This study was a retrospective cohort investigation of the risk of transmission of influenza during a commercial airline flight. The cohort consisted of all passengers seated in the rear section of the aircraft, with a further susceptible subset identified on the basis of interviewing and testing for pandemic A/H1N1. The main outcome measure was the incidence of in-flight infection with pandemic A/H1N1. Secondary outcome measures were the sensitivity and specificity of symptoms of influenza for identifying laboratory confirmed cases and the completeness and timeliness of contact tracing.

The initial public health response focused on controlling transmission from the high school group. Nasopharyngeal swabs were obtained from those with symptoms of influenza. After identification of influenza A in members of the group, the decision was made to trace all passengers on the flight and manage them to contain the spread of influenza (with the recommended protocol including nasopharyngeal swabs from those with symptoms, home isolation or quarantine, and antiviral post-exposure prophylaxis or treatment as appropriate). A flight manifest was obtained from the airline and arrival cards from immigration authorities in Auckland to identify passengers and their onward travel plans. Details were circulated to NZ public health units in the respective districts or international destinations in receiving countries.

A follow-up investigation began during the week of 4 May 2009. All members of the high school group were interviewed with a standard questionnaire covering illness and history of symptoms before, during, and after the flight. Follow-up serological specimens were also collected from the student group 16-23 days after the flight.

The affected students were seated in the rear section of the aircraft, so this population became the focus of the retrospective cohort study. We used the seating plan and passenger lists to construct a record of all passengers in this section. We retrieved the following data on these passengers from public health units involved in the initial response: symptoms during and after the flight, results of any laboratory testing for pandemic A/H1N1, timeliness and types of public health management. We re-interviewed passengers who reported symptoms, and those for whom no symptom history was recorded, by using a standard questionnaire. For assessment of the public health response, we defined passengers as having been “in transit” if they departed for another international destination within 24 hours of arrival or as NZ residents or visitors according to information collected during immigration processing. We used EpiInfo version 3.5.1 to analyse data.

The airline followed up cabin crew by using the same protocol as we used for passengers. We did not use results from this group in this study because we wanted to link the presence or absence of in-flight infection to seating locations, which were not applicable to mobile flight attendants. No illnesses were reported among cabin crew.

We placed nasopharyngeal and throat swabs in viral transport media and tested them by a real time polymerase chain reaction matrix assay for influenza A using primers from the Centers for Disease Control and Prevention, Atlanta, USA. We used an ABI 7000 Sequence Detection System (Applied Biosystems) for amplification as follows: reverse transcription 50°C for 20 minutes, Taq inhibitor inactivation 95°C for two minutes followed by 45 cycles of polymerase chain reaction amplification, 95°C for 15 seconds, and 55°C for 50 seconds. We sent specimens positive for influenza A RNA to a WHO collaborating centre for reference and research on influenza (Melbourne, Australia). Confirmatory testing was done by sequencing of the matrix, haemagglutinin, and neuraminidase genes for comparison with published influenza A sequences in GenBank or by real time polymerase chain reaction using primers that discriminate pandemic A/H1N1 and seasonal influenza A/H1N1, A/H3N2, and B sequences.

Sera were treated with enzymes that destroy receptors and assayed by haemagglutination inhibition using 1% turkey red blood cells for the presence of antibodies against reference viruses A/Auckland/1/2009 (pandemic A/H1N1) and A/Brisbane/59/2007 (seasonal A/H1N1) at the WHO collaborating centre. Seroconversion was defined as a fourfold rise in titre of haemagglutination inhibition between the acute and the convalescent serum or, when only one sample was available, the presence of a titre of at least 80 against A/Auckland/1/2009.

We considered passengers to have influenza-like illness if they had any two of fever or feverishness, cough, sore throat, or rhinorrhoea. We considered them laboratory confirmed if the nasopharyngeal swab was positive for pandemic A/H1N1, serology detected specific antibodies to pandemic A/H1N1, or both.

We considered a passenger to be a post-flight case of influenza if they developed laboratory confirmed or suspected influenza within a plausible incubation period after the flight. Assuming an incubation period of 0.6 to 3.2 days (based on values reported for influenza A8) and given a flight duration of 13 hours, this meant that illness from exposure on the flight could begin at any time after arrival in NZ to 3.2 days (77 hours) later.

We divided passengers into categories of pandemic A/H1N1 infection on the basis of their combination of symptoms, timing of symptoms, and laboratory results, as follows.

Laboratory confirmed symptomatic case during flight—Influenza-like illness starting within two weeks before or during the flight with at least one infectious symptom (cough, sneeze, rhinorrhoea) persisting during the flight and nasopharyngeal swab or serology positive for pandemic A/H1N1.

Suspected symptomatic case during flight—Influenza-like illness starting within two weeks before or during the flight with at least one infectious symptom (cough, sneeze, rhinorrhoea) persisting during the flight and pandemic A/H1N1 not excluded (laboratory investigation for pandemic A/H1N1 either not done or incomplete).

Immune case—Symptoms of influenza-like illness before the flight, no infectious symptoms related to this illness during the flight, and serology positive for pandemic A/H1N1.

Laboratory confirmed post-flight case—Influenza-like illness starting within 3.2 days of arrival in NZ and nasopharyngeal swab or serology positive for pandemic A/H1N1.

Suspected post-flight case—Influenza-like illness starting within 3.2 days of arrival in NZ and pandemic A/H1N1 not excluded (laboratory investigation for pandemic A/H1N1 either not done or incomplete).

Non-case—No symptoms of influenza during or after the flight, or symptoms of influenza during or after the flight and pandemic A/H1N1 excluded (laboratory investigations negative for pandemic A/H1N1).

Unknown status—Could not be contacted or insufficient information to assign to another category.

We considered a laboratory confirmed post-flight case to be a case of in-flight infection if they had no other plausible sources of pandemic A/H1N1 infection (before or after the flight). In addition, we defined pandemic A/H1N1 susceptible passengers as those who were seated in the rear section of the plane, excluding laboratory confirmed and suspected cases during the flight, immune cases, and those with unknown status. This susceptible group became the cohort for calculating the risk of in-flight infection with influenza.

The high school group had arrived on a Boeing 747-400 with a total capacity of 379 passengers. Of the 128 seats in the rear section of the plane, 126 were occupied by passengers, including 24 by the high school group (22 students and two teachers). All members of the high school group were interviewed. Eight of the 24 gave nasopharyngeal swabs, and 23 gave serological specimens.

Of the remaining passengers in the rear section of the aircraft, information on influenza symptoms during and after the flight was collected from 95% (97/102) as part of the initial response or during follow-up. Nasopharyngeal swabs were obtained from 26 of these passengers (14 of whom had symptoms). (See web table A for further details of laboratory testing.)

Table 1 shows the distribution of passengers according to categories of pandemic A/H1N1 infection. Out of a total of 121 passengers with known illness status nine were laboratory confirmed symptomatic cases during the flight. All were in the school group. Three further passengers with influenza-like illness and symptoms during the flight were incompletely investigated: one was a member of the school group (influenza A on nasopharyngeal swab, no pandemic A/H1N1 typing, serology not done); the remaining two were other passengers (nasopharyngeal swab testing not done). These three were categorised as suspected symptomatic cases during the flight.

Five passengers developed influenza-like illness after the flight. One of them developed symptoms six days after arriving and so was excluded. The remaining four developed symptoms within 3.2 days of arriving in NZ and so met the case definition for post-flight cases. Of these, three were laboratory confirmed and one had incomplete pandemic A/H1N1 investigation (see web appendix for details of these individual cases).

The figure shows the seating position of all the passengers and distinguishes them according to the categories of pandemic A/H1N1 infection used in table 1. All four post-flight cases were seated within two rows of confirmed symptomatic A/H1N1 cases. The nine laboratory confirmed symptomatic cases during the flight and four post-flight cases were all in the 10-19 year age group. Web table B shows further demographic and travel history details of the passengers.

Two passengers (A and B in the web appendix) with laboratory confirmed post-flight infection did not have another plausible source of infection and therefore met our criteria for in-flight infection. Passenger A developed symptoms of influenza 12 hours after the flight, and passenger B became ill 48 hours after arrival in NZ. Two others (C and D in the web appendix) also became ill after the flight. One (passenger C) could potentially have been infected before travel from other members of the student group and so is best described as possible in-flight infection. The timing of onset of symptoms of the other (passenger D) excluded infection before travel, but this case also remains inconclusive because the laboratory investigation was incomplete.

Table 1 summarises the susceptible cohort. We considered 107 passengers seated in the rear section of the plane to be susceptible to pandemic A/H1N1 infection during the flight on the basis of excluding laboratory confirmed cases during the flight (n=9), suspected cases during the flight (n=3), immune cases (n=2), and those with unknown status (n=5). Of the susceptible population, 57 were seated within two rows of laboratory confirmed symptomatic cases during the flight.

We estimated the overall risk of in-flight infection in the rear section of the plane to be 1.9% (95% confidence interval 0.3% to 6.0%). For the 57 passengers sitting within two rows of the laboratory confirmed symptomatic cases the risk was higher at 3.5% (0.6% to 11.1%).

Table 2 shows the prevalence of symptoms reported by passengers in the rear section of the aircraft during the period from the start of the flight until four days after its arrival. One symptom (cough) was very sensitive (92.3%) for pandemic A/H1N1 infection, had relatively high specificity (78.8%), and had moderate positive predictive value (63.2%). The surveillance case definition for influenza-like illness used in the United States had low sensitivity (38.5%), high specificity (90.9%), and moderate positive predictive value (62.5%).8 The screening case definition of influenza-like illness subsequently used for detecting potentially infected arriving passengers in NZ was more sensitive (61.5%) than the US definition but had lower specificity (72.7%) and positive predictive value (47.1%).

The New Zealand Ministry of Health introduced a contact tracing protocol shortly after this flight arrived in NZ. This protocol advised that contacts should be located rapidly and interviewed, have a nasopharyngeal swab if respiratory symptoms were present, be given a course of oseltamivir (this was a five day course regardless of symptoms), and be put into home isolation or quarantine for 72 hours from the start of oseltamivir treatment. This investigation assessed the extent and timeliness of this follow-up.

The results (table 3) show that follow-up was relatively complete but not particularly timely. Nearly all (93%) of the passengers outside the initially identified school group received public health service follow-up, including four passengers followed up in Australia. Only 52% were followed up by public health services within 72 hours. Once identified, 81% (77/95) received the full recommended public health management. Follow-up was less complete for visitors and people transiting through NZ.

A limitation of the investigation was that during the initial response phase passengers were interviewed by several personnel, so this process was not as complete and uniform as would be desirable. Some characteristics, notably symptoms, may therefore have been under-reported, which could have reduced case ascertainment. Time delays in interviewing would have produced further recall bias, again probably lowering reporting of symptoms. Incomplete laboratory testing of passengers in the rear section of the plane means that some infected passengers, particularly those with mild symptoms or who were asymptomatic, could have been missed. The course of oseltamivir offered to most of the passengers might also have suppressed symptoms in some (although treatment generally started several days after arrival, so this effect was probably small). Our case definition for in-flight transmission was conservative. On balance, the sources of error in this investigation would tend to underestimate the risk of in-flight transmission of pandemic A/H1N1.

This investigation provides insights into the control measures that are needed during the containment phase of a pandemic response when air travel can rapidly disseminate new infections.17 18 Some equivocal evidence from the 2009 A/H1N1 pandemic indicates that countries which implemented entry screening may have briefly delayed local transmission of this virus. 19 Findings from this investigation support the practice of focusing attention on passengers seated near potentially infectious travellers, rather than following up all passengers on a flight. This episode also suggests that greater effort could be applied to exit screening to reduce the probability that travellers with symptoms board aircraft. In this instance, 10 members of the high school group had symptoms when they embarked at a US airport. Modelling of airport screening for pandemic influenza also supports the greater effectiveness of screening at the point of embarkation.20

This investigation does not suggest that the airline cabin environment is a high risk setting for transmission of influenza. The long-haul flight included nine infectious travellers, and it could be argued that the number of secondary cases was relatively small in comparison. This finding may reflect evidence that pandemic A/H1N1 virus has relatively low transmissibility.21 The number of in-flight infections was also somewhat lower than would be predicted using estimates from a recent quantitative microbial risk assessment of within flight transmission of influenza A (H1N1).22 However, as noted, our investigation methods would have tended to underestimate the risk.

Findings from this investigation also provide some information about the potential effectiveness of screening arriving passengers for symptoms of influenza. Although based on small numbers, one symptom (cough) seemed to be relatively sensitive for detecting cases subsequently found to be infected with pandemic A/H1N1. Combinations of symptoms (such as the NZ or US definitions of influenza-like illness) reduced the sensitivity of screening without greatly improving its positive predictive value. These findings are based on small numbers of passengers screened during the early phase of the A/H1N1 pandemic. They are also based on a single atypical flight that contained an unusually large number of cases of influenza. Larger studies are needed to assess the generalisability of these results to other influenza viruses, populations, seasons, and settings.

Similarly, this investigation provides some insights into the effectiveness of follow-up of passengers after arrival. As part of an extensive public health response, follow-up of these arriving passengers was relatively complete for residents of and visitors to NZ but less so for those transiting to other countries (mainly Australia). Nevertheless, this follow-up was not particularly timely; only 52% were followed up within 72 hours. Because this event came at the start of the pandemic, systems were not in place to support public health follow-up. Subsequently, all passengers arriving in NZ airports were required to complete a detailed locator card.

Future investigations of airline transmission of influenza could be improved in several ways. In particular, collecting suitable laboratory specimens (such as convalescent serological samples) from all passengers in the same section of the aircraft—not just those with symptoms—would be useful to obtain a more valid estimate of the risk of transmission of influenza in these settings. Trying to put such interventions in perspective by assessing their overall effectiveness at delaying the introduction of pandemic influenza into more isolated countries, such as New Zealand, would also be useful, as would estimating the resources needed for such prevention and control measures.

This investigation suggests the existence of a low but measurable risk of transmission of pandemic influenza during modern commercial air travel. This risk is concentrated close to infected passengers with symptoms. Screening of arriving passengers for symptoms may need to focus on the presence of single symptoms (such as cough) to achieve a moderate degree of sensitivity. Follow-up of passengers can be difficult once they leave the airport, even when public health authorities mount a vigorous response. For island countries and those with limited entry points, such border control measures may have a role in delaying introduction and spread within the community during the early containment phase of a pandemic. However, the effort applied to such strategies needs to take into consideration the seriousness of the pandemic, the effectiveness of border control measures, and the resources required to operate them.