The search strategy retrieved 1205 records from scientific databases (164 from PubMed, 756 from Scopus, and 285 from Web of Science) and 247 from grey literature sources. An additional of 35 articles were manually retrieved from the references of included studies and of relevant reviews. A number of 1126 titles and abstracts were evaluated after deduplication, and 1104 of them were discarded because they were not relevant for the review question. Of 22 full-text articles remaining, 4 were finally included in the qualitative analysis (Figure 1).
3.1. Characteristics of included studies
Table 1 describes in detail the four included studies [Caley et al., 2007; Gupta et al., 2012; Nicolaides et al., 2020; Zhang et al., 2013], which targeted the influenza virus infection. All studies were published between 2007 and 2020 and were from four different continents. Three studies used mathematical models [Caley et al., 2007; Gupta et al., 2012; Nicolaides et al., 2020], with two of them using single flight models [Caley et al., 2007; Gupta et al., 2012].
In one study, Caley et al. [2007] simulated a unique 12-h flight per day to travel from an epidemic region to a still infection-free region and to carry 10 to 400 passengers. In the second study, Gupta et al. [2012] simulated a unique 4-h flight in which a fully occupied twin-aisle cabin carried the index passenger occupying the center seat of the cabin. The twenty passengers around the index passenger (six on the same row, and seven on the front and on the back rows) represented the study population. In the third study, Nicolaides et al. [2020] simulated multiple flights to move through a large network of 120 international airports in order to mimic a global spread of viral disease.
Unlike other three studies, the study by Zhang et al. [2013] described the contact tracing of a real influenza outbreak that involved two flights during the 2009 influenza A H1N1 pandemic. One flight carried 274 passengers from New York City (United States) to Hong Kong (China), with stopover in Vancouver. Sixty-three passengers, including the index patient, continued to travel on a connection flight from Hong Kong to Fuzhou (China), which carried 144 passengers in total. Contact tracing identified eight secondary influenza A H1N1 cases, 7 in Fuzhou and 1 in Hong Kong, and all the infected passengers shared the flight from New York to Hong Kong, where transmission could have taken place. The basic reproductive number (R0) was provided in two studies only, and was of 1.5–3.5 in one study [Caley et al., 2007] and 3.0 in the other study [Gupta et al., 2012].
3.1.1. Studies evaluating use of facemask as a preventive measure
Regarding the use of facemask during air travel, Gupta et al. [2012] analyzed specifically N95 respirator, whereas no details on the type of facemask were provided in other two studies [Caley et al., 2007; Zhang et al., 2013].
Caley et al. [2007] explored several variables that might affect the time delaying the epidemic onset in an infection-free region following importation through air travel. The authors found that maximal compliance to facemask use and other non-pharmaceutical control measures (i.e. border screening, flight-based quarantining, or immediate presentation at the onset of symptoms) affected this time much less than the number of travelers per day, with a modest impact of R0 values. For example, during a 12-h travel of 400 passengers per day, facemask use increased the median time delay from 57 to 79 days at an R0 value of 1.5 and from 17 to 20 days at an R0 value of 3.5. Conversely, adopting facemask use together with another non-pharmaceutical control measure and reducing passenger numbers from 400 to 10 per day both delayed the time to 125 days at an R0 value of 1.5 and to 26 days at an R0 value of 3.5.
In the study by Gupta et al. [2012], a computational fluid-dynamic simulation allowed to estimate the quantity and distribution of influenza virus particles of a single-cough exhalation (measured as “quanta” per hour). The effect of N95 respirator on the risk of infection was evaluated as a function of the inhaled influenza virus particles. The infection probability wearing N95 mask was reduced from 15% (3/20) to 0% (0/20) at 103 exhaled quanta per hour or from 100% (20/20) to 55% (11/20) at 5226 exhaled quanta per hour.
In the study by Zhang et al. [2013], a case-control analysis allowed to identify wearing a facemask during the flight as a significant protective factor from influenza A H1N1 infection on the flight from New York City to Hong Kong. Thus, none of the 9 infected passengers compared to 15 (47%) of 32 healthy control passengers wore a mask (odds ratio, 0.0; 95% confidence interval, 0–0.7).
3.1.2. Studies evaluating hand hygiene as a preventive measure
Two studies evaluated hand hygiene as a mitigation strategy [Nicolaides et al., 2020; Zhang et al., 2013]. In the study by Nicolaides et al. [2020] using Monte Carlo simulation, four hand-hygiene scenarios were hypothesized. In one scenario, increasing the percentage of people that in all airports cleaned hands at any time from 20% (i.e. one over five people) to 30%, 40%, 50%, or 60% allowed to reduce the infection prevalence of 18.2%, 33.0%, 45.2%, and 55.4%, respectively. Similarly, the total square displacement of infected people—i.e. the measure of airports’ power to spread a disease across the globe—was reduced of 23.7%, 43.4%, 58.6%, and 69.1%, respectively. The three other scenarios explored the effect of a less expensive and more contained strategy, i.e. hand-hygiene implementation only in a subset of more busy airports, on aforementioned outcomes, thus leading to similar results.
In contrast, the study by Zhang et al. [2013] did not find significant differences in self-assessed hand-hygiene compliance (either hand-washing after toilet use or hand-cleaning by wet-towel before eating) between case and control passengers. In both passenger groups, high rates of compliance were reported, namely 100% (9/9 and 32/32, respectively) for washing hands after toilet use, and 89% (8/9) and 91% (29/32) for cleaning hands before eating, respectively.
3.2. Quality of included studies
The quality of evidence for aircraft-related transmission was possible only for one study based on real outbreak data [Zhang et al., 2013]. We evaluated positively the following criteria: index case classification (+1 point), secondary case definition (+2), contact tracing strategy (+2), and completeness of follow-up (+1). Zero point was assigned to timeliness of contact tracing and alternative exposure means for infection to flight were considered during investigation. Thus, the study was assigned a medium level of evidence (Table 1).