This protocol was designed following the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines (4), and it is registered in the PROSPERO database (registration number CRD42020152621). The PRISMA-P checklist (5) was used to guide the presentation of the items of this protocol (additional file 1).
Review question
In this protocol, the procedures for a systematic review and meta-analysis that is intended to answer the question: "Is antimicrobial exposure among the over 65 associated with kidney, liver, or tissue injury?" have been outlined.
The components of the population, exposure, comparator, and outcome (PECO) are as follows:
- Population: Older adults aged 65 years and above
- Exposure: Any antimicrobial treatment, or long-term antimicrobial treatment, or broad-spectrum antimicrobial treatment
- Comparator: No antimicrobial treatment, or short-term antimicrobial treatment, or narrow-spectrum antimicrobial treatment.
- Outcome: kidney, liver, or tissue injury during or soon after antimicrobial exposure.
Eligible study designs
This study will include all primary epidemiological observational studies, including cross-sectional studies, prospective cohort studies, retrospective cohort studies, case-control studies, and interventional studies that meet the inclusion criteria. However, very few randomised controlled trials are expected to be included in this study. Case reports, case series, reviews, commentaries, editorials, will not be included in this study.
Eligible participants
A study is eligible for this review if it reported exposure to antimicrobial agents among older adults of 65 years or above. Observational and interventional studies are eligible for inclusion in this study. However, studies reporting antimicrobial exposure to children or adults under 65 years of age will be excluded. Studies reporting exposure of any age group to other medicines but not antimicrobials will also be excluded.
Exposures of interest
The exposure of interest is the therapeutic use of antimicrobials of any dose, class/type (narrow or broad-spectrum), duration, or route given for any indication. The comparator is either a group that did not receive any antimicrobials or a group that received an antimicrobial or group of antimicrobials other than the exposure of interest.
Outcome measures of interest
The main outcome of this study is to determine whether antimicrobial exposure increases organ (liver, kidney, or tissue) injury following any antimicrobial exposure among older adults who are 65 years and above. The main measure of effect will be the relative risk (6) of organ injury among the exposed when compared to the unexposed populations. The incidence rate of organ (kidney, liver, or tissue) injury among the antimicrobial exposed group of the older adults will be compared with the respective incidence rate of the control groups (not exposed to antimicrobials) and expressed as risk ratios.
Additional outcome(s)
Additional outcomes will include the determination of whether broad-spectrum antimicrobials increase the risk of organ injury when compared to the narrow spectrum, and whether prolonged exposure increases the risk when compared to short term exposure. The main measure of effect on both outcomes is the risk difference (7) between comparison groups. The risk of organ injury among participants who received broad-spectrum antimicrobials will be compared with those who received narrow-spectrum antibiotics. Secondly, the risk of organ injury among participants with prolonged exposure to antimicrobials will be compared with those on the short term. Risk difference will be the main measure of effect on both cases.
Search methods for the identification of studies
Searches will be conducted in Psych INFO, PubMed and EMBASE databases. Restrictions to publication dates will be applied and limited to the English language. The full search strategy is described in appendix 1.
Selection of eligible studies
Search results from all databases will be uploaded into EndNoteX9 (8) bibliographic software and manually de-duplicated. After removing duplicates, studies will be uploaded into Covidence (9) for the title and abstract screening, as well as full-text screening. Studies identified by the search strategy will be screened independently by two reviewers (TC and PN) using the titles and/or abstracts for their potential to meet the inclusion criteria described above. The full texts of potentially relevant studies will be retrieved and individually reviewed for eligibility using a standardised assessment tool. A decision process flow chart (figure 1) has been developed to standardise full-text selection by both reviewers. Where the two reviewers disagree, the opinion of the third reviewer (SR) will be sought. Further study information will be sought from the study authors about any missing data or clarification about study participants and study outcomes.
Data extraction process
The data to be extracted include study name/article title, authors, journal and full reference, country, study design and setting, participant demographics, exposures (types of antimicrobials including dose, duration and route), a sample size of exposed and comparator groups, confounders, statistical methods, primary outcome, secondary outcomes, and participant missing data and reason for missing. Extracted participant demographics for exposure and control groups will include mean or median age at antimicrobial administration, sex, and ethnicity. For the primary or secondary study endpoints, wherever applicable, we will extract mean (plus standard deviation), median (plus interquartile range), number of cases, and relative risks (with 95% confidence intervals).
Risk of bias assessment for eligible studies
Each reviewer will critically appraise each study and perform the risk of bias using the appropriate tool. Risk of bias for randomised controlled trials (RCTs) will be assessed using version 2 of the Cochrane risk-of-bias tool for randomised trials (RoB 2) (10). In contrast, the Newcastle-Ottawa tool (11) will be used to assess the risk of bias for observational studies. The quality assessment across the studies will be done using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) tool (12) for each outcome. Where the two reviewers differ in opinion, the third reviewer's opinion will be sought. The authors will be contacted for additional information in case of missing or unequivocal information. The following domains will be assessed for the risk of bias:
Type of bias
|
Type studies
|
Low/High risk
|
Selection bias
|
Controlled studies
|
- Low risk if random sequence generation
allocation concealment used
|
observational studies
|
- Low risk if patients enrolled as consecutively observed based on the pre-existing protocol
- Low risk if numbers and reasons for exclusions were reported
- High risk when the association between exposure and outcome is different for study participants compared to non-participants
|
Performance bias
|
Controlled studies
|
- High risk if study personnel were not blinded as to which intervention the elderly patient has received
|
Observational studies
|
- High risk if there are systematic differences in the treatment of participants
|
Detection bias
|
Controlled studies
|
- High risk if personnel evaluating outcomes were not blinded
|
Observational studies
|
- High risk if there were systematic differences in outcomes assessment among comparison groups
- High risk if the measurement of exposure is flawed e.g. recall bias in case-control studies
|
Reporting bias
|
Controlled studies
|
- High risk if reporting of outcomes is not prespecified as of interest to the review
|
Observational studies
|
- High risk if there are systematic differences between reported and unreported findings
|
Confounding
|
Controlled studies
|
- Low risk if allocation was balanced between groups by e.g. matching, stratification etc
|
|
Observational studies
|
· High risk, but can be mitigated using matching by propensity scores etc
· High risk if there is a failure to adjust for important confounders in the statistical analysis.
|
Confounders relevant to all or most of the studies
Important confounding factors in this study are those that affect the association of antimicrobial exposure and organ injury among older adults. The severity of the disease being treated, comorbidities, and non-antimicrobial medication known to risk organ injury will be considered as significant confounders. Sex and age will be also be considered apriori.
Strategy for data synthesis
Descriptive analysis of data
All studies that meet the inclusion criteria will be described, including the following:
- Study design – including the study quality, data collection methods, the validity of tools used and the statistical analysis
- Participants (both exposed and unexposed) – including demographic and socio-economic characteristics and health status such as the severity of the disease.
- Exposure – the types of antimicrobials used, duration and frequency of use and spectrum (broad or narrow) of effectiveness.
- Outcomes – primary and secondary outcomes for each study will be described.
Statistical analysis
Meta-analysis for quantitative data will be carried out for only those observational studies scoring at least six on the Newcastle-Ottawa scale. The analysis will be done using the more recent version of the R statistical software (R i386 4.0.2). Results for studies that show enough homogeneity will be pooled together using the random-effects model. Relative risks, calculated at 95% confidence intervals and two-sided p-values, will be used as the primary effect measure. If homogeneity is insufficient, quantitative results will be tabulated, and a narrative synthesis will be provided to summarise the results displayed in the included studies. Heterogeneity among the observational studies will be assessed using the Cochran's Q test (13) at a 5% significance level and quantified using the I2 statistic (13). An I2 value of more than 50% will be considered to reflect substantial heterogeneity and therefore will trigger sensitivity analysis to investigate the possible source of heterogeneity. For the results of the meta-analysis, publication bias will be assessed using a funnel plot (14). Details of each included study will also be presented in a table of study characteristics, from which exploratory, descriptive analyses will be done.
Analysis of subgroups or subsets
Subgroup analysis by age (65-80 years and more than 80 years), sex, ethnicity, type of organ damage and antimicrobial class will be carried out. Additional subgroups will be analysed, but it is not easy to specify them in advance.
Strategy for the presentation of results
For each database used, the final search strategy, and any other additional searches done, will be provided in a different file. The process of selecting articles from title screening up to final article inclusion into the final review will be shown on a flow diagram following the PRISMA guidelines (4). Information about the rationale of exclusion during full article screening will be presented in this flow diagram. A funnel plot will be used to demonstrate any potential small study effects and potential publication bias for any meta-analysis with at least six studies (15). The characteristics of studies tables and meta-analyses results of this review will be compiled into a publishable journal article. Other information will be included by text, descriptive statistics, and a summary table of findings. The GRADE tool (12) will be used to guide the construction of the summary of findings table. The full results on the quality of individual study assessment using the Newcastle-Ottawa scale will be provided in the additional file. However, the summarised table of these findings will be presented in the main article.
Ethical considerations
This systematic review and meta-analysis will not involve directly working with patients or human tissues. However, during the conduct of the review, ethical principles will still be observed to ensure integrity. Any potential conflicts of interests will be declared, all contributors acknowledged, and plagiarised material will be included in the review (16).
Dissemination
The systematic review and meta-analysis will be submitted for publication in a peer-reviewed journal in geriatrics. The findings will be presented at international conferences in the area of geriatrics or pharmacoepidemiology. This work will contribute to a PhD undertaken at the University of Bath. The results will be communicated to the patient and public engagement networks supported by the NHS Research and Development office, BSW (Bath, Swindon and Wiltshire).