Systematic review and meta-analysis of methodological approaches for characterising airborne SARS-CoV-2 RNA for environmental surveillance

Abstract

SARS-CoV-2 has transformed our understanding of respiratory disease transmission, particularly through aerosols. This systematic review examines studies published up to October 2, 2024 and focused on filter-based sampling methods reviewing key sampling and molecular diagnostic parameters to characterise airborne SARS-CoV-2 RNA from filters. A total of 84 studies were reviewed, yielding 104 datasets from different environments, including 70 studies indoors, 8 outdoors and 6 sampling both indoors and outdoors. The findings suggest that sampling volume, type of filter and storage conditions after sampling affect the detection positivity rate of SARS-CoV-2 genetic material in aerosols sampled near infected individuals indoors. No sampling or analytical parameters were identified to be associated with detection or quantification. Further experiments are recommended to ascertain the optimal methodological parameters for characterising SARS-COV-2 RNA in aerosols for environmental surveillance, including early warning systems, non-intrusive environmental monitoring, managing COVID-19 outbreaks and characterising viral airborne transmission.

Introduction

SARS-CoV-2, responsible of the COVID-19 pandemic, has changed the conception of the routes of transmission of respiratory diseases¹. Among the different ways virus can be transmitted, the airborne route has proven to be particularly important, especially through aerosols—tiny particles that can carry the virus and stay suspended in the air for long periods, especially in enclosed or poorly ventilated spaces^2,3.

The emergence of SARS-CoV-2 greatly accelerated the use of molecular biology techniques for detecting and quantifying viral genetic material in aerosols⁴. Nonetheless, despite the rapid accumulation of data, standardized protocols for airborne virus sampling and characterization have not yet been established^4,5,6, neither the optimal parameters leading to detection and quantification of the virus.

A comprehensive review by Pan et al.⁷ –pre-COVID-19 pandemic - evaluated the aerosol sampling efficiency and virus viability for various sampling techniques, including filters, impactors, cyclones, liquid impactors, electrostatic precipitators, and water-based condensers. They concluded that no single method could efficiently sample aerosols across a broad size range (10 nm to >10 μm) while preserving virus viability⁷. Whilst selecting a sampling method that preserves viability is crucial to evaluate the infectivity of the virus⁸ it is not always a requirement. For some environmental surveillance applications, there is no need to collect viable virus, but to ascertain the presence or abundance of the virus in the environment. These applications include early warning systems, detection of the virus in environments with limited clinical surveillance, monitoring viral circulation in the environment⁹ or identification of virus free spaces to help outbreak management¹⁰.

Despite the different available options for aerosol sampling, filter based methods have been primarily used to detect and quantify SARS-CoV-2 RNA concentrations in the literature^4,11. These include through direct filtration, where particles are captured by interception by forcing the airflow through a porous medium (filter)^12,13. It also includes the use of filters to capture particles via other physical mechanisms such as inertial impaction^14,15, diffusion¹⁵, and electrostatic attraction¹³. Pan et al.⁷ identified that filtration is effective for collecting aerosols ranging from 20 nm to 10 μm or larger, making it one of the most promising methods for virus detection and quantification in air. Other samplers, such as impactors and cyclones, have lower collection efficiency⁷, but some incorporate filters in one of their stages, such as the multistage cyclone developed by the National Institute for Occupational Safety and Health (NIOSH)^7,16 or the Sioutas cascade impactor^7,17.

Data reliability for detecting and quantifying SARS-CoV-2 genetic material in aerosols presents several technical challenges, including variability in aerosol sampling efficiency, preservation of sample integrity during collection and storage, efficiency of extraction of genetic material from filters, and the sensitivity of molecular methods to detect small amounts of genetic material⁵. These challenges underscore the need for standardized protocols that allow for data integration across studies. Given the complexity on the sample collection, preservation and analysis, key parameters leading to robustness of the methods should be identified to ensure the utility and comparability of collected data.

To the best of our knowledge, eight reviews have been published addressing the sampling and analytical parameters of SARS-CoV-2 in air or aerosols. Birgand et al.¹⁸ conducted a systematic review of 24 studies on air contamination in hospital environments published until October 27, 2020, focusing on RNA detection and infectivity¹⁸, although the sampling parameters and molecular analysis were not very detailed. Rahmani and colleagues (2020) presented a mini-review covering air detection methods for coronaviruses, based on 11 studies (9 using filter-based approaches), highlighting the need for further research to assess the effectiveness of these methods for detecting SARS-CoV-2 in the air¹⁹. However, they did not report the molecular analysis methods, nor data on the quantification and viability of the coronaviruses. Robotto et al.²⁰ reported a narrative review describing the methodological challenges and debates surrounding SARS-CoV-2 air sampling²⁰, although it was a preliminary approximation. Borges et al.²¹ reviewed air sampling methods for detecting SARS-CoV-2 in indoor environments, drawing on data from 25 studies (15 using filter-based approaches) published until October, 2020²¹, but they did not report the molecular analysis methods. Bhardwaj and colleagues (2021) evaluated the methods to characterise pathogenic airborne viruses, including 5 studies focused on SARS-CoV-2 (all using filter-based approaches)¹¹. Cherrie and colleagues (2021) evaluated the presence of SARS-CoV-2 RNA in air and assessed the quality of the methods used in 35 studies (24 using filter-based approaches) published until December 24, 2020²², although the sampling parameters and molecular analysis were not very detailed, nor did they mention viral viability. Dinoi et al.⁶ published a review comparing indoor and outdoor environments for detecting and quantifying SARS-CoV-2 RNA in airborne samples, based on 78 articles (53 using filter-based approaches) published until August 31, 2021⁶, although the sampling parameters and molecular analysis were not very detailed. Silva and colleagues (2022) compiled aerosol sampling methods for detecting SARS-CoV-2 in air, based on 76 studies (48 using filter-based approaches) conducted in indoor and outdoor environments published until December 20, 2021⁴, but they did not report the methodology for molecular analysis or the genetic material quantification. Finally, Dias et al.²³ described methods for evaluating exposure to airborne viruses in indoor environments, including 21 studies (9 using filter-based approaches) on SARS-CoV-2 published until June 30, 2023²³, but results of detection, quantification, and viability of the studies were not reported.

Therefore, to date, there is no comprehensive review assessing the sampling factors leading to optimal collection of SARS-CoV-2 genetic material in aerosols using filter-based approaches. Neither exists a review focused on the molecular analysis, along with reporting results on detection, quantification, and viability, with a temporal scope extending beyond mid-2023.

This systematic review aims to examine how sampling and analytical parameters are related to detection, positivity rate and quantification of SARS-CoV-2 genetic material in aerosols collected using filter-based approaches. The review will include studies focused on both indoor and outdoor environments published from the onset of COVID-19 to October 2, 2024.

Results

A total of 499 articles (263 from Web of Science and 236 from PubMed) with potential interest were identified in the initial search. Figure 1 shows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses PRISMA)²⁴ flow diagram followed to identify and select the reviewed studies. These studies were combined into a single database, and duplicates were removed (n = 155). After removal of duplicates, an initial screening of titles and abstracts was performed on 344 articles, resulting in the exclusion of 278. Additionally, 88 articles were identified using the snowballing method. In the second phase, the full text of 154 articles was reviewed, with 70 excluded. Ultimately, 84 articles were included in this systematic review, six of which provided data on both indoor and outdoor sampling.

Fig. 1

PRISMA flow diagram for the identification, screening, and evaluation of records included in this systematic review, following PRISMA guidelines²⁴.

The main characteristics of the studies included in this systematic review are summarized in Table 1, while Table 2 provides details on sampling parameters, analytical methods, and key results. The 84 studies included span 27 countries, with the majority conducted in Europe (36%), Asia (34%), and North America (24%) (Supplementary Fig. 1). USA, Italy, China, Spain and Iran are the top countries where these studies were conducted, whilst no studies were identified from Africa or Oceania. Of the 84 studies included in this systematic review, 30% were published in 2020, 34% in 2021, 26% in 2022, 8% in 2023, and 2% in 2024, showing that most of the publications are from the first two years, with a stark production decrease during the last two years.

Table 1 Summary of studies with detection and quantification results based on sampling environments

Table 2 Summary of sampling parameters, molecular analysis methods, and results on the detection, quantification, and viability of SARS-CoV-2 RNA in air samples

Most of the studies (61/84 studies, 73%) collected samples in healthcare settings, primarily in hospitals, especially in areas dedicated to COVID-19 patients such as intensive care units (ICU), COVID-19 wards, isolation rooms, emergency rooms, patient rooms, medical units, air infection isolation wards, corridors, cafeterias, and medical staff workstations. Six studies (7%) collected samples in sociosanitary settings, including long-term care facilities and elderly care homes. In eight studies (10%) samples were collected in transportation settings. Fifteen studies (18%) included aerosol samples from other indoor environments, including homes with COVID-19 patients (8 studies, 10%). Three studies (4%) collected samples in educational settings. Other minor locations included samples collected from shopping centres, food markets, offices, pharmacies, banks, hair salons, prisons, and a mink farm (Table 1).

Regarding outdoors spaces, samples were taken in 14 studies (17%) (Table 1) across different microenvironments such as urban areas (9), healthcare settings (5), industrial areas (3), public places (2), rural areas (1), and a livestock farm (1). Sampling points were primarily located on sidewalks, bus stations, supermarkets, shopping malls, residential areas, university campuses, and train stations. Additionally, samples were collected in areas near hospitals, such as outdoor terraces, main entrances, and gardens. One study collected samples on a mink farm and its surroundings.

Detection, positivity, quantification and viability

SARS-CoV-2 genetic material was detected in a significant number of studies conducted in indoor environments, with 72% (55/76) the datasets showing positive results in at least one sample^{5,10,17,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76} (Table 3, Supplementary Fig. 2). Detection of SARS-CoV-2 RNA was found in 74% datasets from healthcare^{5,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63}, 83% from sociosanitary^{5,10,63,64,65}, and 63% from transport settings^{10,17,66,67,68}. As regards other indoors, detection was 67% in several datasets collected from homes^{10,59,60,62,63,69,70}, educational settings⁵, commercial spaces⁶⁶ and farms⁷⁶. In outdoor locations, 64% (9/14) of the datasets were positive^{73,74,75,76,77,78,79,80,81} (Table 3; Fig. 2). No statistical differences were found as regards detection rate according to type of environment.

Fig. 2

Number of studies reporting SARS-CoV-2 genetic material detection and non-detection across different sampling environments.

Table 3 Relationship between sampling parameters with the detection and quantification in air of SARS-CoV-2 genetic material in all datasets (N = 104 datasets from 84 studies)

The frequency distribution of positivity rates (Fig. 3 and Table 2), understood as the ratio of filters where SARS-CoV-2 is detected compared to the total amount of filters collected in a setting, was calculated in 74 datasets (both indoor and outdoor) that detected SARS-CoV-2 genetic material in at least one sample in all environments^{5,10,17,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68}. Figure 3 shows the positivity rate based on the number of samples collected in each study, where SARS-CoV-2 genetic material was detected in at least one sample, segmented by sampling environment. Figure 3 shows that 44% of datasets had a positivity rate between 5 and 25% of collected samples^{5,26,32,33,34,35,40,43,44,46,47,48,51,53,56,57,59,62,63,65,67,68,72,73,75,76,79}. A higher positivity rate (25-50%) was observed in 19% datasets^{28,29,31,39,42,49,50,52,60,61,62,70,74}. Likewise, 17% of the datasets reported a positivity rate of 50-75%^{25,41,46,54,59,66,69,73,74,77,80}. Low positivity rate ( < 5%) was found in 12% of the datasets^{5,30,36,37,38,45,58,64,71}, whereas positivity rates > 75% were found in only 8% of datasets, all from indoor environments^10,17,27,55. Datasets collected in healthcare settings showed a large variability of positivity rates (1–100%). In contrast, the positivity rate of outdoor datasets ranged between 6 and 63% (Fig. 3, Table 2).

Fig. 3

Positivity rates, segmented by sampling environment.

According to the meta-analysis, the pooled proportion of SARS-CoV-2-positive samples on air filters (Fig. 4, Supplementary Fig. 3, Supplementary Fig. 4, Supplementary Table 1) was 15% (95% CI: 11–20%), ranging 14% (95% CI: 9–19%) in healthcare settings to 24% (95% CI: 3–53%) in transport settings. Statistically significant heterogeneity was observed across studies (I² = 93.03%; p < 0.001), confirmed by the forest plot (Fig. 4). Individual detection rates ranged from 0% to 100%, reflecting the diversity of sampling settings and methodologies across the reviewed studies.

Fig. 4: Positivity rates based on the number of samples collected in each study, where SARS-CoV-2 genetic material was detected in at least one sample, segmented by sampling environment.

Forest plots of positivity rates and 95% confidence intervals of detection of SARS-CoV-2 in aerosol samples collected in (A) healthcare, (b) sociosanitary settings, (c) other indoors, (d) transport, (e) outdoors.

According to the results of the funnel plot (Supplementary Fig. 5) and confirmed using the Egger test (p-value < 0.001), publication bias could not be ruled out. The sensitivity analysis yielded a combined proportion between 0.15 and 0.16 with a percentage of heterogeneity between 92.52% and 93.10%, thus demonstrating robustness in the results.

The meta-regressions analysis identify that a few variables were statistically significantly associated with the variability in the detected proportions. Specifically, the type of environment (Other indoor vs. Healthcare, p-value = 0.018), sampling volume (p-value < 0.001), the use of filters other than PTFE (excludes gelatine, p-value = 0.033), certain storage conditions, such as storage at 4 °C (p-value < 0.001), and analysis within a few hours of collection (p-value = 0.001), showed statistically significant associations.

However, most of the parameters evaluated—including air flow rate and time, gelatin filter type, and the gene analysed—did not statistically significantly explain the observed variability. I² values remained high in several models (up to 57.93%), indicating that a significant portion of the heterogeneity remains unexplained by the covariates analysed (Table 5). The distribution of viral concentrations in aerosols collected in filters was assessed in 35 datasets (Fig. 5, Table 2). Very low concentrations (< 10 copies/m³) were measured in 18% datasets^{5,46,50,56,57,62,68,72,74,75,79,80}. A third of datasets (35%) had concentrations in the range of 10–100 copies/m³ range^{5,10,46,47,50,56,57,58,61,63,68,72,74,75,79}. Higher concentrations were found in 16% datasets (100–1000 copies/m³)^{5,10,46,50,56,58,63,73,76}, and 16% of datasets (1000–10,000 copies/m³)^{10,17,46,48,49,54,56,58,63,73,76}. Only 9% datasets had concentration in the range of 10,000-100,000 copies/m³^{5,10,46,49,54,56,58,63}, and 6% datasets had a viral load >100,000 copies/m³^{10,46,49,55,58,70} (Fig. 5, Table 2). All sampling environments showed a large variability of viral concentrations. The maximum viral load found in outdoor and transport settings was in the range 1000–10,000 copies/m³. The maximum viral load reported in healthcare, sociosanitary and other indoor settings was considerably higher (> 100,000 copies/m³).

Fig. 5

Distribution of SARS-CoV-2 viral load in air samples from the studies included in this review, segmented by sampling environments.

Only 29 datasets quantified concentrations in indoor environments^{5,10,17,46,47,48,49,50,51,52,53,54,55,56,57,58,61,62,63,68,70,71,72,73,74,75,76,78,79,80,81}, and 8 in outdoor locations^{73,74,75,76,78,79,80,81}. Whist no difference was observed in terms of detection or positivity among indoor or outdoor samples, the median concentration of SARS-CoV-2 RNA was statistically significantly higher (p-value < 0.05) in indoors^{5,10,17,46,47,48,49,50,54,55,56,57,58,61,62,63,68,70,72,73,74,75,76,79,80} than outdoors (Table 3).

Most of the datasets were from studies that conducted part or all of their sampling near diagnosed COVID-19 patients (67/90 datasets, 74% of studies)^{10,17,25,26,27,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,69,70,71,72,73,74,75,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100}, of which 72% detected the presence of SARS-CoV-2 RNA in at least one sample^{10,17,25,26,27,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,69,70,71,72,73,74,75} and 36% quantified the viral load^{10,17,46,48,49,50,51,52,53,54,55,56,57,58,61,62,63,70,71,72,73,74,75}. In the remaining studies where sampling was not conducted near COVID-19 patients (23/90 datasets, 26%)^{5,28,47,66,67,68,76,77,78,79,80,81,101,102,103,104,105}, SARS-CoV-2 genetic material was identified in 16 datasets (70%)^{5,28,47,66,67,68,76,77,78,79,80,81} and quantified in 11 (48%)^{5,10,68,76,78,79,80,81} (Table 3). The results showed a significant association between sampling near COVID-19 patients and quantifying a high viral load of SARS-CoV-2 in air samples (p-value < 0.05) (Table 3).

Many studies conducted in outdoor settings provided concentrations in copies/m³ of air^{73,74,75,76,79,80} (Table 3). Some others, however, reported concentrations in genomic units/Total Suspended Particulates⁷⁸, ng/μL of sample⁷¹ or copies/volume of sample^51,52,53, which preclude to compare their concentrations with other studies. One study indicated that two or three positive targets suggest high viral abundance in aerosol samples, while one positive target indicated low abundance of viral load in the air⁶⁸.

The results of the meta-analysis show that the pooled mean concentration of airborne SARS-CoV-2 genetic material quantified on filters was 48.20 copies/m³ (95% CI: -61.39 to 157.80). Although this represents a mean value, the confidence interval includes negative values that are not physically plausible, reflecting high imprecision in the pooled estimate. This imprecision is due to high variability between studies and the low weight of those with extreme values. Moderate heterogeneity was observed between studies (I² = 60.9%), which was found to be statistically significant (p = 0.009), and was confirmed by the forest plot (Supplementary Fig. 6). The individually reported mean concentrations varied considerably, ranging from 4.8 to 1.50 × 10¹¹ copies/m³, reflecting the wide variability in the settings, sampling techniques, and quantification methods used in the included studies. By removing some outliers, the combined mean concentration and percentage of heterogeneity remained virtually the same (Supplementary Table 2). According to results from the Funnel plots (Supplementary Fig. 7), publication bias could not be ruled out. The sensitivity analysis yielded a combined mean concentration between 45.7 and 4,360 copies/m³ with a percentage of heterogeneity between 42.1% and 64.0%, which showed high variability in the results and a lack of robustness in the overall estimate.

The viability of SARS-CoV-2 collected in filters (Table 2) was evaluated in 16 studies^{29,30,34,37,41,53,54,55,60,63,64,65,76,97,99,100}, of which 6 were able to infect cells with the virus, showing viability of the SARS-CoV-2 after collection on filters from aerosol samples^{55,60,63,65,99,100}. In one study, cytopathic effects were observed in an active sample⁶⁵, while another reported the detection of viable virus after 16 h of sampling⁶³. Additionally, one study successfully established a positive viral culture in one of the three samples analysed⁶⁰, and another found viable virus in cells exposed to aerosol collected from 3 out of 18 samples⁵⁵. Finally, viable virus was also identified in aerosol samples collected using devices that do not rely on filter-based collection^99,100.

Sampling parameters

The most used samplers in the studies were the multistage cyclone developed by the National Institute for Occupational Safety and Health (NIOSH)^{30,32,37,38,48,55,56,65,70,84,85,88,89,99,100} and the MD8 air sampler (Sartorius)^{29,36,40,42,54,60,72,83,87,90,91,92,93,98}. Some studies used a bespoken filter sampling collection system in conjunction with other sampling methods that do not include the use of filters, such as cyclones. No difference was observed according to the use of different sampling instruments as regards frequency of detection, or viral load quantified.

Of the 84 studies included in this systematic review, 3 did not clearly report the airflow during sampling^44,51,79, and another study provided the airflow of the ventilation ducts but did not use sampling equipment⁴¹. Among the datasets, 53% (58/110) employed a flow rate of less than 10 L/min^{10,17,26,27,30,31,32,37,39,43,45,46,48,49,52,53,54,55,56,57,58,63,64,65,66,67,68,69,70,71,72,73,74,76,84,85,88,89,93,97,99,100}, 37% (41/110) used a flow rate between 10 and 100 L/min^{5,25,28,29,31,33,35,36,40,42,47,50,53,54,60,61,62,66,71,72,77,83,86,87,90,91,92,93,95,96,98,101,102,103,105}, and 10% (11/110) used flow rates greater than 100 L/min^{25,34,43,56,59,72,81,82,94,104}.

Among the included studies, sampling time was not clearly reported in three studies^41,51,94. Among the datasets, 26% (24/92) conducted sampling for only a few minutes^{17,26,29,32,34,36,40,42,48,53,54,55,59,60,64,72,83,87,90,91,92,93,96,98}, while a larger portion, 46% (42/92), collected samples over several hours^{10,25,28,30,31,35,37,38,39,45,46,49,50,56,58,61,63,64,65,66,67,68,70,71,72,73,74,76,82,84,85,86,88,89,93,95,97,99,100,101,102,104}. Another group, representing 18% (17/92), performed 24 h sampling^{5,10,27,33,43,47,56,69,71,72,77,78,79,80,81,104,105}, and a limited subset, 10% (9/92), extended their sampling beyond 24 h^{44,52,56,57,62,72,75,76,103}.

Regarding the sampled air volume, 34% (34/101) conducted air sampling with a volume of up to 1 m^{3 17,26,29,30,31,32,36,37,40,45,48,49,53,54,55,56,58,60,64,66,67,70,72,83,87,89,90,91,92,96,97,98,99,100}, while 36% (36/101) collected between 1 and 10 m³ of air^{10,28,31,34,35,38,39,42,43,46,49,50,53,56,59,61,63,64,65,66,68,71,72,73,74,76,78,79,84,85,86,88,93,95,101,102}. Another 22% (22/101) performed air sampling with volumes between 10 and 100 m³ ^{5,25,27,33,47,52,56,57,62,64,69,71,72,74,75,76,77,79,80,82,101,105}, and a smaller subset, 9% (9/101), sampled air volumes exceeding 100 m³ ^{43,56,62,72,79,80,81,103,104}. Four studies did not clearly report the sampling volume^41,44,51,94.

No significant associations were identified between airflow rate, sampling volume, or sampling time and the detection or quantification of SARS-CoV-2 in samples collected in air near individuals infected with COVID-19 (Table 4). However, according to the meta-regression analysis, a larger sampling volume was associated with a higher positivity rate (Table 5).

Table 4 Relationship between sampling parameters and molecular analysis methods with the detection and quantification of SARS-CoV-2 genetic material in aerosols in datasets that collected all their samples near COVID-19 infected people

Table 5 Results of bivariate meta-regression models: association between sampling parameters and molecular analysis methods with the positivity rate of airborne SARS-CoV-2 collected on filters

Different materials were used in the filters for sampling. In 10 studies, two or more types of filters were used for the sampling^{10,37,43,57,66,71,72,75,79,93}. Teflon or polytetrafluoroethylene (PTFE) filters were the most common, used in 33 studies^{10,17,26,28,30,33,35,37,38,43,48,49,58,65,66,67,68,69,70,71,72,73,76,79,80,84,85,86,88,89,93,99,100}. Gelatine filters were used in 27 studies^{29,36,37,40,42,43,50,53,54,55,56,57,60,63,64,72,74,83,87,90,91,92,93,96,97,98,102}, and quartz filters in 12 studies^{5,10,27,47,72,77,80,81,101,103,104,105}. The remaining materials were less common: glass fibre^{52,57,66,75,78,79}, electret filters^34,82,94, cellulose^53,72, HEPA⁴¹, polyurethane foam (PUF)^57,75, membrane^45,95, polycarbonate^37,46,71,79, aluminium foils⁶², polyvinylidene fluoride (PVDF)³¹, polyvinyl chloride (PVC)⁴³, and polyester²⁵. In one study, a surgical face mask was used as a sample filter. In five of the studies, the material of the filters was not specified^{32,39,44,51,61} (Table 2).

No significant association was observed between the type of filter material used in sampling and the results of SARS-CoV-2 RNA detection and quantification in samples collected in air near COVID-19 infected individuals (Table 4). On the other hand, those studies that used filters other than PTFE or gelatine reported a higher positivity rate according to the meta-regression (Table 5).

Nineteen studies reported transporting samples to the laboratory under cold conditions: twelve studies at 4 °C^{26,28,33,35,65,66,67,72,75,76,84,86}, five with ice^{32,34,37,45,57}, and two with dry ice^76,99.

Some studies analysed the samples immediately upon arrival at the laboratory^{45,62,67,74,102}, within 6 h^29,64,86,94, within 24 hours^10,65,96, or within 72 hours^41,50,82. Others pre-treated the samples prior to storage at -80 °C^{17,28,37,66,70,71,84,85,91,92,97}. Other studies directly stored the samples at -80 °C^{25,33,34,38,52,58,76,81,88,90,93,95,98,99,100}, -70°C²⁶, -30°C³⁵, -25°C^101,103, -20°C^{5,43,46,47,78,79,80,105}, or 4 °C^41,63,65.

Twenty-five studies did not mention the conditions of preservation of the sample during transport and storage^{27,30,31,36,39,40,44,48,49,51,53,54,55,56,59,60,61,68,69,73,77,83,87,89,104} (Table 2).

No significant differences were found regarding the storage of samples and the results of SARS-CoV-2 RNA detection and quantification collected in air near individuals infected with COVID-19 (Table 4). However, the results of the meta-regression identified that those studies that stored the samples at 4 °C or those which analysed the samples within hours reported higher positivity rates (Table 5). The factor that increased the highest the positivity rate was storing the samples at 4°C.

Some studies gathered data on temperature^{26,28,33,46,47,49,66,72,77,89,90,100}, relative humidity^{26,28,33,46,47,49,72,77,89,90,100}, concentrations of particulate matter^{33,49,52,53,67}, aerosol concentrations and size distributions⁵⁵, and/or real-time monitoring of particulate matter¹⁰⁴. Other studies also measured CO₂^47,49,89, atmospheric pressure⁷², irradiance⁷⁷ and total volatile organic compounds⁴⁹. Some outdoor studies obtained meteorological or air quality data from the nearest monitoring station^{78,79,80,81,104}. In two studies, gravimetric analysis of the filters was also performed^78,105 (Table 2).

Molecular analysis of SARS-CoV-2 RNA

Incorporating internal controls helps monitor the extraction process and detect potential inhibitors that may affect downstream analyses. This practice is crucial for obtaining reliable and reproducible results in environmental surveillance studies. Despite this, very few studies included an internal control in their molecular analyses to evaluate the efficiency of genetic material extraction (10/84, 12%)^{5,10,28,47,66,73,80,101,103,105}.

Mengovirus was used as the internal control in six studies^{5,10,47,80,101,103}, with recovery rates ranging from 0.8%⁸⁰ to 58%⁵. The Infectious Bronchitis Virus (IBV) was used in two studies, with recovery rates of 20%⁶⁶ and 25%²⁸. One study employed the 77b fragment of the Equine Arteritis Virus achieving a 100% recovery rate⁷³. Finally, one study used synthetic SARS-CoV-2 armoured RNA for extraction efficiency, but no recovery rate was reported¹⁰⁵.

The most frequently targeted regions of the SARS-CoV-2 genome in the studies were E^{5,10,25,29,31,37,39,41,46,47,54,55,56,59,60,63,64,65,68,73,76,77,78,84,87,90,93,101,103}, N^{17,25,26,28,32,33,38,41,46,57,60,63,66,70,74,75,77,88,89,90,91,92,95,101,102,105}, N1^{5,10,30,38,43,44,47,49,52,53,58,69,71,72,78,79,88,104}, RdRp^{26,27,31,33,34,35,59,65,70,77,79,83,87,90,95,101,103}, N2^{5,10,30,38,43,44,47,50,71,72,88,99,100,104} and ORFab^{28,45,60,66,69,74,81,82,84,90,96,102}. Less frequently targeted regions included S^{35,51,62,69,102}, ORFb^34,80,97,105, N3^49,88, ORF1^62,89, IP2⁶⁸, IP4⁶⁸, and UTR³⁷. Additionally, some studies did not clearly specify which gene was targeted in the molecular analysis techniques^{36,40,42,48,61,67,85,86,94,98}.

No significant association was found between the targeted gene and the detection or quantification of SARS-CoV-2 RNA in air near individuals infected with COVID-19 (Table 4). In addition, no clear pattern about which target gene was best suited to detect and quantify the viral load was observed among those studies that used several target genes (Supplementary Table 3). The meta-regression analysis did not show any differences across target genes either as regards positivity rate (Table 5).

In quantitative PCR (qPCR) assays, the cycle threshold (Ct) value—the number of cycles required for the fluorescent signal to cross a predetermined threshold—is commonly used to determine whether a sample is positive. In most studies reviewed, the criterion for determining a positive sample was based on Ct values lower than 40^{5,10,26,27,30,32,37,38,44,46,47,48,62,65,71,72,78,80,88,97,101,105}. Some studies specified slightly different thresholds to consider a sample positive, such as Ct below 35^31,87,104, 36⁶³, 37^69,79,102, 38^{42,49,58,82,96}, 39^56,75, 40.85⁵⁷, 44.25⁷³, or 45³⁴.

Two studies considered a sample positive if there was amplification^84,90. Another study defined strong positive results (low Ct value, ≤32) and weak positive results (high Ct value, >32)⁹³. Finally, one study considered a sample positive when the two replicate qRT-PCRs resulted in a Ct below 40, or when at least one replicate had a Ct below 38⁵³. The rest of studies did not specify the Ct value used to define a positive sample^{17,25,28,29,33,35,36,39,40,41,43,45,50,51,52,54,55,59,60,61,64,66,67,70,74,76,77,81,83,85,86,89,91,92,94,95,98,99,100,103}.

Twenty studies reported the limit of detection(LoD) or the limit of quantification (LoQ)^{5,25,30,34,36,38,50,51,53,54,56,58,71,74,80,88,96,101,103,105}. Some studies reported the limits in copies per sample volume or reaction^{30,36,38,51,53,54,56,74,80,88,96,101,103}, ranging from 0.1 copies/µL⁵⁴ to 1,241 copies/mL⁵⁶. Other studies reported the LoD using different units: 536 (N1), 443 (N2), and 63 (E) copies⁵⁸; 2.18 copies (ORF1ab)⁷⁴; or 0.04 ng/µL⁷¹. Finally, only five studies provided these limits in viral copies per air volume^{5,25,34,50,105}, ranging from 1.04 copies/m³ of air⁵⁰ to 2,200 copies/m^3 34.

Discussion

This systematic literature review compiles studies using filter-based methods for aerosol collection that report detection and quantification of SARS-CoV-2 RNA across various indoor and outdoor environments.

It is noteworthy that, while research on the detection of SARS-CoV-2 RNA in aerosols experienced a marked surge in scientific output during the early stages of the pandemic, it was followed by a progressive decline in subsequent years. This downtrend coincides in time with the resolution of key scientific questions. For instance, airborne transmission of SARS-CoV-2 was acknowledged by the WHO on the 23rd of December 2021^106,107,108. It also coincides with a decrease in research funding for SARS-CoV-2^109,110 from the financial year 2022 onwards.

Nonetheless, the body of evidence generated in this short period deserves to be critically evaluated to identify the sampling and analytical methodological parameters associated with a high detection and positivity rate that could guide in the definition of standardized methods to detect SARS-CoV-2 in aerosols. Likewise, the identified parameters could be useful to conduct further research of airborne pathogens or developing environmental surveillance systems.

This review has identified 84 studies conducted both outdoor and indoors, including healthcare, sociosanitary, transport, residential, and educational settings. Much of the attention of the scientific community centered in detecting and characterising viral load in aerosol samples collected from healthcare settings (Fig. 2). However, the results of this systematic review highlight that, despite receiving less attention, SARS-CoV-2 genetic material was detected also in non-healthcare spaces. Moreover, no statistical differences in detection or positivity rate were observed across all the settings reviewed (p-value > 0.05) (Table 3; Fig. 2; Supplementary Fig. 2).

Sociosanitary centres, such as elderly care homes, represent a high-risk environment due to the vulnerability of their residents and the challenges in implementing strict control measures in these settings¹⁰. Similarly, public transportation and educational spaces could be critical points due to the high density of people and the potential for prolonged exposure in areas with limited ventilation⁵. However, these microenvironments received less attention compared to healthcare spaces despite being potentially critical spaces for viral transmission in public community spaces^4,6.

Most of the studies (74%) focused on collecting aerosol samples in locations with known cases of COVID-19. However, SARS-COV-2 could be present in the air of environments frequented by asymptomatic cases^111,112. Very few studies have focused on characterising viral load in environments where no prior evidence or knowledge of COVID-19-infected individuals existed to evaluate the prevalence of the virus on those locations. Information on the detection rate, positivity rate and concentrations measured in these locations could shed light on the importance of the silent spread of COVID-19 through airborne transmission. This is of relevance since the presence of asymptomatic cases¹¹¹ can represent 0.25% of the total population, and 40% of COVID-19 cases¹¹³. This review found that several studies detected SARS-CoV-2 RNA even in areas where no prior evidence or knowledge of COVID-19-infected individuals existed^{5,28,47,66,67,68,76,77,78,79,80,81}. This reinforces the recommendation to characterise viral airborne levels in public community spaces in future epidemics, with independence of the knowledge of the presence of cases in these environments.

A few studies reporting the detection of SARS-CoV-2 RNA in filters found positivity rates below 5% in healthcare settings^{30,36,37,38,45,58,71} or educational settings⁵, while others reported positivity rates as high as 100% in healthcare⁵⁵, transport¹⁰, and residential settings¹⁰. However, its interpretation should be cautious, especially in studies with small sample sizes. For instance, studies with only 1 or 2 samples^10,61,99 are highly susceptible to variability and may not be representative, potentially yielding inflated positivity rates if even one sample tests positive. This can introduce bias, as results may reflect chance or characteristics specific to a small sample⁶.

The positivity rate - i.e. the percentage of filters where viral load is detected compared to the total number of filters collected - is an important metric for assessing the prevalence of SARS-CoV-2 RNA in the environment surveyed. No differences were observed across different microenvironments according to the meta-regression analysis.

However, differences were observed in airborne viral load. Median viral concentrations in aerosol samples collected indoors were 40 times higher and significantly different (p-value < 0.05) than outdoors (Table 3). This could be primarily attributed to ventilation¹. In outdoor environments, increased air circulation facilitates the rapid dispersion of the virus¹¹⁴. In contrast, indoor environments with limited ventilation lead to aerosol accumulation contributing to higher viral loads in the air¹¹⁵. The finding of higher viral load indoors than outdoors align with previous studies indicating that poor ventilation is a key determinant in the transmission of COVID-19¹. It can also be related to a larger presence of intense sources (i.e. the number of COVID-19 infected individuals) indoors than outdoors, such as in healthcare locations¹¹⁴.

On the other hand, no differences were observed as regards detection and positivity rate in the presence or absence of known cases of COVID-19. This is consistent with results from Birgand et al.¹⁸, who found low positivity rate in samples collected in various hospital sites where patients were present at the time of sampling¹⁸. However, there was a stark difference in the concentrations of SARS-CoV-2 in aerosols, yielding median RNA concentrations 350 times higher across the reviewed studies when samples were collected near COVID-19 patients than when the presence of a case was not known (Table 3). This might be related to the fact that aerosols that yield high concentrations were collected in close distance to a patient (< 2 m) yielding higher RNA concentrations¹¹⁶ than those sampled at larger distances. It might be also associated with the fact that samples were collected in small rooms where patients were isolated^55,70, hence facilitating concentration of the viral load in a small space. It could also be related to the fact that in some studies sampling was collected in the presence of several patients, such as in the ICU⁴⁶, hence increasing the potential emissions sources of viral shedding into the air¹¹⁴.

The information reviewed in the present study provides some insights on the selection of the sampling parameters that enables to capture viral genetic material of SARS-COV-2 present in aerosols. In this regard, the results of detection, positivity rate and quantification have been examined as regards the type of filter used, flowrate, sampling volume, sampling time and preservation conditions during storage of collected samples.

Filters have been identified as the most effective sampling method for capturing particles smaller than 10 μm^7,21, despite presenting some difficulties to maintain viability of the virus due to dehydration during the filtration process⁷.

The global analysis of the datasets where sampling was conducted near individuals infected with COVID-19 reveals that there is no association between the type of filter and the detection or quantification of SARS-CoV-2 (Table 4). However, using filters different from PTFE or gelatine were more effective in detecting viral genetic loads, as suggested by the results from the meta-regression of the positive rates (Table 5). This suggests a preference for using quartz, HEPA or other type of filters different from Teflon or gelatine to characterise SARS-CoV-2 RNA concentrations in aerosols. This is consistent with a previous study that compared the sampling collection efficiency of several filters and found no differences between gelatine and Teflon filters to characterise the influenza virus¹¹⁷.

No significant differences on SARS-CoV-2 detection, positivity rate or quantification according to different sampling flowrates were observed (Table 4, Table 5). This contrast the findings of Raynor et al.¹¹⁸, who observed that low-flow samplers provide more accurate measurements of airborne influenza concentrations than high-flow samplers, although the underlying reasons for this remain unclear¹¹⁸.

No association was observed between the volume and the detection or load of SARS-CoV-2 in the datasets collected near individuals infected with COVID-19 (Table 4). However, the meta-regression analysis indicated that higher volumes increased the positivity rate (Table 5). This is consistent with the results of Dubey et al.³¹, which compared air samples collected with three different volumes—0.09 m³, 1 m³, and 1.6 m³—using the same sampling duration. Their results showed an increase in positivity rate with larger sampling volumes, rising from 28.6% with 0.09 m³ to 45.2% with 1 m³, and reaching 54.8% with 1.6 m³. In contrast, studies like Passos et al.⁷², which used very high sampling volumes (up to 120 m³), did not detect SARS-CoV-2 RNA in filters collected from environments near COVID-19 patients. Robotto and colleagues (2021) proposed an appropriate range of sampling volumes in their review, ranging from several hundred litres to tens of cubic meters²⁰. According to the results on the positivity rate of the current review, larger sampling volumes would be preferable.

The current results indicate that the sampling time does not have a direct impact on the detection or quantification of SARS-CoV-2 in the air. This is supported by the absence of any observed association between sampling time and the detection, positivity rate or abundance of SARS-CoV-2 genetic material in datasets where sampling was conducted near individuals infected with COVID-19 (Table 4, Table 5). In contrast, a study by Chen et al.¹¹⁹ using impingers (instead of filters) reported that the longer the sampling duration, the lower the viral titers could be recovered¹¹⁹.

No differences were observed in the detection or quantification of SARS-CoV-2 RNA in air near individuals infected with COVID-19 concerning sample preservation conditions during storage in the reviewed studies (Table 4). On the other hand, the meta-regression analysis indicated higher positivity rates for filters stored at 4 °C or analysed within hours after collection (Table 5). In contrast, Conte et al.¹⁰¹ suggested that storing filters at -25°C may be appropriate¹⁰¹.

A few studies (23%) have examined whether several environmental conditions^{26,28,33,46,47,49,52,53,55,66,67,72,77,78,89,90,100,104,105}, such as temperature, relative humidity, irradiance, and wind could affect detection, positivity and abundance of SARS-COV-2 collected on aerosol samples⁴. The low number of these studies focusing on a specific parameter precluded analysing the effect of these on airborne SARS-CoV-2 characterisation in this review.

In addition, other variables, such as the density of COVID-19-infected individuals at the sampling site¹¹⁴, their proximity to sampling devices¹¹⁶, and activities performed by patients during sampling (e.g., talking¹²⁰, singing¹²⁰, coughing¹²¹, or sneezing¹²²), can influence the amount of viral particles released into the air, thus affecting the detection and quantification^4,5. These parameters were not accounted for in the analysis, which is a limitation.

Further studies should evaluate the effect of these factors on the detection, positivity rate and abundance of viral load in aerosols.

Regarding the molecular analysis of air samples, no association has been observed between the target genes and detection, positivity rate or quantification of genetic material extraction (Table 4). A further comparison was conducted on those studies that used several target genes (Table S4) reaching the same conclusion. Thus, all target genes could be recommended.

Internal controls in the extraction of genetic material are essential to verify the efficiency of the extraction process. The absence of such information compromises the ability to assess the analytical quality of the reported data¹²³. Very few studies employed an internal control standard to assess recovery efficiency (10/84, 12%)^{5,10,28,47,66,73,80,101,103,105}, which represents a significant limitation in the validation of the results. In fact, some studies have failed to detect SARS-CoV-2 RNA, even when sampling took place near individuals infected with COVID-19^{82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100}. Of these latter studies, none included an internal control to assess the efficiency of genetic material recovery during extraction, making it impossible to rule out that negative results may have been due to low recovery efficiency. This highlights the importance of including recovery controls in environmental studies, particularly when evaluating the presence of the virus in low concentrations or in high ventilation conditions.

Reporting the LoD and LoQ values is necessary to assess the quality and reliability of the reported data¹²⁴, as information on the analytical sensitivity is especially critical in environmental studies where viral loads are often near the detection limit. Only a quarter of the studies have reported the LoD or LoQ (20/84, 24%)^{5,25,30,34,36,38,50,51,53,54,56,58,71,74,80,88,96,101,103,105}, and only five studies provided these values in terms of RNA per unit of air volume^{5,25,34,50,105}. The lack of LoD and LoQ data hinders the accurate interpretation of viral detection results in air. Moreover, when results are expressed as genomic copies per volume of air, it is essential to report LoD and LoQ in terms of RNA per volume of air²², rather than by RT-PCR well.

Whilst most studies that report Ct values, use a cutoff of 40, the Ct threshold ranged between 35 and 45. Typically, a sample is considered positive when its Ct value is below a threshold of 40, as higher Ct values may indicate nonspecific signals or reduced assay sensitivity. Therefore, setting a Ct cutoff of 40 is standard practice in many laboratories to balance sensitivity and specificity in PCR detection. This threshold helps minimize false positives that could arise from nonspecific amplification at higher Ct values. Setting the Ct threshold below 40 might imply that some environmental samples where SARS-CoV-2 RNA might be present, are not considered positive, becoming a false negative. Low Cts, in the range of 17 and 32 are typically defined in clinical samples as at these concentrations the virus is assumed to be infectious^125,126. Cts higher than those set in clinical samples are more common in environmental samples as shown in this review.

Standardizing the units used to report viral load in air is crucial for comparability. While most studies present results as genetic copies per volume of air (or equivalents), other studies report different units, such as genetic copies per reaction volume^51,52,53, genomic units per total particles in suspension⁷⁸, or nanograms per reaction volume⁷¹. This variability complicates comparisons across studies. The homogenization of these metrics would greatly enhance data interpretation. The recommendation is to present the concentration (i.e. viral load), as well as the LoD and LoQ, in terms of genomic copies of RNA per volume of air. This is to account for the volume of air in which the genetic load has been quantified. Detailed calculations for expressing the results in genomic copies per unit volume of air are provided in Supplementary Note 1.

The detection of SARS-CoV-2 RNA in air samples emerges as a promising indicator for understanding viral presence in different environments, with potential applications for environmental surveillance. The studies reviewed reveal substantial variability in viral loads across environments, including healthcare, transport, residential, and outdoor spaces. Nonetheless, viral load has been detected and quantified in a wide range of public common spaces. The presence of SARS-CoV-2 RNA in environments without prior knowledge of infected individuals underscores its potential as a tool for environmental surveillance^{5,28,47,66,67,68}. This could include monitoring viruses as early warning systems, which would be useful to detect clusters or outbreaks of infections⁹. Early warning could facilitate reducing the spreading of infectious diseases by allowing public health officers to implement appropriate preventive measures on a timely manner^127,128. For instance, some studies using wastewater tracking were able to detect increases in SARS-CoV-2 RNA in environmental samples several days prior to clinical surveillance¹²⁹. A similar approach could be conducted with aerosol sampling in frequently used common public indoor spaces. According to the results of this review, suggested locations for environmental surveillance could be enclosed main public transport stations, hospital accident and emergency departments, city libraries, or indoor markets among other public common spaces.

In line with this application, some studies have recommended bioaerosol sampling as an early warning system for emergent or novel viruses. For instance, routine aerosol sampling in live animal markets combined with metagenomic analysis was suggested as an early warning screening tool system for novel viruses^130,131. Positive aerosol samples could be followed by swabs in animal or cages, thus helping to back-trace to specific farms and contributing to controlling the spread of viruses¹³¹.

Another application of viral characterisation in aerosol samples could be to monitor the prevalence and temporal trend of circulation of the virus in the population in a non-intrusive way⁹. This could complement the existing epidemiological surveillance systems^132,133, as already suggested from results of wastewater surveillance systems¹³⁴.

The ability to detect genetic material in filters suggests that air sampling could also complement traditional epidemiological methods to manage COVID-19 outbreaks in residential locations in a non-invasive manner^{10,40,44,63,65}. It could have the potential utility of sampling the air in different locations within a building for identifying areas at risk of spreading the virus, or virus-free safe spaces. This application was successfully implemented and helped public health officials to manage and control COVID-19 outbreaks in several nursing homes and a merchant ship¹⁰.

Public health containment and preventive measures are based on the mode of transmission of infectious diseases¹³⁵. The detection of viral genetic material in aerosols could contribute to understanding the possible airborne transmission of infectious diseases. Accordingly, appropriate preventive measures to reduce or stop viral airborne transmission, such as ventilation of indoor spaces^106,107,108 could be implemented.

Overall, these potential applications highlight the role of detection and quantification of SARS-CoV-2 RNA from aerosols samples in environmental surveillance. These tools should be developed further to increase the capabilities for pandemic prevention and surveillance following recommendations of the Pandemic Agreement (Article 4)¹³⁶ and consistent with the International Health Regulations (Article 5)¹³⁷.

On the other hand, very few studies have shown positive results for viral viability in filters^55,60,63,65. In line with this, Pan et al.⁷ argued that filtration was not a good sampling method for evaluating viable viruses. During the filtration process, viruses are exposed to dehydration, which may inactivate them, and reduce their ability to remain viable⁷, representing a key limitation for conducting infectivity studies. Similarly, Tang et al.¹³⁸ pointed out that current air sampling technologies do not accurately reproduce the actual processes associated with human respiratory infection through inhalation¹³⁸. This is because the natural airflow rates during human exhalation and inhalation differ significantly from the conditions used by existing sampling techniques. Therefore, studies aiming at collecting aerosol samples to test viability of the virus should consider alternative sampling methods, such as cyclones⁵⁵, or impingers¹¹⁹. In addition, Chang et al.¹³⁹ suggested that prolonged collection of aerosols on liquid samples could not guarantee viral viability, although it was useful for molecular diagnostics (i.e. detection and quantification). They suggested collecting samples over a short period to maximise viral viability¹³⁹. Notwithstanding, whilst the detection of viral RNA in air samples does not necessarily guarantee the presence of viable viruses, its identification can be an indicator of the possible existence of viral particles capable of infecting^4,5.

The characterisation of SARS-CoV-2 RNA load in aerosol samples highlights the need for a multidisciplinary approach combining aerosol scientists, microbiologists and epidemiologists (among others) to address the complexities of studying viral detection in air. The experience gained during the COVID-19 pandemic through the synergy of these interdisciplinary teams should be exploited further to advance the knowledge and develop environmental surveillance tools for airborne respiratory viruses relevant for public health.

As the characterisation of viral genetic material in aerosol samples is a relatively emerging field, the lack of a standardised methodology is noticeable. This makes it difficult to respond quickly in a coordinated manner during outbreaks, as variations in the methods used by different teams can result in inconsistent or non-comparable data. Establishing sound and common protocols would allow for better integration of the information, facilitating both early detection and the implementation of more effective control measures.

Establishing the optimal sampling and molecular RNA analysis conditions is also essential to maximise collection and recovery of viral genetic material from aerosols samples. This is critical when sampling in environments where viral aerosol concentrations may be low, such as in locations that could be integrated in a viral environmental surveillance network. Further method development is recommended to optimise the sampling and analytical parameters under controlled experimental conditions.

Regarding molecular analysis, it is recommended to incorporate quality control and quality assurance measures, such as internal controls, LoD and LoQ. It is also recommended to report abundance of SARS-CoV-2 (i.e. concentrations) in terms of genetic copies of RNA per volume of air.

The heterogeneity of all the studies involved in the analysis for each of the sampling and molecular analysis parameters represents the main limitation of this study, which might have affected the identification of the optimal choice for maximizing SARS-CoV-2 genetic material detection. Furthermore, although statistical analyses were employed to assess differences among the datasets where sampling was conducted near individuals infected with COVID-19, the following factors were not controlled for in the analysis. Distance of the sampler inlet from the patients was not taken into consideration, whereas there is evidence that samples collected in close distance to a patient ( < 2 m) yield higher RNA concentrations¹¹⁶. The number of patients present in the room was not taken into consideration, whilst the number of patients shedding titers into the air would increase the viral load in the air¹¹⁴. Information on the volume of the sampled space has not been accounted for. The smaller the environment, the higher the potential concentration of viral load in the air, in the absence of effective ventilation. The day post-infection of the patients was not taken into account. The amount of infectious viral shedding would be higher between the 3rd and 7th day post infection, with the intensity of the shedding varying depending on the SARS-CoV-2 variant¹⁴⁰. The volume of infectious units exhaled by the patients is also unknown, whilst there is evidence on the variability of the viral load exhaled by different patients¹¹¹. There could be also super spreader patients that could contribute to elevated viral load in the aerosol collected¹⁴¹ in some studies, but this information was unknown. Information on the use of facemask was not considered in the analysis, which might have attenuated the shedding of viral load into the air¹⁴². Therefore, it is recommended to conduct independent experimental evaluations for each of the parameters, where as many variables as possible can be controlled to determine the optimal sampling and analysis parameters.

Another limitation is the exclusion of sampling methods other than those that use filters. On the other hand, to the best of our knowledge, this is the first study to evaluate critically sampling and molecular analysis parameters in relation to the detection, positivity rate and quantification of SARS-CoV-2 genetic material.

This review was conducted using two databases, one targeting specialized coverage in biomedical sciences and another with a broad multidisciplinary scope, including additional references via backward snowballing method. It also focused only on articles published in English, being the predominant language in international scientific literature. Despite not including more than two databases and restricting the review to English-only studies, this review included the largest number of studies, 84 studies, all focused on filter-based methods for sample collection, that yielded 104 different datasets to review. Finally, the review has used statistical analysis, including meta-analysis, to compare differences among parameters in the detection and quantification of SARS-CoV-2 load in aerosol samples.

In conclusion, this systematic literature review found that several sampling factors increased the positivity rate of detection of SARS-CoV-2 RNA in aerosols collected using filter-based methods. A meta-regression analysis identified that larger volumes, using filters other than PTFE or gelatine, storing filters at 4 °C post sampling and analysis within a few hours of collection increased the positivity rate. On the other hand, no differences were observed as regards detection or quantification of SARS-CoV-2 RNA in datasets collected near individuals infected with COVID-19. Given some of the limitations identified in this review, it is highly recommended to perform further methodological development experiments to optimize sampling and analytical conditions under controlled conditions.

SARS-CoV-2 has been detected in a wide range of environments (71% of the datasets surveyed). A higher viral load was observed in indoor environments compared to outdoors, as well as when the sampling was conducted near patients. This suggests the role of ventilation and presence of infected people on aerosol accumulation indoors and the importance of implementing appropriate prevention measures to mitigate the transmission risk in indoor environments, especially where cases might be present. Notwithstanding, the virus was even detected in public indoor spaces where the presence of COVID-19 cases was not known.

The critical appraisal of the body of evidence generated during the COVID-19 pandemic provides a solid foundation for advancing environmental surveillance and addressing future challenges related to respiratory pathogens. Several applications of viral characterisation on aerosol samples were suggested. These included early warning systems, non-intrusive monitoring of environmental viral prevalence and temporal trends, managing COVID-19 outbreaks, and characterising the airborne transmission of viruses.

Methods

Search strategy and eligibility criteria

This systematic review encompasses studies published up to October 2, 2024, across the PubMed and Web of Science databases. Terms such as “aerosol”, “PM”, “air sample”, “SARS-CoV-2”, “indoor”, and “outdoor” were used in the search, employing Boolean operators AND and OR to construct the search syntax for each database: PubMed search algorithm (n = 236) (“aerosol” OR “PM” OR “air sample”) AND (“SARS-CoV-2” OR “SARSCOV2”) NOT (“model” OR “modelling” OR “modeling”) AND (“indoor” OR “outdoor”) and ISI Web of Science algorithm (n = 263) (“aerosol” OR “PM” OR “air sample”) AND (“SARS-CoV-2” OR “ SARSCOV2”) (All Fields) not “model” OR “modelling” OR “modeling” (All Fields) and “indoor” OR “outdoor” (All Fields). To refine the results, a restriction was applied using the Boolean operator NOT for terms such as “model” or “modelling”. A complementary snowballing strategy was applied by screening the reference lists of key articles to identify additional relevant studies.

The inclusion criteria comprised articles published in English; original scientific research or journal articles on the topic; articles that sampled both indoor and/or outdoor air; and studies employing direct filtration or any method that uses filters as the sampling method, such as impactors or cyclones that include filters in some of their stages. Exclusion criteria included articles in languages other than English; review articles; studies on mathematical modelling; on preventive measures or atmospheric pollution; studies that use sampling methods in which no filter is used; and studies aimed at virus elimination.

Review and extraction process

The article selection process was conducted using the free tool Rayyan. Two reviewers (Alfaro, C. and Barberá-Riera, M.) screened all titles, abstracts and full-text. Discrepancies and queries on selection were resolved by a third reviewer (Delgado-Saborit, JM).

The included studies were analysed by two researchers (Alfaro, C. and Barberá-Riera, M.) to extract key data on detection, quantification, and viability, as well as sampling parameters, such as environments and microenvironments, proximity to COVID-19 infected individuals, air volume sampled, sampling duration, airflow rate, filter type, sample transportation and storage conditions, and the recording of additional parameters during sampling. Molecular analysis parameters for SARS-CoV-2 RNA were also reviewed, including internal controls, targeted genes, criteria for positive samples, and LoD and LoQ. All airflow rates were standardized to L/min, and air volumes were converted to m³ for consistency.

Analysis of the reviewed evidence

The results of SARS-CoV-2 genetic material detection have been presented as detection (yes/no), which refers to those settings where at least one sample was positive. It also includes the positivity rate, which refers to the percentage of filters where genetic material was detected compared to the total number of filters collected. The results of SARS-CoV-2 RNA quantification have been reported as the median and Interquartile Range (IQR) (25th percentile–75th percentile) of the viral load (i.e. concentration), in the units reported by the original authors.

Fisher’s Exact Test was used to evaluate differences of genetic material detection (Detection/No Detection) between categorical variables such as sampling environments (healthcare, socio-health, transport, other indoor, outdoor), and proximity to COVID-19 infected individuals among the reviewed studies (Table 3). Differences in concentrations of genomic copies/m³ of air (gc/ m³, or equivalent metrics reported in some studies) and sampling environments or proximity to infected individuals in the reviewed studies were tested with the Kruskal-Wallis Test (Table 3). The Chi-Square Test and Mann-Whitney U Test were used to assess differences between indoor versus outdoor sampling with genetic material detection and concentration reported in the reviewed studies, respectively (Table 3).

The evaluation of sampling and molecular analysis parameters in relation to the SARS-CoV-2 detection and viral load, was conducted only with studies in which sampling was conducted in proximity to COVID-19 infected individuals. This would ensure, in some way, the presence of an infection source near the sampling point. For associations between sampling parameters (airflow rate, air volume, sampling duration, type of filter and sample storage) or molecular analysis parameters (target gene) versus detection and viral load variables, Mann-Whitney U Test, Fisher’s Exact Test, Spearman’s Correlation, and Kruskal-Wallis Test were applied as appropriate (Table 4). Spearman’s correlation coefficient (rho) between the viral concentration and the flowrate, sampling time or sampling volume was reported where applicable.

In the case of the positivity rate, a meta-analysis of proportions was performed using a random-effects model to estimate the pooled proportion of SARS-CoV-2-positive samples on air filters (i.e. positivity rate), as well as its 95% confidence interval. Since some studies presented proportions close to 0 or 1, the Freeman-Tukey double arcsin transformation was applied to stabilize the variance before performing the analysis.

As regards, the concentrations, a meta-analysis of means was performed using a random-effects model to estimate the pooled mean concentration of SARS-CoV-2 genetic material (expressed in copies/m³) detected on air filters, along with its 95% confidence interval. For each study, the mean concentration of SARS-CoV-2 in the air, estimated from the filters analysed, was used as a summary measure. Only studies conducted indoors, with sampling near COVID-19-infected individuals, and that quantified viral RNA concentrations were included in the meta-analysis.

Heterogeneity between studies was estimated using the I² statistic and the DerSimonian-Laird test. The results of the meta-analysis were graphically represented using a forest plot.

The existence of potential publication bias was analyzed using the funnel plot and the Egger test. In the case of the positivity rate, since the proportions were transformed using the Freeman-Tukey double arcsin, the Egger test was only exploratory, as this transformation does not fully meet the linear symmetry assumptions required for its formal application.

A sensitivity analysis was performed by repeating the meta-analysis, successively excluding each of the studies, in order to evaluate the robustness of the results, both for the positivity rate and the concentrations meta-analysis.

In the case of the positivity rate, a bivariate meta-regression was also performed to explore whether sampling characteristics, such as sampling setting (healthcare, social care, transportation, other indoors, outdoors), type of setting (indoors, outdoors), and proximity to infected individuals, explained some of the observed heterogeneity in the positivity rate. This variable was introduced as a categorical factor, and the transformed proportion and its standard error were used as the dependent measure. The analysis was weighted by the precision of each study. This analysis included only studies that sampled near COVID-10 infected people and sampling conducted indoors.

Statistical significance was defined as a p-value below 0.05. All analyses were performed using RStudio, except the meta-analysis, which was performed in STATA.

It is important to note that some studies reported results collected in different environments (e.g. indoors and outdoors); or used several flow rates, sampling durations and volumes, filter types, or target genes. Each sampling combination was considered an independent dataset in both qualitative and quantitative analyses. Thus, whilst only 84 studies were included in this review, the number of independent datasets is larger, totalling 104 different microenvironment locations.