Analysis of the effect of PM10 on hand, foot and mouth disease in a basin terrain city

Hand, foot, and mouth disease (HFMD) is a common childhood infection that causes a substantial disease burden in the Asia-Pacific region. Various climate variables, such as humidity and temperature, have been associated with HFMD. However, few studies have assessed the impact of PM10 on childhood HFMD. This study investigated the association between PM10 and HFMD. We fitted a standard distributed lag non-linear model to investigate the temporal lagged relationship between PM10 and HFMD, and then further assessed whether this relationship varied by gender and pathogen. Between 2011 and 2015, a total of 122,564 HFMD cases under 15 years of age were reported in Chengdu. The PM10-HFMD associations were shown to be non-linear in all subgroups, with the peak at 101–218 μg/m3. Male children were more sensitive to PM10 effects. For pathogen-specific relative risks, we found that the risk estimates were generally higher in cases of CVA16 infection. Our study provides evidence that PM10 increases the risk of HFMD. Authorities and parents should be fully aware of the impact of PM10 on childhood HFMD. Furthermore, appropriate protective measures should be taken to reduce risks.

In this study, we investigated the relationship between PM 10 and HFMD, and then further assessed whether this relationship varied by gender and pathogen. Specifically, we fitted a standard distributed lag non-linear model to examine the nonlinear lagged effects of daily PM 10 on HFMD incidence. A better understanding might be attained for the association between PM 10 and HFMD by adopting a more sophisticated model.

Results
From January 1, 2011 to December 31, 2015, a total of 122,564 HFMD cases under 15 years of age were reported in Chengdu, of which 5,000 (4.08%) were laboratory confirmed. About 97% of HFMD cases occurred in children aged 0-5 years. Of 122,564 HFMD cases, 72,401 were males and 50,163 were females (male-to-female sex ratio: 1.44). Among the laboratory confirmed cases, 1,418 (28.36%) were associated with EV71, and 971 (19.42%) were associated with CVA16. Table 1 shows the descriptive statistics for meteorological variables, PM 10 , and HFMD cases. On average, there were 67.12 daily HFMD cases during the study period. The mean values of daily PM 10 concentration, relative humidity, and mean temperature were 119.51 μg/m 3 , 78.47%, and 16.33 °C, respectively. The mean value of daily SO 2 and NO 2 were 25.59 μg/m 3 and 55.21 μg/m 3 , respectively. The daily mean PM 10 concentration significantly exceeded the values recommended by the national air quality standards and the World Health Organization (WHO).
The 3D plot displays the association between PM 10 and HFMD incidence during 14 lag days (Fig. 1). Overall, there was a non-linear association between PM 10 and HFMD incidence. In addition, the results suggested that the association between HFMD and Pm 10 might have a different lag structure. For instance, low PM 10 (50 μg/m 3 ) had a minimum relative risk (RR) at lag day 0 (the current day) and arrived the peak at lag day 9, and declined  www.nature.com/scientificreports www.nature.com/scientificreports/ slowly on the following days. High PM 10 (300 μg/m 3 ) had a maximum RR at lag day 5, which declined rapidly on the following days. Figure 2 illustrates the cumulative impact of PM 10 on HFMD over the 14-day period. The results indicated that PM 10 was statistically associated with HFMD. We found that the PM 10 -HFMD relationship exists as an approximately inverted V-shape curve. The cumulative RR increased with PM 10 . The curve peaked at 118 μg/m 3 , and then the curve began to decline. Figure 3 presents the cumulative RRs of PM 10 on HFMD incidence over 14 days in different subgroups. For different gender groups, the exposure-response curves followed similar patterns. The cumulative RR reached a peak at 113 μg/m 3 for male children and 131 μg/m 3 for female children.
The cumulative effects of PM 10 on different pathogen groups are also shown in Fig. 3. We found that the relationship between PM 10 and HFMD varied across pathogen groups. For EV71-related cases, the exposure-response curve reached a peak at 218 μg/m 3 , and then the curve began to decline. For CVA16-related cases, the exposure-response curve reached a small peak at 101 μg/m 3 , and then the curve started to decrease. Again, the curve increased when PM 10 was above 383 μg/m 3 .
The relative risks of different PM 10 values along the lags for total, gender-, and pathogen-specific HFMD cases are displayed in Table 2. For gender-specific RRs, the risk estimates were generally higher in male children. For female children, the highest RR was 1.36 (95% CI: 1.17-1.58) at lag 0-14. For male children, the highest RR was 1.59 (95% CI: 1.40-1.81) at lag 0-14. When comparing risks for different pathogen groups, we found that the risk  www.nature.com/scientificreports www.nature.com/scientificreports/ estimates were generally higher in CVA16-related cases. For CVA16-related cases, the highest RR was 9.46 (95% CI: 3.54-25.25) at lag 0-14, while for EV71-related cases, the highest RR was 4.40 (95% CI: 1.93-10.06) at lag 0-14 when PM 10 was 101 μg/m 3 .
In the sensitivity analyses, similar trends were observed when the degrees of freedom for weather factors and air pollutants were changed between 3 and 6.

Discussion
Ambient air pollution has become one of the world's biggest public health issues because of its profound impacts on the vast majority of the world's population. The present research quantified the effect of PM 10 on childhood HFMD, using Chengdu as the research area. The daily mean PM 10 concentration in Chengdu significantly exceeded the values recommended by the national air quality standards and WHO. The current study has several strengths. To our knowledge, this is the first research to investigate the association between PM 10 and childhood HFMD in a basin city, and to further evaluate gender-and pathogen-specific effects.
In this study, we fitted a standard distributed lag non-linear model to investigate the temporal lagged relationship between PM 10 and HFMD. Our findings showed that an association exists between PM 10 and HFMD. The exact mechanisms are not known; however, one possible explanation is that viruses may attach to the tiny particles in the air, thus facilitating the transmission of HFMD. In addition, we found an inverted V-shape pattern for the exposure-response curve. One possible reason is that the susceptible population might prefer staying at home or taking measures (e.g., wearing masks) to protect themselves on high air pollution days. EV71 and CVA16 have been widely considered to be the predominant pathogens causing outbreaks of HFMD worldwide 19,20 . However, no studies have quantified the effects of PM 10 on HFMD among different pathogen groups. We found that the PM 10 -HFMD relationship differed between EV71 infection and CVA16 infection. Further research is needed to examine the pathogen-specific effects of meteorological factors and air pollutants on HFMD incidence.
The results of our study also suggest that the relationship between PM 10 and HFMD varies by gender. The results indicated that males appeared to be more sensitive to PM 10 effects. The results of our study are consistent with the previous study. A recent study also indicated that male children owned more prominent cumulative effects than female children 21 . Compared to female children, male children are generally more active and more likely to play outside, which increases their chances of being exposed to air pollutants and enteroviruses.
The pattern of the relationship between PM 10 and HFMD risk has not been consistent among studies. Our study suggests an inverted V-shape pattern for the exposure-response curve over 14 days. Huang et al.   www.nature.com/scientificreports www.nature.com/scientificreports/ between PM 10 and female HFMD cases in Ningbo city (from 2012 to 2016) 15,22 . Luo et al. reported a negative correlation between PM 10 and HFMD in Yuexiu district from 2010 to 2011 16 . A recent study in Guilin city provided a clue that a high PM 2.5 level increased the risk of HFMD, though it was not statistical significant 21  Some limitations of the current study should be acknowledged. First, there are complicated associations among air pollutants. However, only SO 2 and NO 2 were adjusted in the distributed lag nonlinear model (DLNM) since the data of other air pollutants (such as CO and O 3 ) were only available since 2013. Those air pollutants will be examined once a longer time series is obtained. Second, the present research focused on the effect of PM 10 on HFMD. Multi-pollutant models might have a great influence on the robustness of parameter estimation, which might results in the difficulty and limitations of result interpretation 23 . Therefore, when estimating the health impacts of particulate matter, the results of the single-pollutant models were more often reported and may be more representative 23 . In addition, potential interactions may exist between meteorological factors and air pollutants. Further research is required to investigate the interaction effects of meteorological factors and air pollutants on HFMD incidence.
In conclusion, our study provides evidence that PM 10 could increase the risk of HFMD. Authorities and parents should be fully aware of the impact of PM 10 on childhood HFMD. Appropriate protective strategies should be taken to reduce the risks. Furthermore, different PM 10 -HFMD relationships among different subgroups would be helpful in optimizing prevention and control strategies.

Materials and Methods
Data sources. In 2004, a nationwide notifiable infectious diseases reporting information system (NIDRIS) was established in China. In 2008, HFMD was listed as a class C notifiable infectious disease. All of the HFMD cases were required to be reported to the NIDRIS within 24 hours after diagnosis. The diagnosis of HFMD cases was based on the national clinic guide published by the Chinese Ministry of Health 24 . Daily cases of HFMD in Chengdu during 2011-2015 were obtained from NIDRIS. According to the preliminary analysis, 99.60% of HFMD cases occurred among children aged 0-14 years during the study period. So HFMD cases under the age of 15 years were chosen in the present study.
Daily data of air pollutants from 2011 to 2015 were collected from the Sichuan Environmental Monitoring Center, including PM 10 , SO 2, and NO 2 . Meteorological data were obtained on a daily basis from the China Meteorological Data Sharing Service System.

Statistical analyses.
In the present study, a time-series regression based on a generalized linear model with a poisson family was used to examine the effect of PM 10 on HFMD incidence. In this regression, the association with PM 10 was specified with a standard distributed lag nonlinear model (DLNM) 25 where t indicates study days, Y t denotes the daily number of HFMD cases, and cb(pm 10 ) means the cross-basis function for the space of the predictor and the lag dimension of daily PM 10 . We selected a cross-basis composed of a natural cubic spline for the exposure response, and a natural cubic spline for the lag response. The Akaike Information Criterion was used to evaluate the choice of degrees of freedom (df) 25 . The lag period was extended to 14 days to investigate the potential lag associations based on the incubation period of HFMD infections (usually around 3-5 days) 24 and previous studies 11,12 . ns(weather variables) represents natural cubic spline fits to relative humidity and mean temperature, both with three df. ns(air pollutants) represents natural cubic spline fits to SO 2 and NO 2 , both with three df. Trend represents a variable of the calendar month and year, which is used to control for long-term trends and seasonal effects. Dow represents the day of the week. Holiday is a binary variable indicating whether day t was a public holiday. β 1 , β 2 , and β 3 represent the regression coefficients. Stratified analyses were conducted by gender (female and male) and pathogen group (CVA16 and EV71). We performed sensitivity analyses to evaluate the robustness of the results: changing the df (3-6) for climate factors and air pollutants. R packages, including "spline" and "dlnm", were adopted to create the model. Ethics approval. In 2008, HFMD was listed as a class C notifiable infectious disease. The present study was based on official HFMD surveillance data in Sichuan Province. No confidential information was included because analyses were performed at the aggregate level. The research protocol was approved by the institutional review board of the School of Public Health, Sichuan University. In the current study, all of the methods were conducted in accordance with the approved research protocol. Since all of the patients' records were anonymized and no individual information can be identified, informed consent was not required.