Concentrations of nitrogen compounds are related to severe rhinovirus infection in infants. A time-series analysis from the reference area of a pediatric university hospital in Barcelona.

BACKGROUND: There is scarce information focused on the effect of weather conditions and air pollution on specific acute viral respiratory infections, such as rhinovirus (RV), with a wide clinical spectrum of severity.
OBJECTIVE: The aim of this study was to analyze the association between episodes of severe respiratory tract infection by RV and air pollutant concentrations (NOx and SO2 ) in the reference area of a pediatric university hospital.
METHODS: An analysis of temporal series of daily values of NOx and SO2 , weather variables, circulating pollen and mold spores, and daily number of admissions in the pediatric intensive care unit (PICU) with severe respiratory RV infection (RVi) in children between 6 months and 18 years was performed. Lagged variables for 0-5 days were considered. The study spanned from 2010 to 2018. Patients with comorbidities were excluded.
RESULTS: One hundred and fifty patients were admitted to the PICU. Median age was 19 months old (interquartile range [IQR]: 11-47). No relationship between RV-PICU admissions and temperature, relative humidity, cumulative rainfall, or wind speed was found. Several logistic regression models with one pollutant and two pollutants were constructed but the best model was that which included average daily NOx concentrations. Average daily NOx concentrations were related with the presence of PICU admissions 3 days later (odds ratio per IQR-unit increase: 1.64, 95% confidence interval: 1.20-2.25)).
CONCLUSIONS: This study has shown a positive correlation between NOx concentrations at Lag 3 and children's PICU admissions with severe RV respiratory infection. Air pollutant data should be taken into consideration when we try to understand the severity of RVis.

INTRODUCTION

Rhinovirus (RV) is one of the main causes of respiratory tract infection in children. RV infection (RVi) ranges from asymptomatic or mild symptomatic infection to being often the only etiological agent in patients requiring advanced life support in intensive care units. Viral and bacterial coinfections have been considered as factors associated with severe disease, but these results are not consistent between series , and the wide clinical spectrum of RVi has not been clearly explained by a unique element.

On the other hand, exposure to air pollution (i.e., sulfur dioxide [SO2], ozone [O3], nitrogen dioxide [NO2], nitrogen oxides [NO ], and particulate matter [PM]) has been associated with respiratory exacerbations in asthmatic children and adults. In recent years, the literature collects some studies that show associations between short‐term and long‐term changes in air pollutants and increased rates of respiratory admissions among young children, especially in those under 4 years old. , , ,

Overall, there is scarce information focused on the effect of weather conditions and air pollution on specific acute viral respiratory infections, such as RV. RV is a respiratory virus with a poor seasonal pattern of circulation, being detected along all the year seasons in our setting, whereas other respiratory viruses cause epidemic peaks only during winter.

The aim of this study was to analyze the association between episodes of severe respiratory tract infection by RV and air pollution concentrations in a reference area of a university hospital. We hypothesized that either the susceptibility to RVi or its severity could be partially explained by air pollution. Weather variables, circulating pollen grains, and mold spores were also considered. Therefore, this study focuses on elucidating if the air pollutant data should be taken into consideration as another of the variables that may help the scientific community to understand the wide severity range of RVi in children.

MATERIALS AND METHODS

Variables

This study analyses a temporal series of daily counts of admissions in a pediatric intensive care unit (PICU) due to severe respiratory RVi, in patients from 6 months to 18 years old, residing in the reference area (Barcelona‐South and Baix Llobregat) of a pediatric reference center (Hospital Sant Joan de Déu, Barcelona). This reference area includes a pediatric population of about 300,000 children. Daily weather variables, concentrations of pollen grains, mold spores and air pollutants (NO and SO2) were collected from regional stations. The study spanned from 2010 to 2018.

Daily weather variables (average temperature, average wind speed, and cumulative rainfall) were collected from El Prat Airport Station, Baix Llobregat. Relative humidity was also collected from April 2011 to December 2018 from a station of the same municipality (this data was not available from January 2010 to March 2011). Regarding air pollutants, daily arithmetic mean, as well as the daily maximum peak values of sulfur dioxide (SO2) and nitrogen oxides (NO , which included nitric oxide [NO] and nitrogen dioxide [NO2]) were monitored from El Prat station (Baix Llobregat). These data were obtained from the Surveillance Network and Air Pollution Forecast from Generalitat of Catalonia. Mean daily pollen and spore concentrations from the station in Barcelona city center were collected from the Aerobiological Network of Catalonia.

Definitions

Severe RVi was defined as an acute lower respiratory tract disease (bronchiolitis, bronchitis, or pneumonia) with the need of mechanical ventilation and microbiologically confirmed identification of RV presence in nasopharyngeal aspirates by RNA polymerase chain reaction (PCR) detection. Patients with respiratory (except asthma), cardiac, and neurological chronic diseases were excluded from the count. Addresses of all cases were checked and those who did not live in the reference area (Barcelona‐South and Baix Llobregat) were excluded. For reference, 90% of the respiratory admissions in the PICU of this hospital correspond to patients living in this area.

Statistical analysis

Continuous non‐normal distributed variables were described as median value and interquartile range (IQR). Several logistic regressions with “day with PICU admission (yes/no)” as the outcome variable and income variables were built (generalized linear models). The choice of the optimal degrees of freedom of cubic functions and the choice of the model, which combined the highest goodness of fit and the lowest number of variables was made by minimization of the corrected Akaike Information Criterion (AIC). This strategy also helps to avoid overfitting.

An initial model was created with:

‐ a natural cubic smooth function of calendar time, with 7 degrees of freedom (df) per year to exclude unmeasured long‐term and seasonal trends. Models with 3 and 5 df were also considered, but AIC values were similar among them and several studies in international literature used 7 df per year with the same purpose. ,

‐ an indicator of the day of the week, as a potential confounder.

‐ a lagged variable of the daily average temperature (0‐ to 5‐day shifted series), as the risk of RVi had been associated with lower temperatures in several series. A linear function was finally used after exploring a nonlinear effect with 3 df.

To select those variables to be added to the initial model, associations between the number of daily PICU admissions and other main variables were explored. For each variable with a Mann–Whitney U test p < 0.10, a model was created. Those variables were introduced using lagged values (0‐ to 5‐day shifted series), which were simultaneously entered in the models (“distributed lag model”). Other studies have used similar methodologies. , Lagged variables were used from 0 to 5 days, because this captures the incubation period of RVi. If more than one variable was associated at some lag‐time with the outcome with a p < 0.10, a model with >1‐pollutant would be considered. A strict p (<0.01) was considered statistically significant to minimize the possibility of chance findings. The corrected AIC was used to compare and to identify the best final model among those with an associated variable with the outcome. Residual autocorrelation was checked in the final model.

Statistical analysis was performed with R software v 4.0.2 and the following packages: dlnm, ggplot2, splines, AICcmodavg.

Ethics approval

This study was approved by the Ethics Committee and Institutional Review Board of the Sant Joan de Déu Hospital (PI179‐21) and informed consents from patients' parents were waived.

RESULTS

During the study period, 150 patients (independent observations) were admitted to the PICU with RV detection by PCR in nasopharyngeal aspirate. These patients corresponded to 26 municipalities of Barcelona and the metropolitan area. Mean number of admissions to PICU per day was 0.046 (95% confidence interval [CI]: 0.04–0.05). There were admissions in 146/3285 days (142 days with 1 admission and 4 days with 2). Median age was 19 months old (IQR: 11–47) and 90 (61%) were males (Figure 1).

Figure 1

Temperature and other variables associated with the number of pediatric intensive care unit (PICU) admissions (SO2, NO , and air pollen concentrations) and number of PICU admissions during the study period.

 *Bold spots: number of rhinovirus‐related PICU admissions; soft spots: values of weather, pollen or air pollutant variables.

Regarding contaminants, median of daily average SO2 was 2 µg/m3 (IQR: 1–3). Median of the daily peak values of SO2 was 5 µg/m3 (IQR: 3–7). The median of daily average NO was 49.7 µg/m3 (IQR: 33.5–73.2) and the median of daily peak values was 149 µg/m3 (IQR: 87–223). More detailed information about weather conditions and airborne concentrations of contaminants, pollen grains, and mold spores during the study period can be found at Table 1.

Table 1

Weather variables and average daily concentrations of air pollutants, pollen grains, and mold spores of a reference area and number of RV‐related PICU admissions of children living in this reference area (2010–2018).

	2010	2011	2012	2013	2014	2015	2016	2017	2018	2010–2018
Max temperature (°C) a	20.0 (14.2–25.5)	21.2 (16.4–26.9)	21.0 (16.3–26.7)	20.3 (15.1–25.8)	21.0 (16.6–26.1)	20.7 (16.7–26.2)	20.2 (16.8–26.1)	21.1 (18.9–26.8)	21.6 (16.5–27.2)	20.7 (16.3–26.4)
Min temperature (°C) a	11.2 (6.2–18.2)	13.7 (7.4–18.6)	12.7 (7.5–19.1)	12.8 (7.5–18.7)	14.1 (8.9–18.9)	13.4 (7.9–18.8)	12.9 (8.6–18.9)	12.9 (7.8–19.1)	13.9 (8.1–20.3)	13.1 (7.9–18.9)
Average temperature (°C) a	15.2 (10.2–21.6)	17.3 (11.9–22.7)	16.7 (11.7–22.9)	16.5 (11.1–22.2)	17.3 (12.4–22.6)	17.1 (12.2–22.6)	16.6 (12.7–22.5)	16.7 (12.3–22.5)	17.8 (12.1–23.9)	16.8 (11.9–22.6)
Relative humidity (%) a	–	75 (69–80)	72 (64–79)	72 (66–78)	75 (69–81)	74 (67–79)	73 (66–79)	72 (66–78)	74 (67–81)	73 (67–79)
Cumulative rainfall (mm) a	0 (0–0.1)	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–0.2)	0 (0–0)
Average wind‐speed (km/h) a	15.1 (11.9–18.0)	15.1 (12.4–18)	15.1 (11.8–18.3)	23.0 (18.0–28.0)	15.1 (19.7–19.1)	15.1 (12.9–18.0)	15.1 (12.9–19.1)	15.1 (12.9–19.1)	15.1 (12.4–19.1)	15.8 (12.9–20)
Average SO 2 (µg/m3) a	2 (1–3)	3 (2–4)	2 (2–3.2)	2 (1–3)	2 (1–2)	3 (2–4)	3 (2–4)	2 (2–3)	2 (1–2)	2 (1–3)
Average NO x (µg/m3) a	57.1 (40.3–86.2)	50.1 (36.3–73.4)	54.8 (37.8–80.8)	47.8 (30.4–74.6)	47.8 (30.4–75.8)	53.1 (33.4–78.6)	49.2 (31.7–72.4)	47.4 (33.5–70.2)	47.4 (34.4–62.4)	49.7 (33.5–73.2)
Average pollen grains (g/m3) a	20.3 (7.7–105.5)	24.5 (8.4–165.4)	22.4 (8.4–118.3)	34.3 (10.8–157.1)	21.0 (9.1–168.7)	18.9 (8.4–124.6)	24.5 (9.8–140.0)	19.9 (9.8–138.2)	27.3 (9.8–141.4)	23.1 (9.1–140.3)
Average mold spores (n/m3) a	1519 (963–2444)	1671 (955–3528)	1596 (777–3151)	1601 (757–3257)	1243 (700–2596)	1126 (419–2507)	1159 (659–1984)	1080 (622–1900)	1697 (871–9277)	1400 (742–2746)
Number of PICU admissions (n)	16	6	15	14	15	28	25	11	20	150

Abbreviations: PICU, pediatric intensive care unit; RV, rhinovirus.

Median and interquartile ranges.

Average SO

(µg/m a

Average NO

(µg/m a

(g/m a

(n/m a

SO2, NO , and airborne pollen concentrations were potentially associated with the outcome in the univariate analysis and, therefore, these variables were introduced in the time‐series regression models (Table 2).

Table 2

Spearman ρ matrix correlation between the values of weather variables and airborne concentrations of pollutants, pollen grains, and mold spores, and associations of these variables with the outcome (PICU admissions yes/no).

	Average temperature (°C)	Cumulative rainfall (mm)	Average wind speed (km/h)	Relative humidity (%)	Average daily SO2 concentrations (µg/m3)	Average daily NO x concentrations (µg/m3)	Daily peak SO2 (µg/m3)	Daily peak NOx (µg/m3)	Airborne pollen grains (g/m3)	Average mold spores (n/m3)
Cumulative rainfall	−0.056c
Average wind speed	0.072c	0.016
Relative humidity	0.001	0.327c	−0.199c
Average daily SO2	−0.200c	−0.136c	−0.209c	−0.017
Average daily NO x	−0.464c	−0.050c	−0.281c	0.178c	0.492c
Daily peak SO2	−0.165c	−0.149c	−0.231c	−0.046b	0.823c	0.488c
Daily peak NO x	−0.349c	−0.068c	−0.300c	0.178c	0.418c	0.892c	0.478c
Airborne pollen	0.015	−0.102c	0.140c	−0.192c	0.047c	−0.204c	0.100c	−0.170c
Average mold spores	0.404c	0.097c	0.000	−0.155c	−0.101c	−0.141c	−0.133c	−0.166c	−0.015
PICU admissionsa
Yes, median (IQR) No, median (IQR)	14 (11–20) 17 (12‐23)c	0 (0–0) 0 (0–0)	15 (13–20) 16 (13–20)	73 (67–79) 73 (65–80)	3 (2–4) 2 (1–3)c	70 (43–100) 51 (35–74)c	6 (4–8) 5 (3–7)c	194 (119–285) 147 (86–220)c	16 (7–99) 24 (9–143)b	1268 (683–2908) 1407 (745–2733)

Abbreviations: IQR, interquartile range; PICU, pediatric intensive care unit.

Mann–Whitney U test was performed to make these comparisons.

p < 0.10.

p < 0.01.

Average daily NO concentrations

Five logistic regression models were built, one with each of the following variables added to the initial model: average daily SO2, average daily NO , daily peak value of SO2, daily peak value of NO , and airborne pollen. Models with daily average SO2 concentrations, the one with daily average NO , and the one with the peak values of NO had their pollutant values at Lag 3 potentially related to the outcome with a p < 0.1 (Table 3). Average daily temperature was not associated with the outcome in any of these models. As NO daily average and peak values were strongly correlated between them, two models with two pollutants were built: one with daily average SO2 and NO concentrations, and another one with daily average SO2 values and peak NO concentrations (Table S1). The model including a potential variable related with the outcome with a p < 0.01 and with the lowest AIC value was that of NO daily average. In this model, NO concentration at Lag 3 was associated with PICU admissions with an odds ratio (OR) per IQR increase of 1.64, 95% CI: 1.20–2.25). There was no remaining significant autocorrelation in the deviance residuals.

Table 3

Logistic regression models constructed to explain RVi‐related PICU admissions

	Univariate models
	Model with average airborne pollen concentrations (OR per 131 u increasea)	p	Model with average daily SO2 concentrations (OR per 2 u increasea)	p	Model with average daily NO x concentrations (OR per 40 u increasea)	p	Model with daily SO2 peaks (OR per 4 u increasea)	p	Model with daily NO x peaks (OR per 136 u increasea)	p
Day 0	0.89 (0.73–1.07)	0.209	1.25 (0.89–1.77)	0.198	1.20 (0.90–1.60)	0.207	1.06 (0.94–1.20)	0.352	1.28 (0.96–1.71)	0.091b
Day 1	1.09 (0.93–1.29)	0.292	1.03 (0.69–1.53)	0.896	0.92 (0.67–1.27)	0.613	1.04 (0.89–1.22)	0.630	1.07 (0.78–1.46)	0.666
Day 2	1.00 (0.84–1.18)	0.971	0.88 (0.59–1.31)	0.535	0.99 (0.72–1.36)	0.932	0.96 (0.79–1.16)	0.670	1.01 (0.74–1.39)	0.924
Day 3	1.01 (0.85–1.20)	0.936	1.49 (1.01–2.20)	0.045b	1.64 (1.20–2.25)	0.002c	1.11 (0.95–1.29)	0.183	1.39 (1.03–1.89)	0.033b
Day 4	1.07 (0.89–1.28)	0.497	0.84 (0.56–1.26)	0.397	0.88 (0.64–1.22)	0.448	0.90 (0.70–1.16)	0.420	1.09 (0.79–1.48)	0.596
Day 5	0.88 (0.73–1.06)	0.189	1.04 (0.72–1.49)	0.852	0.92 (0.69–1.23)	0.575	0.98 (0.78–1.23)	0.866	0.78 (0.57–1.06)	0.110
cAIC	1097		1123		1108		1132		1123

Abbreviations: cAIC, corrected Akaike Information Criterion; OR, odds ratio; PICU, pediatric intensive care unit; RVi, rhinovirus infection.

An interquartile range unit increase.

p < 0.10.

p < 0.01.

Model with average daily NO concentrations (OR per 40 u increasea)

Model with daily NO peaks (OR per 136 u increasea)

DISCUSSION

We investigated the association between PICU admissions with RVi and air pollution concentrations in Barcelona South‐Baix Llobregat for 9 years. Average daily NO concentrations the third day before the PICU admission was related to PICU admissions with RVi. Our observations were made in a metropolitan area with more than 1 million of inhabitants (Barcelona‐South and Baix Llobregat). Nitrogen oxides are traffic pollutants from on‐road vehicles and Barcelona is the European city with the densest traffic, twice as much as Madrid and three times as dense as London.

Our analysis included a natural cubic smooth function of calendar time with 7 df and, therefore, only short‐term effects can be captured. Taking this into consideration, our study showed no relationship of the outcome with changes in temperature, relative humidity, cumulative rainfall, or average wind speed. Other studies carried out in Mediterranean countries support our findings, as they did not find any association between meteorological variables and viral infection caused by RV. On the other hand, respiratory exacerbations caused by other respiratory viruses, such as RSV and influenza, have been related to changes in weather variables. , Specifically, relative humidity has been positively associated with RSV activity, whereas absolute humidity has been inversely correlated with influenza viruses' circulation, , being a good predictor of the onset of each influenza season. Other authors who analyzed the incidence of respiratory infections without considering their viral etiologies have not observed any association between these episodes and changes in relative humidity. Many viral infections have non‐coincidental seasonal peaks and others, such as RV, are all‐year viruses. All these data illustrate that it might be important to discern between viral etiologies when we try to understand the weather effect contribution to respiratory viruses spreading.

With regard to air pollen concentrations, it has been associated with children's respiratory admission cases. Probably, we did not observe this correlation, because most of the patients requiring PICU admission in our series were children under 2 years old and sensitizations to air allergens in this age group are not yet fully developed.

About the association of air pollutant concentrations and respiratory admissions in children, some studies in the international literature have observed similar very‐short term effects, most of them in lag periods of up to 5 days. , , Moreover, Fusco et al. observed that same‐day levels of NO2 were associated with an increase of 4% in admissions due to respiratory infections in children.

It is known that exposure to air pollutants (NO2, NO , and PM < 10 µm) decreases lung function , and induces oxidative stress, resulting in the production of free radicals, which may damage the respiratory system, reducing the resistance to viral infections. Wong et al. described a significant association between the level of NO2 and admissions with pneumonia and influenza, indicating that exposure to NO2 increases the susceptibility to influenza virus infection. Spannhake et al. investigated the effects of NO2 exposure and RVi in human nasal and bronchial epithelial cells in vitro. They found that exposure to NO2 after RVi increases the production of interleukin‐8 greater than only NO2 exposure, making it evident that a direct additional effect of NO2 exposure over the RVi exists. In the same way, Greve et al. showed that NO2 exposure enhances the expression of intercellular adhesion molecule 1. This molecule is a receptor for the RVs and this could be a potential mechanism explaining the observed increased susceptibility to RVi by NO2 exposure in other settings. Moreover, assessing the associations between severity of RVi and NO in children aged between 8 and 11 years, Chauhan et al. described that a high exposure to NO during the week before starting a respiratory viral infection by RV or respiratory syncytial virus was associated with an increase in the severity of asthma exacerbations caused by these viruses. These results are in agreement with our findings about the involvement of nitrogenous compounds in RVi severity.

The main limitation of this study is the low number of admissions with severe RVi. However, as far as we know, this is one of the largest series exploring the associations between admissions with severe RVi requiring intensive care and air pollutants. Second, despite using a strict cutoff p, chance findings are still possible due to the many variables analyzed. Finally, our time‐series analysis relies on data from only one fixed‐site monitoring station. Thus, we want to remark that our findings are the result of an exploratory study and further multicenter research with more outcomes and with data from other settings are needed.

In conclusion, this study shows a positive association between average daily NO concentrations and children's PICU admissions with severe RV respiratory infection 3 days later. Therefore, the wide range of clinical severity of RVi could partially be explained by air pollutants concentrations (nitrogen compounds). These results could be helpful to adopt policies that mitigate environmental pollution and the burden of this common viral infection in young children.

AUTHOR CONTRIBUTIONS

Georgina Armero: Conceptualization; methodology; investigation; data curation; formal analysis; writing—original draft. Daniel Penela‐Sánchez: Data curation; investigation; writing—original draft; methodology; formal analysis. Jordina Belmonte: Data curation; investigation; writing—review and editing. Diana Gómez‐Barroso: Investigation; methodology; validation; formal analysis; writing—review and editing. Amparo Larrauri: Investigation; methodology; validation; formal analysis; writing—review and editing; supervision. Desiree Henares: Conceptualization; investigation; writing—original draft. Violeta Vallejo: Writing—review and editing; investigation; data curation; validation. Iolanda Jordan: Writing—review and editing; conceptualization; investigation; supervision. Carmen Muñoz‐Almagro: Validation; visualization; conceptualization; investigation; data curation; writing—original draft; writing—review and editing. Pedro Brotons: Funding acquisition; writing—review and editing; writing—original draft; conceptualization; methodology; formal analysis; investigation. Cristian Launes: Conceptualization; investigation; funding acquisition; writing—original draft; methodology; writing—review and editing; formal analysis.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information.

Click here for additional data file.

Supporting information.

Click here for additional data file.