Open Collections

<span lang="EN-US"><strong>Background</strong>:</span><span lang="EN-US"> People living with cystic fibrosis have an increased risk of lung infection with non-tuberculous mycobacteria (NTM), which is reportedly increasing. We conducted a systematic review of the literature to estimate the burden (prevalence and incidence) of non-tuberculous mycobacteria in the cystic fibrosis population. </span></p>

<span lang="EN-US"><strong>Methods</strong>: Electronic databases, registries, and grey literature sources were searched for cohort and cross-sectional studies reporting epidemiological measures (incidence and prevalence) of NTM infection or NTM pulmonary disease (NTM-PD) in cystic fibrosis. The last search was conducted in September 2021; we included reports since database creation and registry reports published since 2010. The methodological quality of studies was appraised with the Joanna Briggs Institute tool. A random-effects meta-analysis was conducted to summarize the prevalence of NTM infection, and the remaining results are presented in a narrative synthesis. </span></p>

<span lang="EN-US"><strong>Results</strong>: Ninety-five studies were included in this review. All 95 studies reported on NTM infection, and 14 of these also reported on NTM-PD. The pooled estimate for the point prevalence of NTM infection was 7.9% (CI 95%, 5.1–12.0%). In meta-regression, sample size and geographical location of the study modified the estimate. Longitudinal analysis of registry reports showed an increasing trend in NTM infection prevalence between 2010 and 2019. </span></p>

<span lang="EN-US"><strong>Conclusions</strong>: The overall prevalence of NTM infection in CF is 7.9% and is increasing over time based on international registry reports. Future studies should report screening frequency, microbial identification methods, and incidence rates of progression from NTM infection to pulmonary disease.</span></p>; <b>Methods</b><br />

<strong><span lang="EN-US">Review question</span></strong></p>

<span lang="EN-US">We designed our review question based on population, condition, outcome (epidemiological measure) and study design, as recommended by current guidelines. Briefly, we screened for cross-sectional or cohort studies reported in English including people with CF (population) and evaluating NTM infection or NTM-PD (condition). NTM infection was defined as isolation of any NTM on at least one occasion per patient; the criteria for NTM-PD were specified in each study. Reporting of at least one epidemiologic measure among incidence rate, incidence proportion, point prevalence, or period prevalence was required for inclusion. The full criteria are described in Supplementary Table 1. The review protocol was registered to the </span>International Prospective Register of Systematic Reviews, PROSPERO<span lang="EN-US"> (CRD42020200418) in July 2020. In October 2020, before the abstract screening, we updated the grey literature sources and screening procedures.</span></p>

<strong><span lang="EN-US">Literature search</span></strong></p>

<span lang="EN-US">EMBASE and MEDLINE were searched in September 2020 using the criteria specified in Supplementary Methods 1; an updated search was conducted in September 2021. We manually reviewed the conference proceedings from relevant research meetings between 2010 and 2020 (North American Cystic Fibrosis, European Cystic Fibrosis Society, American Thoracic Society, and the Infectious Diseases Society of America conferences). Also, we performed forward and backward searches for highly cited references in Web of Science (listed in Supplementary Table 2) using Google Scholar and Web of Science. Finally, the United States of America (CF Foundation), Canadian (CF Canada), European (European CF Society), Australian, and Brazilian registry reports published between 2010 and 2021 were included.</span></p>

<strong><span lang="EN-US">Screening and data extraction</span></strong></p>

<span lang="EN-US">All records were retrieved and exported in </span>Research Information Systems <span lang="EN-US">format. Initial manual deduplication evaluated Author, Title, and Year of publication. Then, we performed automated deduplication using the SRA De-Duplicator software and Covidence. Screening of reports and full-text manuscripts, data extraction, and risk of bias assessment was conducted independently by two reviewers (M.P. and M.A.); discrepancies were solved by consensus or by a third reviewer (B.Q.). Epidemiological measures of interest reported in each study were included for analysis. Abstract screening evaluated language, study type, the inclusion of CF population, and reporting of any measures of interest. Full-text screening evaluated all eligibility criteria defined in Supplementary Table 1. For unretrievable reports, we requested access to unpublished full manuscripts from authors via email on at least two separate occasions. The Joana Briggs Institute tool was used to assess methodological and reporting quality. Overall low risk of bias was defined as low risk in the assessments of the sampling frame, sample size, population description, and statistical methods. High risk was determined by a high-risk assessment in any of the following: sampling frame, sampling scheme, sampling size, population description, identification methods, or statistical calculation. Data extraction was based on a pre-specified data dictionary piloted with 10 studies (Supplementary Table 2). For period prevalence, point prevalence, and incidence proportion, we extracted proportions, the number of cases, and the sample size. We did not impute any missing data. In studies with unclear years of data collection, we assumed that data was obtained from the year before publication. The body of evidence was not evaluated for certainty given the lack of adapted tools for single-proportion measures.</span></p>

<strong><span lang="EN-US">Data analysis</span></strong></p>

<span lang="EN-US">Data were analyzed with the meta and metafor packages in R studio and R version 4.1.1. Risk of bias plots were produced with the robvis and ggplot2 packages, and tables with the flextable package. We pre-specified the use of random-effects models based on expected heterogeneity by study region and dates. To model proportion data, we used generalized linear models with LOGIT transformation. Point and annual prevalence of NTM infection were summarized together in the meta-analysis because they contain comparable time frames of evaluation (a year or less). The remaining epidemiologic measures including period prevalence of NTM infection, incidence of NTM infection, prevalence (point or period) of NTM-PD, and incidence of </span>NTM-PD<span lang="EN-US"> are reported in supplementary tables and text only. Period prevalence of NTM infection and NTM-PD were not pooled due to varying time intervals among studies, while the rest of the epidemiologic measures had a small number of studies. To avoid the overrepresentation of registry reports in the meta-analysis, we included only the most recent report per registry with both numerator and denominator available to calculate prevalence. Secondary data analyses of registry data were also excluded from meta‑analysis to reduce redundancy with the registry reports. Heterogeneity was assessed with the I2 index and 95% confidence interval, with a significance level established at p &lt; 0.10. Publication bias was explored graphically using sample size as a predictor of bias in the funnel plot.</span></p>

<span lang="EN-US">We pre-specified subgroup analyses by study design, age category (pediatric vs adult), year of data collection (before 2000, 2001 to 2009, and 2010–2019)</span><span lang="EN-US">, </span><span lang="EN-US">geographical region (grouped as North America, Europe, and others), and the most common individual NTM species reported in CF (<em>Mycobacterium</em> <em>abscessus</em> complex (MABs) and <em>Mycobacterium</em> <em>avium</em> complex (MAC)). The prespecified meta-regression model was optimized by maximum likelihood and used the same transformation as the meta-analysis (LOGIT). We evaluated the goodness of fit in the model using Akaike’s information criteria by stepwise inclusion of pre-specified coefficients. Exploratory (unspecified) analyses include a longitudinal trend of prevalence in registries and subgroup analyses by region for MAC and MABs. Sensitivity analyses included three meta-analyses of NTM infection point (and annual) prevalence. The first excluded a study that screened patients only in the presence of increased symptoms, the second included only registry data, and the third excluded studies that did not use standardized culture media for identification of NTM. Reporting is based on the recommendations of the Joanna Briggs Institute and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist.</span></p>; <b>Usage notes</b><br />

Requires R version 4.1 or higher and a spreadsheet program (MS Excel) to explore the collected datasets. The R studio IDE was used to conduct R analyses and document procedures. The list of packages (with versions) necessary to reproduce the results is specified inside the scripts. </p>