This study estimates the overall attack rate (AR) for influenza A(H1N1)pdm09 on a nine-hour commercial flight with six potentially infectious passengers on board to be 4·3%, but did not demonstrate an increased risk to those seated within two rows of an infectious case.
The AR in our present study is largely consistent with the published data from four previous reports that have suggested transmission of influenza A(H1N1)pdm09 during long-haul modern commercial air travel; 0·5% for all sections of a 13-hour flight with a single infectious passenger; 4·7% in a single section of a 15-hour flight with a single infectious passenger; 1·9% in a single section containing 12 infective passengers on a 13-hour flight; 2·4% estimated using pooled data from four flights of between 1·5- and 2·5-hour durations. Han et al. reported a single case (AR 1·1%) infected on a 45-minute flight containing two infectious cases.
Our study does not demonstrate an increased risk of becoming infected with influenza A(H1N1)pdm09 in those seated within two rows of an infectious case compared with those who were not. Two separate reports of single cases infected in-flight demonstrated that they sat two and six rows away from the known infectious case.[7, 8] Ooi et al.demonstrated that contact tracing of passengers within two rows of an infectious case would have only detected two of five of cases newly infected on a long-haul flight with a single known infectious passenger. The attack rate for those seated within two rows of any of 12 infectious cases on a flight from Los Angeles to Auckland was shown to be two of 57 (3·5%) compared with 0 of 46 for those seated elsewhere, with only the rear section of the plane studied. Following the study of two long-haul flights, Foxwell et al. suggested that the risk of contracting influenza A(H1N1)pdm09 from fellow passengers was increased by 1·4% (95% CI 0·5–3·4%) in those seated within two rows of an infectious case, but the overall response rate in this study was low at 43%, raising the possibility of selection bias. An investigation of two flights landing in North America, of between 1·5- and 2·5-hour duration, demonstrated a risk ratio of 3·0 (95% CI 0·7–12·7) for those seated within two rows of an infectious case of influenza A(H1N1)pdm09 compared with those who were not.
The results of laboratory testing in this present study are largely consistent with the clinical case definitions. All four infected in-flight cases tested using PCR were positive, suggesting specificity of the CDC clinical case definition; only one of four infectious cases tested was positive, but this is unsurprising given the length of time (5–15 days) between symptom onset in Mexico and return to the UK. Of the four infected in-flight cases confirmed using PCR testing, two of four (50%) cases were seated within two rows of an infectious case (RR 0·7; 95% CI 0·10–4·7). Serological testing was only available for two cases (both infected in-flight) with ILI and was negative in one case with a previously positive PCR, perhaps as a result of early timing of the test or a false-negative result following administration of oseltamivir. Serological data were considered in the light of potential cross-reactivity of antibody from seasonal influenza A(H1N1); antibody titres ≥1:32 by HI were only considered ‘suggestive’, not as confirmation of previous infection. Furthermore, serological data were only interpreted in the context with additional information, namely PCR and ILI data.
Influenza transmission in-flight can occur via several pathways: direct physical contact, fomites, direct droplet spread and suspended small particles. Aircraft cabins contain air that is recycled through high-efficiency particulate air (HEPA) filtration via a laminar flow pattern, reducing longitudinal spread fore and aft of the vessel. The risk of airborne transmission of pathogens is therefore likely to be greatest in adjacent rows, although computational studies have demonstrated spread of droplets up to seven rows within 4 minutes; also perturbations of airflow are likely to occur as passengers walk through the cabin. Influenza virus-containing aerosol shedding can occur during normal tidal breathing, leading to suspension of particles in ambient air and potential disease transmission during air travel.[21, 22] In one study, 70% of exhaled aerosol particles in subjects with influenza were between 0·3 μm and 0·5 μm, therefore largely removed if passed through HEPA filtration. Low absolute humidity on aircraft has the theoretical potential to increase influenza survival and transmission. Using quantitative microbial risk assessment, Wagner et al. demonstrated that the risk of contracting influenza A(H1N1)pdm09 is unlikely to extend beyond each individual cabin and is lower in first compared with economy class; their estimates of 5–10 new infections from a single case over an 11-hour flight are consistent with the findings of this present study.
Limitations of this study include selection and measurement biases, limited sample size and the potential for influenza A(H1N1)pdm09 transmission beyond the flight.
Information sufficient to determine case status was available for 239 of 278 (86·0%) passengers on board the flight, with more detailed clinical data obtained from 224 of 278 (80·6%) passengers, the high response reducing the potential effect of selection bias. No data are available regarding the demographics or case status of non-responders to the cohort study and contact tracing exercise, unless recorded during enhanced surveillance. However, there is no suggestion that this group were aware of, and therefore could be selected by, the exposure studied. The majority of passengers with unknown case status belonged to groups of four to six travellers, where only contact details of the lead passenger were available, further suggesting that non-response was unrelated to individual case status or exposure, and therefore unlikely to bias the results of this study. The use of a clinical case definition has the potential to result in incorrect classification of case status, thereby under- or overestimating the attack rate. The specificity of 100% for the four infected in-flight cases tested with PCR provides some validation to the CDC case definition chosen. The validity of defining true infectious cases was potentially reduced by the low PCR positivity rate (1 of 4; 25%) and lack of serological testing among this group. The potential for observer bias was minimised as interviewers would not be aware of the location of infectious cases in the cohort study, but recall bias among respondents aware of the reported cases remains a possibility.
Despite the large number of passengers on the flight, the low attack rate reduces the power in this study to test the primary hypothesis, the AR in those seated or not seated within two rows of an infectious case was almost identical, but a much larger study, in practical terms a meta-analysis of several flights, would be required to test the hypothesis with statistical integrity.
In calculating the AR, the presumption is made that all newly infected cases contracted influenza A(H1N1)pdm09 during the course of the flight, but this requires the validity of several assumptions. The incubation period used in this study was 0–6 days, estimated from modelling studies among the UK population following the pandemic. A systematic review of influenza A suggested an incubation period of 0·7–2·8 days, but a longer time period in this study was used because it is influenza A(H1N1)pdm09 specific and onset of disease up to 6 days after the flight was considered more likely to be from passenger contacts than elsewhere. Reported median incubation periods for influenza A(H1N1)pdm09 include 0·8, 2·1, 3 and 4·3 days, all consistent with the pattern of the epidemic curve in this present study.[13, 27-29]
Despite the first reported cases of influenza A(H1N1)pdm09 arriving on the flight studied, there is evidence to suggest circulating disease in the UK prior to this. A phylogenetic study estimated from divergence that two of 13 strains of influenza A(H1N1)pdm09 may have been circulating up to 8 days before the arrival of the flight, although the risk of transmission to the passengers studied remains unknown.
Exhaustive attempts were made in the cohort study to identify common links, such as accommodation and excursions in Mexico, family groups and toilet use on board. No clear links were identified apart from that between infectious case 5 and infected in-flight case 13, but the potential for becoming infected prior to the flight in Mexico remains for all infected passengers, and therefore the AR may be an overestimate.
Other potential opportunities for infection include via aerosol spread while travelling in close proximity to an infectious case on a low-humidity air-conditioned charter bus to the airport and via direct contact in a potentially crowded departure lounge.
At the time of the cohort study, the aircraft toilet locations were not known to health protection staff, and there is likely to be considerable non-differential misclassification in self-reported use. Therefore, any increased risk of illness, if present, associated with using the same toilet as an infectious case was unlikely to be identified by this study. The onset of symptoms 5–15 days prior to the flight for infectious cases raises the possibility of resort exposure leading to the infection of those infected in-flight and demonstrates the additional possible exposure to circulating disease in Mexico at the time of the flight. The shape of the epidemic curve, albeit for a small number of cases, does suggest increased transmission around the time of the flight.