Search strategy

Using concepts related to virus, transmission, and HVAC, a research librarian (GMT) searched three electronic databases (Ovid MEDLINE, Compendex, Web of Science Core) from inception to June 2020 (see Appendix A in S1 File for the Ovid MEDLINE search strategy). Prior to implementing the searches, two librarians peer-reviewed the strategies (TL, AH). An updated search was conducted in January 2021. Reference lists of all relevant papers and review articles were screened. Using Compendex and Web of Science, conference abstracts were identified and were not included, but literature was searched to identify if any relevant abstracts had been published as complete papers. Limits for language or year of publication were not placed on the search. However, only English-language studies were included due to the volume of available literature and resource constraints. References were managed in EndNote and duplicate records removed prior to screening.

Study selection

Study selection occurred in two stages: title/abstract screening and full-text screening. In the first stage, two reviewers independently screened the titles and abstracts of all references identified by the searches of the electronic databases. Relevance of each record was classified as No, Yes, or Maybe. Conflicts between No and Yes/Maybe were resolved by one of the review team. Pilot testing was conducted with three sets of studies (n = 199 each) to develop consistency among the review team. The review team met to discuss discrepancies and develop decision rules after each set of pilot screenings. In the second stage, two reviewers independently reviewed the full-text articles and applied the inclusion/exclusion criteria. Reviewers classified studies as Exclude or Include. Conflicts between Exclude or Include were resolved through consensus by the review team. One reviewer resolved conflicts when different exclusion reasons were given. Pilot tests with three sets of studies (n = 30 each) were used for the second stage of screening. The review team met to resolve discrepancies after each pilot round. Covidence software was used to conduct screening.

Inclusion and exclusion criteria

Exclusion and inclusion criteria are listed in Appendix B in S1 File. This systematic review was part of a larger effort to examine virus transmission and different HVAC design features. While all four design features were included in the search and screening process, only studies evaluating humidity were synthesized here. In addition, literature examining humidity in combination with ultraviolet radiation was addressed in a separate systematic review on ultraviolet radiation. A variety of agents were included in the search with priority placed on studies of viruses or agents that simulated viruses. Other agents (e.g., fungi, bacteria) would be included if studies were not available specific to viruses. Studies using bacteriophages, which are viruses that infect bacterial cells [26], were included. For this specific review, the synthesis was further narrowed from viruses to coronaviruses and influenza viruses. Studies of the indoor built environment (e.g., office, public, residential buildings) which had mechanical ventilation were of particular interest. Primary research providing quantitative results of the association between humidity and virus transmission was included. Only English-language, peer-reviewed publications were included.

Risk of bias assessment

For experimental studies, the risk of bias was determined based on three key domains: selection bias, information bias and confounding [27, 28]. Reviewers assessed domains as high, low, or unclear risk of bias using signalling questions [23] for the different study types that were included (e.g., animal studies, laboratory experiments, epidemiological studies) from guidance documents [27–30]. Modelling studies were assessed using the following three key domains: definition, assumption, and validation [30, 31]. Definition considered model complexity and data sources, assumption considered the explanation and description of model assumptions, and validation considered model validation and sensitivity analysis [31]. Reviewers assessed each domain as high, low, or unclear risk of bias based on signalling questions [30–32]. Pilot tests were conducted among three review authors for risk of bias items, then two reviewers (DD, EK, or NF) applied the criteria to each relevant study independently and met to resolve discrepancies.

Data extraction

General information about the study (authors, year of publication, country of corresponding author, year of publication, study design) and methods (setting, population [as applicable], intervention set-up, agent studied) was extracted. Details on humidity treatment parameters (where available) were extracted, including relative humidity (RH), absolute humidity (AH), medium, exposure time, and temperature, where applicable. The studies were grouped as aerosolized virus, modelling, animal, and field studies. Quantitative data were extracted, in addition to the results of any tests of statistical significance related to humidity. The primary outcome of interest was quantitative measures of the association between virus transmission and humidity. As such, data on actual transmission were extracted where available (i.e., infections), as well as information regarding virus survival, persistence, infectivity, viral load per hour, concentration, recovery, decay rate, death rate, and virus detection in air. In the animal and aerosolized virus tables, humidity was categorized as low (<40%RH), mid (40–60%RH), and high (>60%RH) RH. Symbols were used to denote high virus survival (+), low virus survival (-), mid virus survival (/) (i.e., between low and high), and no effect (*). The abbreviation ND was used when virus was not detected, and NR was used when the virus viability in a particular humidity category was not reported. One reviewer extracted data and a second reviewer verified data for accuracy and completeness using a data extraction form spreadsheet to ensure consistency. The review team discussed discrepancies.

Data synthesis

Meta-analysis was not possible due to heterogeneity across studies in terms of study design, humidity levels tested, outcomes assessed, and results reported. Evidence tables were developed to describe the studies and their results. A narrative synthesis of results was conducted by study grouping (aerosolized virus, modelling, animal, and field studies).

Aerosolized viruses

Coronaviruses.

Five experimental studies examined coronaviruses using SARS-CoV-2 (BetaCoV/USA/WA1/2020) [35], SARS-CoV-2 England-2 [36], MERS-CoV isolate HCoV-EMC/2012 [37, 38], and hCoV-229E [39] (Table 1). These studies were conducted by aerosolizing the virus into a rotating drum [35, 36, 38, 39] or environmental chamber [37].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Aerosolized virus.

https://doi.org/10.1371/journal.pone.0275654.t001

Two studies that examined SARS-CoV-2 showed different results. Smither et al. [36] found that increased humidity from mid to high RH was associated with increased survival in both artificial saliva (AS) and tissue culture medium (TCM), although survival in TCM was less at higher RH than in AS at the same RH. Schuit et al. [35] found that humidity alone did not significantly affect virus survival. Discrepancy in results could be due to differences in the studies’ experimental set-up and test procedures, e.g., exposure time up to 60 [35] vs 90 [36] minutes.

Two studies analyzed the effect humidity on MERS viruses, including MERS-CoV [37] and MERS-CoV isolate HCoV-EMC/2012 [38]. Van Doremalen et al. [37] found increased humidity from mid to high RH was associated with decreased virus survival (i.e., the highest survival was at mid RH), noting a significant effect of humidity on virus survival. Van Doremalen et al. [37] did not test at low RH. Pyankov et al. [38] found that increasing RH was associated with increasing virus survival when coupled with decreasing temperature; these results were statistically significant during the 30- and 60-minute exposure times. Mid RH levels were not tested.

Ijaz et al. [39] examined a full spectrum of RH ranges and found that increasing humidity from low to mid RH was associated with increased virus survival for hCoV-229E, with the highest survival for coronavirus at mid RH. As well, increased humidity from mid to high RH was associated with decreased hCoV-229E survival.

Aerosolized coronaviruses were not consistent as to minimum and maximum survival versus humidity. Two studies found that high humidity was associated with minimum virus survival for MERS-CoV [37] and hCoV-229E at 20±1°C [39] (70%RH and 80±5%RH, respectively). Two studies found low humidity was associated with minimum survival for MERS-CoV [38] and hCoV-229E at 6±1°C [39] (24%RH/38°C and 30±5%RH, respectively). One study found that mid RH was associated with minimum virus survival [36]. Two studies found that mid RH was associated with maximum virus survival (50±5%RH and 40%RH) [37, 39] and two studies found that high RH was associated with maximum virus survival (79%RH/25°C and 68–88%RH) [36, 38]. Schuit et al. [35] did not find a significant effect of humidity so minimum and maximum survival could not be determined.

Modelling studies

Coronaviruses.

One modelling study [42] examined the effect of humidity on SARS-CoV-2 in terms of viral survival load per hour to determine optimal temperature/RH pairs for virus inactivation (Table 2). Spena et al. [42] used experimental data from Pyankov et al. [38] and Van Doremalen et al. [37] for MERS-CoV, MERS isolate HCov-EMC, SARS-CoV-1, and SARS-CoV-2 in the development of the model. Spena et al. [42] noted that specific enthalpy is a better predictor of ideal virus control than humidity; their study indicates a target value of 55kJ/kg is optimal. Unfortunately, this target results in high absolute humidity values well above typical set points for mechanical systems. To achieve 55kJ/kg specific enthalpy, HVAC settings require an indoor RH of approximately 93% at 20°C, decreasing almost linearly to 60%RH at 25°C (Fig 5 in Spena et al. [42]). Spena et al. [42] indicate a triangular subsector on the psychrometric chart within the ASHRAE recommended quadrangular comfort zone which is both optimal for virus control and comfort. Their findings indicate an important trade-off exists between controlling virus activity and typical building indoor air design parameters.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Modelling studies.

https://doi.org/10.1371/journal.pone.0275654.t002

Animal studies

Influenza.

Six animal studies examining the effect of humidity on viruses used influenza (Table 3) [14, 15, 47–50]. Three studies came from the same research group [14, 15, 50]. Strains used were Influenza A (PR8) [47, 48], Influenza A/Panama/2007/99 (Pan/99; H3N2) [14], Influenza A/Panama/2007/1999 (H3N2) (Pan/99) [15, 49, 50], Influenza A/Netherlands/602/2009 (H1N1) (NL/09) [15, 50], and Influenza A/Indiana/8/2011 (H3N2v) [49].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Experimental animal studies.

https://doi.org/10.1371/journal.pone.0275654.t003

Five studies found that increased humidity from low to mid RH was associated with decreased virus persistence [47], infectivity [48], and transmission [14, 15, 49].

Two studies found that increased humidity from mid to high RH was associated with decreased virus persistence at 5°C, 20°C, and 27–29°C [14, 47]. However, three other studies found that increased humidity from mid to high RH was associated with increased infectivity at 22.2–23.8°C [48] and transmission at 20°C (65%RH) and 23°C [14, 49]. Steel et al. [50] found that increased humidity from low to high RH (mid RH not tested) was associated with decreased transmission at 20°C and 30°C.

Like that of aerosolized influenza studies, six studies found that low RH was associated with maximum virus survival (23–43%RH [47], 23%RH [48], 20%RH [14], 20%RH [15], 20%RH [50], and 30%RH [49], respectively). Three studies found that high RH (89%RH [47], 80%RH [14], and 80%RH [50]) was associated with minimum virus survival and two studies found mid RH was associated with minimum virus survival (45–60%RH [48] and 50%RH [49], respectively).

Field studies

Influenza.

One study found no significant effect of absolute and relative humidity on Influenza A and B detection in different settings on a university campus in Hong Kong (Table 4) [51].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Field studies.

https://doi.org/10.1371/journal.pone.0275654.t004

Table 5 shows a visual representation of the relative change (↑ increase, ↓ decrease,—no change) in virus infectivity between low (<40%), mid (40%-60%), and high (>60%) RH.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Virus infectivity relative change between low RH (<40%RH), mid RH (40%-60%RH), high RH (>60%RH).

https://doi.org/10.1371/journal.pone.0275654.t005

Risk of bias

All animal and field experimental studies had low risk of bias for the three domains: selection bias, information bias and confounding. Seven of the aerosolized virus experimental studies had low risk of bias for all three domains. For the remaining aerosolized virus experimental studies, one had unclear information bias due to lack of clarity regarding exposure time [18] and one had unclear information bias and high selection bias because the test and tracer material were not identical [33, 34]. One was assessed with high risk of bias due to confounding for our comparison of interest because both humidity and temperature were changed, where 79%RH and 25°C was compared with 24%RH and 38°C [38]. The six modelling studies had low risk of bias for all three domains: definition, assumption, and validation.

Implications for research

Based on the included studies, several key implications for research were found, such as the influence of medium, temperature, and exposure time; the need for statistical analysis to better understand and interpret results; and the need for standardized testing procedures. For example, for SARS-CoV-2, increased humidity from low to high RH was associated with an increase or no change in infectivity, where the difference was attributed to differences in exposure time and/or suspending medium. Interestingly, increased humidity from low to high RH was associated with an increase in infectivity for MERS-CoV, where one study decreased the temperature while increasing the RH. While this is one example, the complexity of other factors can be seen through the review content.

While one study [36] directly compared mediums, it was not the only study to comment on the perceived influence of medium on the results. Smither et al. [36] compared Artificial Saliva (AS) and tissue culture medium (TCM), finding that there were fewer particles observed in AS than in TCM, perhaps contributing to the amount of viable virus present. While the difference in medium has implications for research, practitioners may want to consider the results from AS as they are more applicable to real-world transmission or virus survival scenarios. Additionally, using human bronchial epithelial extracellular material (HBE ECM), Kormuth et al. [41] found no significant effect of humidity on viruses. Kormuth et al. suggested that a lack of results could be due to “protection conferred by supplementation of the viruses with HBE ECM” [41 (p744)]. This theory was further tested using Phi6 both with and without HBE ECM. Additionally, Kormuth et al. questioned “how well the media composition represented that of actual aerosols and droplets expelled by an infected host” [41 (p744)]. As such, researchers should be aware of the influence of medium when testing humidity while also considering their research goals (e.g., choosing a medium such as AS to simulate real-world scenarios).

Five studies used multiple temperatures when examining the influence of humidity on viruses [14, 15, 33, 34, 39, 50]. The relationship between temperature and humidity is complex [42]. For example, Ijaz et al. found that “the fluidity of the lipid-containing envelope is stabilized at low temperature, thus protecting the virion; however, further studies are needed to explain these phenomena” [39 (p2747)]. As well, the Lowen group [14, 15, 50] found that humidity and temperature as a combined approach could have an impact on virus survival, suggesting that “influenza virus transmission indoors could potentially be curtailed by simply maintaining room air at warm temperatures (20°C) and either intermediate (50%) or high (80%) RHs” [14 (p1475)]. Additionally, Harper [33, 34] found that high temperatures were associated with the lowest survival at all levels of humidity, finding the influence of humidity “negligible” [34 (p68)]. As a result, researchers should consider the influence of temperature in addition to the influence of humidity.

Exposure time was also found to be a factor when testing the effect of humidity [42]. Schuit et al. [35] questioned whether insufficient exposure time explained the inability to detect a RH relationship for SARS-CoV-2 that was similar to hCoV-229E. Smither et al. [36] found that, in conjunction with the effect of medium, increased exposure time resulted in two patterns of virus survival. For SARS-CoV-2 in AS, increased exposure time was associated with increased differences between virus survival between mid and high RH. In TCM, increased exposure time was associated with decreased differences between survival at mid and high RH. Similarly, Harper [33, 34] found that aerosols were able to remain viable for a considerable amount of time “in favourable conditions” [33 (p485)]. However, Harper [33] also noted that favourable conditions vary by virus.

Other research implications arising from this review are the need for reporting of statistical analyses and standard procedures for testing. Only four of the 18 experimental studies included statistical analysis in their presentation of results [16, 35, 37, 41]. Two of the four studies found significant results [16, 37]; while two studies found nonsignificant results which the authors attributed to study design issues [35, 41]. A limitation of this review is that studies with statistical analyses were presented alongside studies that did not conduct or report statistical analyses in order to allow for a full understanding of the available humidity/virus literature. While attempts were made to ensure clarity when reporting results, differences between survival were reported without supporting statistical analysis from the original documentation and as such, could potentially influence overall findings. Additionally, these findings are further complicated by inconsistencies in testing procedures across the included studies. For example, Schuit et al. [35] attributed their nonsignificant results to short exposure times and Kormuth et al. [41] suggested that medium choice may have affected study outcomes. Additionally, as Derby et al. [12] previously noted, not all of the studies tested a full spectrum of RH levels. Two experimental studies [36, 37] did not test low RH (<40%RH) [36, 37] and two experimental studies [38, 50] did not test mid RH (40–60%RH). As a result, it can be difficult to make comparisons even among similar viruses.

Implications for practice

In their January 2021 Core Recommendations for Reducing Airborne Infectious Aerosol Exposure ASHRAE recommended “maintaining temperatures and humidity at set points” [11]. Spena et al. [42] considered the ASHRAE comfort zone for domestic and office-like environments in their model of SARS-CoV-2. They found that SARS-CoV-2 infectivity would be effectively suppressed for only a portion of the temperature and RH in the ASHRAE comfort zone. This indicated target zone is of high humidity and would be challenging in buildings where mould/mildew control is important, or in older buildings in cold climates where condensation can be a problem. Cold regions would require very significant humidification efforts when outdoor make-up air has very low absolute humidity at inlet. As this modelling study by Spena et al. [42] was released early in the SARS-CoV-2 pandemic timeline, the model includes data for Phi6, HCoV-EMC (MERS), SARS-CoV, and one SARS-CoV-2 study to provide inputs for SARS-CoV-2 modelling. While the enthalpy was not calculated within the scope of this review, the results of Schuit et al. [35] and Smither et al. [36] potentially do not support the findings of Spena et al. [42]. Schuit et al. [35] did not find a significant effect of humidity, possibly due to short exposure times. However, Smither et al. [36] found that high RH was associated with maximum survival at 19–22°C, whereas the modelling by Spena et al. [42] suggests that mid to high RH is associated with increased inactivation or decreased survival. As such, it would be interesting to see how the results of Spena et al. [42] change with new research.

Strengths and limitations

Comprehensiveness and use of methods to avoid bias, including an a priori protocol, pre-specification of inclusion/exclusion criteria and involvement of at least two reviewers at all stages, are the strengths of this study. The limitations of this study are the inconsistencies across the included studies regarding statistical analysis and standardized testing procedures. To gain a full understanding of the available literature, included studies with statistical analysis were presented alongside included studies without statistical analysis. Comparison of included studies of the same virus was challenging due to a lack of standardized testing procedures regarding exposure time, temperature, and medium.

Future research and practice priorities

Even though increasing relative humidity from mid to high RH was associated with decreased SARS-CoV-2 survival, all coronaviruses do not behave the same. As a result, blanket prescriptive humidity levels for coronavirus mitigation are difficult to ascertain. While influenza survival varied from mid to high RH, increased humidity from low to mid RH was associated with decreased virus survival with maximum survival at low RH. When controlling humidity as an HVAC feature, practitioners should take into account virus type and temperature. Future research should also consider the impact of exposure time, temperature, and medium when designing experiments, while also working towards more standardized testing procedures and statistical analysis.