixture’. In the ‘Group only’ set, the treatments were described without reference to any individual or class of antibiotic, instead these ‘Group only’ sets included groups defined by, for example, guideline concordant therapy or guideline non-concordant therapy, delayed or early therapy, aggressive or conservative therapy, appropriate or inappropriate therapy. Ninety-one studies had treatments categorised as ‘Group only’ (66% retrospective cohort, 34% prospective cohort). The 42 studies categorised as ‘Group mixture’ (48% retrospective cohort, 26% prospective cohort and 26% RCTs) typically compared a single named antibiotic or class of antibiotics with a ‘Group’, for example ertapenem versus ‘other agents’ or carbapenems versus ‘without carbapenems’.

The 236 studies with treatments that were classified as ‘Individual’, ‘Class’, ‘Mixed’ or ‘Group mixture’ were also classified by which antibiotic(s) were the focus in the study ( Table 8). With such a wide variety of studies and different antibiotics in use, it was not possible to categorise the studies by individual antibiotics (with the exception of linezolid and lefamulin); therefore, the antibiotics were classified by class. The aim was to identify subsets of the included studies according to the class of antibiotics of interest to allow further investigation of the subset and identification of data required for the economic model. Seventy-eight of 236 studies (33%) were mapped to a single drug class, the most common being polymyxins. These studies will include a mixture of those that focus on a single antibiotic in that drug class but will also include, for example, studies that compare two different antibiotics from within the same drug class. Sixteen studies (7%) were mapped to ‘Other drugs’ and the remaining 142 (60%) were mapped to two or more drug classes (including the category of ‘Other drugs’).

The final way in which the treatment was categorised in the 236 studies with treatments that were classified as ‘Individual’, ‘Class’, ‘Mixed’ or ‘Group mixture’ was by indicating if a WHO reserve list antibiotic had been used ( Table 9). At least one WHO list antibiotic was used in 44% (104/236) of these studies, with some studies (n=11) using more than one antibiotic (2 in 9 studies, 3 in 1 study, and 4 in 1 study). Therefore, in total, there were 118 instances of a specific WHO antibiotic being used in a treatment ( Table 9), with the most common being colistin (injection), which was predominantly used in nosocomial pneumonia and VAP; linezolid which was used across all the pneumonia types categorised; and tigecycline, which was predominantly used in nosocomial pneumonia and VAP.

Among the 327 included studies, 23 included inhaled adjuvant therapies.

Outcomes

Studies reporting on mortality, clinical outcomes, adverse events, or quality of life were included. Fifty-two (52/129, 40%) of the 129 studies that focused on patients with severe CAP or a mixed-severity CAP population reported at least one outcome for patients with severe CAP ( Table 10). For pneumonia types other than CAP, very few studies stated that patients had severe pneumonia; therefore, this restriction was not used when looking at which outcomes were reported by the studies. Across all pneumonia types, the most commonly reported outcome was mortality, followed by clinical cure or stability. No studies reported on health-related quality of life (HRQoL) among patients with severe CAP (three studies reported on HRQoL for mixed-severity CAP) and only one study reported on HRQoL in patients with nosocomial pneumonia (severity not stated or unclear).