In most healthy individuals without
However, none of the existing reviews (10, 11) systematically assessed observational studies among beekeepers worldwide, addressing the epidemiology of SAR to
A comprehensive systematic literature review and meta-analysis were performed following the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement (14). The study protocol was preregistered in the International Prospective Register of Systematic Reviews (PROSPERO) under the identification number CRD42021260922 and is described elsewhere (15).
We included questionnaire-based observational (cohort, cross-sectional) studies, assessing the estimated prevalence of self-reported SAR to
Exclusion criteria comprised studies assessing reactions other than self-reported SAR to
A systematic electronic literature search was carried out by two reviewers (TC, AK) across seven databases: MEDLINE via PubMed, Web of Science Core Collection, Scopus, Academic Search Complete (EBSCO host), ScienceDirect, Cumulative Index to Nursing and Allied Health Literature (CINAHL, EBSCO host), and Zoological Record (Web of Science). The search spanned from their inception up to August 3, 2021, and was subsequently repeated between July 11, 2022, and finalized on August 1, 2023. The search strategy, employing two search terms (“hypersensitivity” AND “beekeeping”), was reviewed by an experienced librarian and initially formulated in MEDLINE via PubMed, with subsequent adaptation for use in other electronic databases (data available on request). Manual search of the reference lists was conducted to identify any relevant publications that might have been missed in the electronic search.
The search results underwent initial screening for duplicate removal and record management using the Zotero reference manager, either automatically or through manual upload. The selection process, encompassing title and/or abstract screening, as well as a full-text review based on the eligibility criteria, was independently conducted by two reviewers (TC, AK). When necessary, a third reviewer (IL) was consulted.
Prior to final data tabulation, a pre-designed data extraction form in Excel was prepared. Two independent reviewers (TC, AK) manually extracted the following data of interest: study characteristics (first author; year of publication; location, study design, aim); observed population characteristics (sample size, age, gender); observed health outcome, method of data collection and statistical analysis. The observed health outcome was defined as the estimated lifetime (≥10 years) and/or one-year prevalence of self-reported SAR to
The quality assessment of the included studies was conducted using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Studies Reporting Prevalence Data (16) (data available on request).
The meta-analysis of the prevalence data and forest plot construction were performed under the R statistical environment (version 4.3.1), utilizing the function metaprop within the “meta” package. Due to expected high heterogeneity, the random effect model was applied and the restricted maximum-likelihood estimator (REML) was used for calculating between-study variance (□2). Overall prevalence was calculated using the logit transformation. Confidence intervals of prevalence for individual studies were calculated based on exact binominal intervals.
The initial database search yielded a total of 468 publications. After removing duplicates (n=235) and conducting screening based on titles and/or abstracts (n=233), we identified eight articles that met the criteria for full-text assessment. All of these (17,18,19,20,21,22,23,24) were eligible for inclusion and no additional studies from the reference lists were added. The PRISMA 2020 flow diagram (Figure 1) outlines the selection process.
The majority of the included studies (n=5) were conducted outside Europe, with four in Turkey (18, 22,23,24) and one in Mexico (21) (Table 1). There was one study each from Northern (Finland) (17), Central (Germany) (19) and Western Europe (Great Britain) (20). The studies were published between 1996 and 2020, and the majority were published after the year 2000 (18,19,20,21,22,23,24).
The primary objectives of the studies were to assess the prevalence and types of HVA, specifically stinging
Key items of the self-reported observed health outcome.
Finland | 191 | 164 | 51.8±12.3 | ✓ | ✓ | Müller | bee wasp ( |
|
Turkey | 1245 | 489 | 48.2±11.5 | NA | bee | |||
Germany | 1053 | 973 | 61.8±13.9 | ✓ | ✓ | Müller | bee | |
Great Britain | 852 | 545 | range 51–60 | ✓ | modified Müller | bee | ||
Mexico | 1541 | 1289 | average 37 | ✓ | Müller | bee | ||
Turkey | 301 | 295 | 48.2±11.5 | Müller | bee | |||
✓ | Ring-Messmer | |||||||
Turkey | 221 | 213 | 49.9±11.8 | ✓ | Ring-Messmer | bee | ||
Turkey | 69 | 57 | 48.4±12.0 | ✓ | NA | bee |
Legend: N=number of participants; NA=not available; SD=standard deviation
a:*additional data gathered upon request
b: ✓ indicates the observed measure(s) of occurrence
Epidemiological data were collected using questionnaires, distributed through various methods, (sending by mail (17, 21), being included in selected journals and sent to subscribers, or made available in electronic form on the internet (19), or only in electronic form (20)). In one study, printed questionnaires were completed during a beekeeping congress meeting under the supervision of the researchers (24). In a few studies how the questionnaires were distributed was not clearly specified (18, 22, 23). Survey response rates varied widely, ranging from as low as 3.0% (19) to 79.6% (17).
Except for three studies (17, 18, 22), the prevalence period for the estimated lifetime self-reported SAR to
A summary of the RoB assessments for each study is available on request. None of the studies met all 10 evaluation points for quality assessment, with more than half (n=5) exhibiting a high RoB. Measurements of the outcome (Q7.1 and Q7.2) and the statistical analysis (Q8) applied to a high RoB in most cases.
The estimated overall lifetime prevalence of self-reported SAR to bee venom, graded for severity according to different classification systems, was 23.7% (95% CI: 7.7–53.4). A substantial level of heterogeneity was observed among the studies (I2=99%, p<0.01, Figure 2A). The estimated lifetime prevalence of self-reported SAR to bee venom for grades III–IV, graded for severity according to different classification systems (classification data not provided in one study (24)), was 6.0% (95% CI: 3.0–11.7). A significant degree of variability in reported event rates was observed (I2=93%, p<0.01, Figure 2B).
The estimated overall one-year prevalence of self-reported SAR to bee venom, graded for severity according to different classification systems (classification data not provided for one study (18)), was 7.3% (95% CI: 5.8–9.2). There was no heterogeneity observed among the studies (I2=0%, p=0.43, Figure 3).
To the best of our knowledge, this systematic literature review and meta-analysis represents the first comprehensive attempt to estimate the global prevalence of self-reported SAR to
Two major reasons could contribute to the observed heterogeneity in the study. Firstly, they may reflect methodological differences, including data collection technique, definition of AR and utilization of diverse classification systems for grading the severity of SAR across different regions (12, 25, 26).
With regard to the definition of AR, variability in how studies categorized and reported AR was observed. With the exception of one study (23), the questionnaires did not distinguish between allergic and non-allergic reactions (i.e., systemic toxic reactions, psychogenic reactions), raising the potential for a false history of self-reported SAR to
Importantly, putting aside the fact that a classification other than Müller was used, a high estimated overall lifetime prevalence of self-reported SAR to bee venom (37.6%) was also reported by Ediger et al. (23). However, these results may actually reflect the incidence of new cases rather than the overall prevalence, as the authors observed the course of symptoms over the years following bee stings. Meanwhile, Münstedt et al. (19), aimed to report the incidence of bee venom allergy, but a detailed textual analysis revealed that they reported the prevalence instead. Compared to the other studies, the authors also reported the lowest overall lifetime prevalence of self-reported SAR to bee venom, partly attributed to variations in the age of participants, with the mean age being approximately a decade higher compared to the data from other studies (17, 18, 22,23,24). Nevertheless, an important shortcoming of the German study is its very low response rate (3.0%), which may affect the accuracy of the prevalence estimates.
Secondly, the observed heterogeneity might be attributed to genuine disparities in sting exposure across different regions, influenced by geographic locations, climate, and beekeeping practices (11, 12, 25, 26).
In relation to bee sting exposure, Bousquet et al. (31) noted a strong correlation between the degree of sensitization to bee venom and the annual number of bee stings. This correlation is most prominent when the annual number of stings falls below 25 and reaches an optimum when it exceeds 200 (31). Aligning with this data, the potential protective effect of higher sting frequencies, as observed in Turkish studies (18, 22), and less so in the Finnish beekeepers (17), may explain the lower estimated overall one-year prevalence of self-reported SAR to bee venom in Turkey (6.5% (18), 7.0% (22)) compared to Finland (9.4%) (17). However, in Bousquet et al.’s study (31) the specific selection criteria employed (exclusion of numerous allergic beekeepers and individuals with variations in the number of annual bee stings over the previous five years) may have influenced the study’s outcomes.
Nonetheless, an intriguing pattern emerges within the estimated lifetime prevalence of self-reported SAR to bee venom in grades III–IV (severe SAR). Studies consistently indicate a higher estimated prevalence of severe self-reported SAR to bee venom in colder European regions (Finland, Great Britain) (17, 20), and a lower one in warmer non-European ones (Mexico, Turkey) (21, 24). In the latter case, it is plausible that favourable climatic conditions permit beekeepers to be exposed to bees throughout most of the year (18, 22, 32), leading to a lasting form of immunological protection (33), presumably on an immunological basis. Notably, heavily exposed beekeepers exhibit higher levels of bee-venom specific IgG4 (sIgG4), reflecting their degree of exposure to stings and believed to induce immune tolerance while mitigating the inflammatory response (34). However, the results of one Turkish study led by Ediger et al. (23) deviate from this expected pattern. It is suggestive that these findings may not be solely attributable to geographic location and climatic conditions, as observed in other studies. Instead, it is conceivable that methodological concerns, as mentioned previously, could significantly impact the observed outcome. However, the risk of developing SAR to bee venom cannot be entirely ruled out, even among beekeepers with a history of numerous bee stings and no prior AR (35, 36).
In contrast to beekeeping in regions with milder climates, apiculture in Europe is inherently seasonal, characterized by the distinct absence of bee sting exposure during the winter months, with this seasonal break lasting from the end of October throughout the entire winter (33). Moreover, the length of the beekeeping season varies across the regions, i.e., in Finland it extends from May to August (17), while in France it spans from early spring to late fall (31). It is conceivable that the natural history of sting reactions may exhibit disparities between the northern and southern regions (17). Moreover, it is plausible that differences between countries may also be due to beekeeping with different subspecies of honeybees (e.g.,
However, regardless of the location, the temporal gap between two working seasons may potentially attenuate the protective effect conferred by prior bee stings, consequently increasing the susceptibility to the development of AR (37). This aligns with the conclusions drawn from a literature review (10), as initial stings in spring were identified as a definitive risk factor for the onset of allergic bee sting reactions among beekeepers. Furthermore, Münstedt et al. (19), reported the occurrence of more severe non-allergic reactions to bee venom during the spring months when compared to later periods. It is also important to consider the impact of climate change, as the available data support the presence of positive correlations between climate change and HVA (38).
Knowing that factors such as geographical location, climate differences, temperature fluctuations, and insect behaviour patterns can heighten the risk of insect stings within this population group, targeted public health interventions are essential. This includes implementing comprehensive risk assessment and management strategies, as well as launching public health campaigns and educational initiatives aimed at raising awareness about SAR and promoting preventive measures.
For allergic beekeepers, the most critical measure to mitigate risk is to reduce exposure by considering cessation of beekeeping activities. However, our meta-analysis reveals that many allergic beekeepers continue beekeeping, thereby exposing themselves to recurrent and potentially life-threatening SAR. Therefore, allergologists, public health professionals, and occupational, traffic and sports medicine specialists should intensify efforts in counselling, emphasizing 1) the importance of wearing full protective equipment during all beekeeping activities, 2) self-medication in emergencies, including regular training in proper use of adrenaline autoinjectors, and 3) considering Venom Immunotherapy (VIT) as a causal treatment option when indicated. In particular, life-long VIT should be considered for individuals with inherited or acquired risk factors.
Finally, when considering the extent of exposure, the differences in beekeepers’ status (professional or hobbyist) may also affect the outcome. As only one study included professional beekeepers (17), it would be intriguing to investigate potential differences between these two groups in future research. Moreover, an important knowledge gap is the lack of information regarding the location of hives (i.e., in a rural or urban environment). Urban beekeeping considerations are essential for public safety, as the estimated prevalence of self-reported SAR to
For future cross-sectional studies, detailed reporting of study design, settings, study participants, and the use of validated questionnaires should be employed to ensure high-quality assessment of the observed health outcomes. In order to reduce the overestimation of the self-reported data, the observed health outcomes should be confirmed by an allerogologist. In terms of data collection, comprehensive reporting of health outcomes should include essential elements such as the classification system used (e.g., Müller grading system), grade of SAR, identified culprit
The quality of our work is subject to several limitations, primarily stemming from the high heterogeneity in self-reported data among the included studies and a high RoB. Additionally, the predominant reliance on Turkish data in more than half of the eligible publications, along with the absence of data about the beekeepers’ status, raises concerns about the generalizability of these findings. Therefore, caution is needed when interpreting the results, especially considering the exclusion of clinical studies from our analysis. Specifically, the lack of conformity between self-reported SAR to bee venom observed by beekeepers and verification by physicians suggests potential overestimation, particularly for mild/moderate SAR (grades I–II).
Moreover, the meta-analysis was conducted on cross-sectional studies, thereby limiting the ability to infer causality or temporal relationships. Additionally, our study was constrained by the small number of available studies, which is reflected in the very wide confidence intervals of the reported estimates. However, our research was able to clearly distinguish between the one-year and lifetime prevalence of self-reported SAR to bee venom, thereby explaining much of the variability in the results. Consequently, individual forest plots included only three, four or five studies. We acknowledge that some of the studies are of poor quality, with inadequate reporting. However, the estimates provide valuable indicative trends that can guide further research and highlight areas where larger, better designed, and more comprehensive studies are needed. For public health professionals and policymakers, even these unstable estimates can raise awareness about the significance of SAR among beekeepers, prompting preliminary guidelines and interventions aimed at mitigating risks until more robust data become available.
Nonetheless, our study’s main strength lies in its rigorous methodology. It included comprehensive searches across seven electronic databases without language or publication date restrictions, adhering to PRISMA 2020 guidelines. We made extensive efforts to obtain additional information from the authors, not available in the original articles, and used a JBI algorithm with predetermined criteria, facilitating objective quality assessments. Moreover, we identified methodological aspects warranting improvement in future research, which will help mitigate potential sources of bias and enhance the robustness of estimates. By identifying research gaps and exploring the major sources of heterogeneity across the included studies, our findings could contribute to a more comprehensive understanding of the existing research limitations in this field of science.