Methods

Selection criteria and data collection

We performed a systematic search of SciVerse Scopus, PubMed MEDLINE, and Web of Science databases from inception of each database up to the 5th December 2022 using the following boolean search commands: [“biodiversity”] AND [“allerg*”] OR [“asthma”] AND [“environment*”] AND [“microbio*”]. The methodology used in this systematic review and meta-analysis followed the PRISMA reporting standards. The protocol was published in the PROSPERO (registration number: CRD42022381725) in December 2022 (https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42022381725). The full search strategy can be found in the online supplement (p 6).
A study was included in the systematic review and meta-analysis if (i) it was an original randomized controlled trial or an epidemiologic study (cohort, case-control or cross-sectional study); (ii) provided data on the relations between inner and/or outer layers of biodiversity and the risk of asthma, wheezing, and/or allergic sensitization, (iii) included an adequate definition and/or indexes of biodiversity and the main health outcomes (asthma, wheezing and allergic sensitization). Studies reporting only asthma exacerbation as the outcome or assessing the effect of antibiotics, medication or treatment were excluded. Studies assessing only the effect of endotoxins, fungi or virus, diet, breastfeeding, mode of delivery, use of probiotics/prebiotics/symbiotics, pet exposure, exposure to farm, rural, or urban environments, and to natural and green spaces were also excluded.

Exposure

The exposures of interest were the outer and/or inner layer biodiversity.6 The outer layer is dependent on the environment we live in (including soil, natural waters, plants, and animals). The inner layer inhabits human body (including gut, skin, and airways) and is influenced by colonization from the outer layer.6 The outer layer biodiversity was defined based on the environmental microbiota (including dust or soil) or on a score of environmental biodiversity (e.g., species richness index, land-use gradient, and plant diversity). The inner layer biodiversity was defined based on the diversity of human inner microbiota in the skin, stool, urine, and airway samples (including nasal, oropharyngeal, nasopharyngeal, and/or throat) (online supplement p 6).

Outcomes

The primary outcomes included asthma and the symptom wheezing. Different asthma definitions were considered for the inclusion of studies, such as the medical diagnosis, use of asthma medications, or the GINA-based definition.12 The secondary outcome was allergic sensitization which was defined based on a positive response to IgE antibodies for specific allergens, or a positive skin prick test (SPT).

Data extraction

Three reviewers (IP, NS, BH) independently performed the initial screening by the title and the abstract. After this initial screening, full articles were reviewed by two independent reviewers (TH, AR) and those fulfilling the inclusion criteria were selected for data extraction. When a conflict in assessment arose at any stage of the review, a consensus between the original and two additional reviewers (TH, AR) was required to achieve the resolution. Additional description of the data extraction process is presented in the online supplement (p 6).

Quality assessment and publication bias

The risk of bias was independently assessed by two reviewers (IP and NS) applying the Newcastle-Ottawa Scale (NOS) for cohort and case-control studies.13 An adapted version of the NOS developed by Herzog et al. 14 was used for cross-sectional studies. The results from the NOS were translated into the Agency for Health Research and Quality standards, and applying these the studies were classified as good, fair or poor.15

Statistical methods

A study was included in the meta-analysis if (i) it used a priori recognised measure to assess biodiversity (Shannon diversity index and/or bacterial richness) and (ii) it analysed biodiversity as a continuous variable. Meta-analysis applying the random-effects model was performed on the available data applying the R software, Version 1.4.1106 (dmetar package) and SAS software. To evaluate the effect of biodiversity, the data were summarised as the standardized mean difference (SMD) or the effect estimate (EE) with 95% confidence interval applying the Hedges method and the restricted maximum likelihood (REML) estimator, respectively. The magnitude of heterogeneity between the included studies was estimated using the Higgins I2 statistic and τ2. Publication bias was assessed through visual examination of a funnel plot and by applying the Egger’s test. Sensitivity analysis was performed according to the risk of bias/quality of evidence by comparing the summary effect estimates obtained by excluding studies with a high risk of bias with those studies considered as being at low and moderate risk of bias.