Systematic review: the bioavailability of orally administered antibiotics during the initial phase of a systemic infection in non-ICU patients

This study was performed and reported according to the PRISMA (Preferred Reporting items for Systematic reviews and Meta-analysis) statement (Supplemental Table 1) [15].

Studies reporting data on the pharmacokinetics of oral antibiotics in the early phase of infection were searched, preferably, but not necessarily, in comparison with the convalescence phase of infection. Studies were eligible if they included patients aged 16 years or above and febrile or acutely ill due to an infectious disease, which had to be clearly documented or illustrated with elevated infectious laboratory parameters, i.e. CRP, leucocytosis, SIRS or qSOFA criteria [16]. We chose a subset of antibiotics which are widely used and known to have a moderate to good bioavailability, namely amoxicillin, flucloxacillin, ampicillin, clindamycin, macrolides, fluoroquinolones, metronidazole and trimethoprim-sulfamethoxazole. The pharmacokinetic outcome parameters of interest were those related to oral bioavailability: the (maximum) serum concentrations (Cmax), time of achieving maximum serum concentrations (Tmax), the area-under-the-plasma-concentration-time curve (AUC) or bioavailability itself (F). Blood samples for these outcome parameters had to be taken at the first day of antibiotic therapy, when patients were in the initial phase of their infectious disease. Intravenously pre-treated patients were excluded. Studies had to be reported in English or Dutch. We allowed all clinical study types, as long as they presented sufficient information to retrieve the patient inclusion criteria and the predefined outcome parameters. We excluded studies investigating healthy volunteers; patients admitted to the Intensive Care Unit; patients with impaired renal or hepatic function, because the impairment itself can already influence the predefined outcome parameters; and febrile neutropenic patients, because mucosal injury might make its findings not be generalizable to the general population.

Together with an experienced clinical librarian, we conducted a systematic literature search in OVID MEDLINE and EMBASE for all relevant studies up to July 2020, based on the predefined objectives and eligibility criteria. In addition, we searched the reference lists of retrieved reviews. The primary records obtained were imported and de-duplicated in EndNote (complete search strategies can be found in supplemental Table 2 and 3). One reviewer (A.v.d.B.) screened all the titles and abstracts, to identify studies that potentially met the eligibility criteria. 10% was randomly assigned to and independently screened by the other reviewers (R.M.v.H, C.E.V.,J.M.P) to ensure reliability and completeness. Differences in decision were resolved by consensus. We allowed a 2.5% margin of difference between the reviewers. If after discussion the difference remained more than 2.5%, all articles had to be screened by the other reviewers. Next, the full text articles of the potentially relevant studies were retrieved and assessed for eligibility by all reviewers. Any disagreement on inclusion of studies was discussed by all reviewers and resolved by consensus. Finally, the reference lists of the eligible articles were screened by A.v.d.B for additional suitable studies.

A standard form was used to extract and summarize the predefined outcome data and data necessary for the quality/risk of bias assessment of the included studies. Study design, patient characteristics, predefined pharmacokinetic outcome parameters and conclusions were extracted by one reviewer (A.v.d.B.) and fully checked for accuracy by another reviewer (R.M.v.H). Discrepancies were resolved by discussion, together with the other reviewers if necessary. Next, the risk of bias of the included studies was assessed independently by three reviewers (A.v.d.B., R.M.v.H., J.M.P.), using an adjusted form of the Newcastle-Ottawa Quality Assessment Scale (NOS) for non-randomized studies, and the Cochrane risk of bias tool for Randomized Controlled Trials [17, 18]. Studies could score points (or stars) on three dimensions: Selection (max. 5 stars), Comparability (max. 2 stars) and Outcome (max. 3 stars). The more stars, the lower the risk of bias. We adapted sub questions of these domains to enable more appropriate quality evaluation for descriptive studies reporting pharmacokinetic parameters. The maximum NOS score was 10 and we considered the risk of bias high when the score was 5 or lower and low when above the median score of 5.

Our literature search yielded 6011 potentially relevant studies. After removing the duplicates 4989 papers remained. Based on the eligibility criteria, 4879 studies were excluded in the initial screening phase based on title and abstract, leaving 110 records for full text screening, including two records which were added after reference screening of two reviews addressing the pharmacokinetics of ciprofloxacin and intracellular pharmacokinetics of antibiotics [19, 20]. Of these, 103 were excluded, and 7 included after full-text screening. In addition, we identified 2 papers by reviewing the reference lists of the included studies, resulting in 9 papers for qualitative analysis [21‐29]. (Fig. 1).

The study characteristics of the 9 included studies are presented in Table 1. The studies were non-randomized, observational studies [21‐29] All studies were performed in hospitalized patients. 6/9 studies reported fever or other signs of acute phase of febrile illness [21, 25‐29]. In 3/9 studies, which were the studies that were added after reference screening of the retrieved reviews and the reference lists of the included studies, the febrile state of patients was unclear, which may explain why these records were not captured in our initial search. However, the patients were hospitalized for acute purulent respiratory exacerbations and were in their initial phase of illness, making it highly likely that these patients were febrile or acutely ill as well. Therefore, we chose to include these studies [22‐24]. Four studies originated from The Netherlands [21‐24], two from the USA [27, 29], one from Canada [26], one from Egypt [25] and one from Guatemala [28]. The pharmacokinetics (PK) of amoxicillin was investigated by two studies [22, 25], the PK of azithromycin [21], ampicillin [22], clarithromycin [26] and enoxacin [24] by one study each and the PK of ciprofloxacin by four studies [23, 27‐29].

The risk of bias results are listed in Table 2. Six out of nine studies had a low score and therefore a high risk of bias, mainly due to the sample selection, in which a sample size calculation was missing, and the outcome measurement, in which a clear description of the laboratory procedures for the measurement of drug concentrations was missing [21‐25, 28].

To the best of our knowledge, this is the first study that systematically reviewed the bioavailability of orally administered antibiotics during the initial phase of infection in non-ICU patients in accordance with the PRISMA statement. A major strength of our systematic review is that we had a very broad and thorough search strategy, and three title and abstract screening reviewers, reducing the risk that articles have been missed. Our systematic review is also subject to several limitations. Because we included studies that used very heterogeneous methods, study endpoints and outcome measurements, we were not able to pool the data and process them in a meta-analysis. Also, most of the included studies were dated from 1975 to 1995. It is uncertain whether the used laboratory methods to measure the antibiotic serum concentrations were sensitive enough to present reliable results. Most studies did not report whether their method for antibiotic concentration measurement was validated (Table 2). Finally, it is possible that relevant pharmacokinetic data have not been published. For example, studies sponsored or performed by a pharmaceutical company are less likely to be published, regardless of the results [34].

Our findings showed that knowledge of the bioavailability of orally administered antibiotics during the acute phase of a febrile illness in non-ICU patients is scarce. In previous studies the PK of antibiotics in a broader sense, so not only bioavailability, but also clearance and volume of distribution, has been mainly investigated in healthy volunteers and in critically ill patients [8‐10, 35]. Non-ICU patients cannot be automatically equated with the latter, as systemic infection might profoundly alter the PK of antibiotics depending on its severity [35]. For example, in critically ill infectious patients an increased volume of distribution is often seen, due to capillary permeability negatively impacting exposure [7‐10]. And in terms of clearance, as stated before, both toxic antibiotic serum concentrations and subtherapeutic antibiotic serum concentrations can be seen, depending on the perfusion alterations of the kidney. Especially in cases of augmented renal clearance both the area-under-the-concentration-time curve as well as the percentage of time of a dosing interval the antibiotic concentration is above the minimum inhibitory concentration will be lower [7‐10]. In addition, the results from critically ill patients do not learn anything about bioavailability of antibiotics, as antibiotics are almost always administered intravenously in these patients. The same accounts for patients newly admitted to a general ward with an acute infection. This reluctance to administer antibiotics orally from the beginning of a course proves that clinicians are not a priori convinced that the acute phase of an infection does not alter bioavailability, neither in ICU patients nor in non-ICU patients, even if the gastrointestinal tract of the patient is intact. The two studies on ciprofloxacin [27, 29] and the one study on clarithromycin [26] in which the same patients in the febrile and afebrile phase of infection were compared, suggest that bioavailability is not altered during the initial phase of infection. Although these results are promising concerning the possibility to switch from IV to oral therapy within the first 24 h of treatment, the value of these studies is limited due to their small number of included patients (respectively n = 12, 7 and 12). For the other orally administered antibiotics evidence is completely lacking whether febrile illnesses influences the extent of bioavailability [21‐25, 28]. If febrile illnesses do have an effect on the bioavailability of orally administered antibiotics, it is likely that there are clinical consequences of oral administration: reduced bioavailability could lead to insufficient serum concentrations, negatively affecting clinical outcome and the risk of development of antibiotic resistance, while increased bioavailability might increase the risk of toxicity [12, 36, 37]. We therefore believe that current and new oral antimicrobial agents should be tested in non-ICU patients during the acute phase of febrile illness as well, and not only in healthy volunteers or during the convalescent state.