In principle, influenza viruses can be transmitted by 3 routes: aerosols, large droplets, and direct contact with secretions (or with fomites). These 3 routes are not mutually exclusive and, as noted above, may be difficult to disentangle in natural infections.
For the purpose of deciding on the use of N95 respirators in a pandemic, showing that aerosol transmission occurs at appreciable rates is sufficient. Evidence supporting aerosol transmission, reviewed above, appears compelling. Despite the evidence cited in support of aerosol transmission, many guidelines or review articles nevertheless routinely state that "large droplets transmission is thought to be the main mode of influenza transmission" (or similar statements) without providing supporting evidence from either previously published studies or empirical findings. Despite extensive searches, I have not found a study that proves the notion that large-droplets transmission is predominant and that aerosol transmission is negligible (or nonexistent). Reports on many outbreaks suggest that influenza aerosols are rapidly diluted because long-range infections occur most spectacularly in situations of crowding and poor ventilation (25,26
). However, even if long-range infections do not readily occur when sufficient ventilation exists, this does not rule out the presence at closer range of infectious particles in the micron or submicron range, against which surgical masks would offer little protection (29,30
). Many infection control practitioners have argued that the introduction of large-droplets precautions in institutions has proven sufficient to interrupt influenza outbreaks and therefore that aerosol transmission appears negligible. This evidence is, unfortunately, inconclusive because of several confounding or mitigating factors. First, unless precise laboratory diagnosis is obtained, respiratory syncytial virus outbreaks can be mistaken for influenza outbreaks (9
), which would artificially increase the perceived "effectiveness" of large-droplets precautions against influenza. Second, serologic studies are often not conducted, and therefore asymptomatic infections are not documented (among healthcare workers a large fraction of influenza infections are asymptomatic or mistaken for another disease [31
]). Third, since we are in an interpandemic period and the viruses currently circulating have been drifting from related strains for decades, we all have partial immunity against these viruses, immunity that is further boosted in vaccinated healthcare workers. It has even been argued that after several decades of circulation the current human influenza viruses are undergoing gradual attenuation (32
). Finally, surgical masks (used in large-droplets precautions) do not offer reliable protection against aerosols, but they nevertheless have a partially protective effect, which further confuses the issue (29,30
In contrast, the situation with a pandemic strain of influenza A (H5) would become only too clear because no one would have any degree of immunity against such a virus, vaccines would not be available for months, and these viruses would likely be highly virulent. Even though efficient human-to-human transmission of the A (H5N1) virus has not yet been observed (by any mode), transmission of influenza A (H5N1) by aerosols from geese to quails has been demonstrated in the laboratory (33
). Thus, even in the current incarnation of A (H5N1), infection by the virus can generate aerosols that are infectious for highly susceptible hosts. As far as we know, 1 of the main blocks to efficient human-to-human transmission of influenza A (H5N1) is the virus's current preference for specific sialic acid receptors. The current strains still prefer α-2,3–linked sialic acids, which is typical of avian influenza viruses, whereas human influenza viruses bind preferentially to α-2,6–linked sialic acids (34–36
). In all likelihood, 1 of the mutations required for influenza A (H5N1) to give rise to a pandemic strain would be to change its receptor affinity to favor the α-2,6–linked sialic acids. For the influenza A (H1N1) pandemic strain of 1918, this change required only 1 or 2 amino acid substitutions (36
). Once a highly transmissible strain of influenza A (H5) has arisen, it will likely spread in part by aerosols, like other human influenza viruses.
Recent studies have shown that whereas epithelial cells in the human respiratory tract express predominantly the α-2,6 sialic acid receptor, cells expressing the α-2,3 receptor were detected only occasionally in the upper respiratory tract; however, measurable expression of α-2,3–linked sialic acid receptors was found in some cells in the alveolar epithelium and at the junction of alveolus and terminal bronchiole (35
). Binding of influenza A (H5N1) virus can be demonstrated in human tissue sections from the respiratory tract in a distribution corresponding to that of the α-2,3 receptors in the respiratory tract (34,35
). This pattern of virus binding correlates well with autopsy findings, which show extensive alveolar damage (34,37
), and also correlates well with the observation that recovery of the A (H5N1) virus is much more difficult from nasal swabs than from throat swabs (37
). Thus, in the respiratory system the current strains of A (H5N1) appear to infect mostly (perhaps exclusively) the lower respiratory tract. If that is indeed the case, it in turn suggests that human cases of avian influenza were acquired by exposure to an aerosol, since large droplets would not have delivered the virus to the lower respiratory tract. (Another hypothesis might be gastrointestinal infection, followed by viremia and dissemination, but not all patients have gastrointestinal symptoms [37
]). Given the strong evidence for aerosol transmission of influenza viruses in general, and the high lethality of the current strains of avian influenza A (H5N1) (37
), recommending the use of N95 respirators, not surgical masks, as part of the protective equipment seems rational.
Several infection control guidelines for influenza have recently been published, some specifically aimed at the current strains of A (H5N1), others as part of more comprehensive pandemic plans that address the emergence not only of a pandemic form of A (H5) but also of other types of pandemic influenza viruses. Even though to date human-to-human transmission of A (H5N1) remains very inefficient, the high lethality of the infection and potential for mutations call for prudence. The use of N95 respirators is included in the 2004 recommendations of the Centers for Disease Control and Prevention for healthcare workers who treat patients with known or suspected avian influenza (38
). The World Health Organization's current (April 2006) guidelines for avian influenza recommend the use of airborne precautions when possible, including the use of N95 respirators when entering patients' rooms (39
Currently, several pandemic plans differ considerably in their recommendations for infection control precautions and PPE. The current version of the Canadian pandemic plan recommends surgical masks only, disregarding data that support the aerosol transmission of influenza (4
). The US pandemic plans (5
) and the British plans, from both the National Health Service (available from http://www.dh.gov.uk/PublicationsAndStatistics/Publications/PublicationsPolicyAndGuidance/