Quantitative aerobiologic analysis of an influenza human challenge‐transmission trial

Abstract Despite evidence that airborne transmission contributes to influenza epidemics, limited knowledge of the infectiousness of human influenza cases hinders pandemic preparedness. We used airborne viral source strength and indoor CO2 monitoring from the largest human influenza challenge‐transmission trial (EMIT: Evaluating Modes of Influenza Transmission, ClinicalTrials.gov number NCT01710111) to compute an airborne infectious dose generation rate q = 0.11 (95% CI 0.088, 0.12)/h and calculate the quantity of airborne virus per infectious dose σ = 1.4E + 5 RNA copies/quantum (95% CI 9.9E + 4, 1.8E + 5). We then compared these calculated values to available data on influenza airborne infectious dose from several previous studies, and applied the values to dormitory room environments to predict probability of transmission between roommates. Transmission risk from typical, moderately to severely symptomatic influenza cases is dramatically decreased by exposure reduction via increasing indoor air ventilation. The minority of cases who shed the most virus (ie, supershedders) may pose great risk even in well‐ventilated spaces. Our modeling method and estimated infectiousness provide a ground work for (a) epidemiologic studies of transmission in non‐experimental settings and (b) evaluation of the extent to which airborne exposure control strategies could limit transmission risk.


| INTRODUC TI ON
The substantial, global disease burden attributed to seasonal influenza and the threat of pandemic influenza demand improved preparedness. Each year influenza has caused up to 650 000 respiratory deaths globally, and as many as 960 000 hospitalizations and over $11 billion in economic burden in the United States. [1][2][3] Efforts to develop and improve vaccines (related to possible differential immunity stimulated by infection initiation in lung vs upper respiratory mucosa) and other prevention strategies are hindered by deficient understanding about the competing risk contributions of contact, large droplet, and droplet nuclei (ie, aerosol or airborne) transmission modes. Influenza virus has been detected in exhaled breath droplet nuclei, 4 and there is evidence supporting airborne transmission as a driver of epidemics and severe illness. [5][6][7][8][9][10][11][12] Quantifying airborne risk is widely accepted as a necessary step for quelling outbreaks.
Wells postulated an airborne quantum generation rate q as the infectious doses generated by an infected individual during exposure with a susceptible. 13 The Wells-Riley equation (Equation S1) facilitates the computation of q with knowledge of secondary attack rate from exposure between primary cases and susceptibles and indoor ventilation rates. 14 Rudnick  to estimate indoor occupant exposure to exhaled breath (Appendix S1). 15 It is based on the knowledge that (a) droplet nuclei are emitted through the exhaled breath of infectious individuals, and (b) CO 2 contained in exhaled breath is constant at 3.8E + 4 ppm and is the predominant source of CO 2 in buildings, a valid assumption for environments without significant combustion sources. Assuming a wellmixed space, the rebreathed air equation uses measured CO 2 levels to directly estimate exposure to exhaled breath that may be contaminated with a quantifiable level of infectious particles.
Despite the challenge of accurately quantifying exposure between infected and susceptible humans, 16  gives an observed secondary attack rate with airborne viral exposure data (EMIT; ClinicalTrials.gov number NCT01710111). 17 The trial tested the effect of using face shields and stringent hand hygiene on transmission risk; these measures were considered to prevent droplet and contact transmission and hence isolate the airborne mode of transmission. We applied the rebreathed air equation to these data to characterize the relationship between airborne exposure and infection risk. We estimated the airborne quantum generation rate q for influenza H3, and we estimated σ, the RNA copies in fine particle exhaled breath aerosols per quantum. We compared these calculated values to available data on influenza airborne infectious dose: NIOSH exposure room air samples from EMIT, a proof-of-concept human challenge-transmission study, 18 an airliner outbreak, 19 and symptomatic, naturally infected cases. 4 Finally, we applied the values to CO 2 -monitored dormitory room environments to predict transmission probability between roommates.

| EMIT trial design
The EMIT challenge-transmission trial methods are described elsewhere. 17 In brief, seronegative volunteers were randomized to viral "Donor" (N = 52), "Intervention Recipient" (N = 40), and "Control Recipient" (N = 35) groups. Intervention recipients wore face shields and observed strict hand hygiene to eliminate droplet and contact transmission routes, but still enable airborne transmission. Control Recipients wore no face shields and had normal hand hygiene, and were considered to be exposed through airborne, droplet, and contact transmission. Donors were inoculated with 0.5 mL per nostril of a suspension containing 5.5log 10

| Pulmonary ventilation rates assumed homogeneous
Here, we assumed the pulmonary ventilation rates of Donors and Recipients (and study monitors who also spent time in the rooms) were similar. Given that volunteers participated in similar, lightly or non-physical activities, and never experienced severe illness, the main differences in the contribution to exhaled breath in the room would be related to baseline respiratory function, which was not likely to be substantially different between healthy, young adult volunteers.

Practical Implications
• Quantitative aerobiologic analysis of the EMIT transmission trial and comparison with a proof-of-concept study suggests that the airborne mode may have driven transmission in these settings. The calculated airborne infectious dose generation rate and airborne RNA copies per infectious dose, and the presented method support new efforts to estimate, predict, and validate airborne infectious doses and transmission risk given knowledge of indoor air CO 2 or ventilation and estimates of exhaled breath viral shedding rates.

| Quantifying viral shedding in exhaled breath
Viral shedding into exhaled breath aerosols was collected from Donors with a Gesundheit-II, 4,20 into "fine" (≤5 µm and >0.05 µm in diameter) and "coarse" (>5 µm) fractions. Evaluation of nasopharyngeal swabs and exhaled breath aerosols by qRT-PCR was done in duplicate using the protocol from Yan et al. 4 The limit of detection (detection of at least one replicate) was 500 RNA copies/sample; the limit of quantification (detection of all replicates) was 2000 RNA copies/sample.  More information on viral shedding estimates is presented in Appendix S3.

| Linking q with measurable airborne virus
One transmission event was observed in Quarantine 2 yielding an overall secondary attack rate (SAR) of 1/75 (1.33%) from exposure to viral Donors. Nguyen-Van-Tam and colleagues suggested a role for airborne transmission when discussing the findings in context. 17 We assume for this analysis that the trans- We applied the rebreathed air equation to the EMIT transmission trial data to estimate an airborne influenza infectious dose generation rate q to give rise to the observed 1.33% SAR. The relationship between q and the observed rate of RNA copy shedding into fine particle exhaled breath aerosols/h, V is defined by where σ is the number of RNA copies per quantum (ID 63 ) and represents the difference between estimated RNA copy airborne exposure, and the viral RNA quantity that reaches a vulnerable locus in the respiratory tract and evades the host immune system. Substituting for q gives Equation 3, which can be applied to the trial data to evaluate σ.
The development of new equations to evaluate the relationship between q and aerosolized RNA copy exposure is described in Results.
Empirical bootstraps with 10 000 samples were used (base R) to produce 95% confidence intervals for q and σ. Residual standard error for the cumulative viral exposure in each EG was computed by linear regression of inhalation exposure on EG. (1) Concentrations measured at 5-min intervals over the entire course of the 4-d exposure period

| Exposure to exhaled breath from infectious donors
Exposure room CO 2 concentrations varied ( Figure Figure 2B, parameter estimates in Tables S1-S4, and model diagnostic Figure S1).
Coarse aerosols were assumed to not contribute to airborne risk in this model given their higher settling velocity and the contribution to substantially less exhaled breath influenza RNA (Appendix S4

| Computation of q
The transmission trial represents a discrete exposure quantity for each of 75 susceptibles, where one became infected. By summing the probability of infection, P for each Recipient, we computed q for the trial as a whole. A similar approach was used to sum risk of transmission across multiple exposure periods between school children during a measles outbreak. 14  Theoretically raising or lowering CO 2 by 10% altered q values by no more than a factor of 0.85 (Table S6 and Figure S3). Doubling the number of transmission events (ie, from 1 to 2) would have produced

| Computation of σ
We adopted a cumulative viral exposure term for each Recipient

| NIOSH sampler aerosol detection limit
NIOSH bioaerosol samplers were deployed in the EMIT exposure rooms for up to 3 hours sampling at a flow rate of 3.5 L/min. 22 No viral RNA was detected from any of the samples immersed in 1mL UTM.
Given an average human breathing rate P of 8 L/min, the samplers collected air at 44% that of human inhalation. The maximum collected by the sampler would have been 5.5E + 2 RNA copies/mL, which is below the limit of quantification for the assay given the dilution factors from nucleic acid extraction and qRT-PCR protocols. 4 Therefore, the lack of RNA detection in the NIOSH samplers is consistent with the low quantities of RNA estimated in the exposure room air.

| Applying q to influenza challenge-transmission "proof-of-concept" study
A proof-of-concept study prior to EMIT demonstrated the feasibility of human transmission following nasal inoculation of seronegative volunteers. 18

| Estimating q for a population of symptomatic, naturally infected cases
Yan et al 4

| Estimating airborne transmission risk for roommates
The computed σ can be used to estimate probability of airborne transmission in a variety of non-experimental settings using the Wells-Riley or rebreathed air equations (Equations S1 and S2). We used hypothetical scenarios in a "high" and "low" ventilated dormitory, each with one infected and one susceptible roommate, given  of exposure, the follow-on transmission trial was expected to produce a 16% SAR. Thus, the observed SAR of 1.33% was much lower than expected under identical study conditions (P < .001). 17 Nguyen-Van-Tam and colleagues point out that the main difference between these studies was likely the room ventilation. 17 The proof-of-concept study was performed in hotel rooms, with relatively little ventilation compared with the follow-on trial in a controlled environment.

| D ISCUSS I ON
That the SAR did not increase between the proof-of-concept and follow-on trials, yet the magnitude and duration of direct and indirect contact between infectious and susceptible volunteers more than Phylogenetic studies could evaluate the theory that a minority of viral shedders are responsible for the bulk of transmission using a previously described analytical framework. 28 Studies that can refine predictors of high-level aerosol shedding, as done by Yan and colleagues, or can refine predictors of transmission related to indoor environments are of great importance to new research efforts and population level disease prevention efforts.

Studies of influenza transmission in households have reported
SARs of 8% and 21%. 29,30 Analysis of these household trials-which used hand hygiene and facemask interventions to control for transmission mode-reported that airborne influenza could be responsible for about half of influenza transmission events and that interventions to interrupt contact and large droplet modes may not reduce overall risk, but rather shift transmission mode. 12 Thus, the household study SARs due to airborne risk alone may be about 4%-10%, which is very similar to that observed in the dormitory scenario with typical shedders (Figure 3). An analysis of a separate household cohort found that among 52 sample pairs between primary and potential secondary household transmission cases with sequence data of sufficient quality, 47 (90%) were considered phylogenetically supported transmission events. 31  shedder could be substantially mitigated by increased ventilation, but perhaps not in the presence of a supershedder. This work highlights the potential for airborne transmission as a driver of epidemics and underscores the need to better characterize drivers of infectious viral shedding and the effect of built environments and exposure controls and their roles in transmission risk, population surveillance, and epidemic and pandemic prevention and readiness.

ACK N OWLED G EM ENTS
We thank the EMIT Consortium (a complete list of the EMIT Consortium can be found in the Supporting Information). We should be inferred.

CO N FLI C T S O F I NTE R E S T
There are no conflicts to declare.