Medical knowledge | Studies, essays, thesises » Particles Size and Composition in Mediterranean Countries, Geographical Variability and Short-term Health Effects

Source: http://www.doksinet MED-PARTICLES Project 2011-2013 Under the Grant Agreement EU LIFE+ ENV/IT/327 Particles size and composition in Mediterranean countries: geographical variability and short-term health effects Particles size and composition in Mediterranean countries: geographical variability and short-term health effects MED-PARTICLES ACTION 10. Health effects of PM10, coarse particles (PM2.5-10) and fine particles (PM25) on daily cause-specific mortality: city-specific results and meta-analysis Summary: Protocol to define the standardized methodological steps to investigate the association between mortality and PM concentrations, with regards to different causes, age groups and different PM fractions. -------------------------------------------------------------------------- 1 Source: http://www.doksinet BACKGROUND The knowledge on the effects of air pollution on human health have been growing in last decades as the result of the enormous scientific effort to

design and conduct large epidemiological studies. At present, we rely on a growing consolidated knowledge that airborne particles with a diameter < 10 μm is the most important airborne pollutant associated with the short-term effects on human health, while fine particles (PM2.5) is the fraction responsible for the most severe health effect There are several additional aspects that we understand in a more comprehensive way. We know that coarse fraction (PM 2.5-10μm) has a predominantly natural origin while the fine fraction is produced by combustion (vehicles, industries and electric power plants). We know that traffic and heating are the most relevant sources in determining the increases of mortality and morbidity due to respiratory and heart diseases. We understand the shape of the concentration-response function and the existence of no-effects threshold between concentration of airborne particles and health effects. We know that particles are able to increase not only

exacerbations of diseases, but also mortality as well as the onset of respiratory and cardiovascular diseases. We understand that the oxidative stress is the most important mechanism of damage at cellular level, but inflammation is not the only consequence, since increasing blood coagulability, heart autonomic functions, and the atherosclerosis processes are also influenced by air pollution. Notwithstanding the growing of epidemiological research, scanty data are available in the European Mediterranean countries, especially on PM2.5 and PM25-10, so that multi-center studies on the shortterm health effects of fine and coarse fractions on mortality and morbidity are virtually non-existent Objective The objective of this protocol is to define the methodological steps to investigate the association between mortality outcomes and PM concentrations, with regards to: 1) different causes of mortality; 2) different PM fractions; 3) different modeling choices of confounding adjustment. The

analyses will be conducted within each city participating the Med-Particles project, and then pooled estimates will be derived with random-effects meta-analytical procedures. Study population and outcomes Data on daily deaths will be collected for at least 10 cities from Italy, Greece, Spain and France. For each city and for each day of the study period, data on daily deaths will be collected with regards to the population resident of the city that died in the city. For each city, daily counts of deaths will be available, for the following causes. We distinguish two groups of outcomes: primary outcomes, on which all analyses will be conducted, and secondary outcomes, in italics, on which only selected analyses will be carried out: 2 Source: http://www.doksinet Cause of death ICD-9 code Natural causes 001-799 A00 – R99 Diabetes 250 E10 – E14 Cardiovascular diseases 390-459 I00 – I99 390-429 I00 – I52 Acute coronary events 410-411 I21 – I23 Conduction

disorders 426 I44 – I45 Arrhythmias 427 I46 – I49 Heart failure 428 I50 430-437 I60 – I68 460-519 J00 – J99 466, 480-487 J09 – J18, J20 – J22 Acute bronchitis, bronchiolitis, unsp. LRTI 466 J20 – J22 Pneumonia 480-486 J12 – J18 Influenza 487 J09 – J11 490-492,494,496 J40 - J44, J47 Cardiac diseases Cerebrovascular diseases Respiratory diseases Low respiratory tract infections (LRTI) Chronic obstructive pulmonary disease (COPD) ICD-10 code Further details are reported in the protocol for data collection on health endpoints, Action 6. For the primary mortality outcomes the epidemiological analysis will be implemented for all ages as well stratified for deaths among those above and below 75 years of age for deaths from natural causes and cardiovascular diseases and for respiratory deaths among those above 75 years. For all secondary mortality outcomes the analyses will be done only for all ages except for COPD that will be done only for the

age group above 65 years of age. For the investigation of effect modification by sex we will analyze only deaths from natural causes, for the age groups outlined above. Environmental variables For each city, daily average exposures to PM 2.5 , PM 10 and PM 25-10 will be derived from monitor-specific data, as detailed in the protocol for data pooling, Action 7. In addition, data are available for each city and day on other air pollutants (NO2, O3, CO, O3) and meteorological parameters (temperature, humidity, barometric pressure). Other confounders 3 Source: http://www.doksinet Data have been collected on other daily variables, useful for confounding adjustment in the PMmortality association. They include influenza epidemics and indicator variables relating to population dynamics during holidays and summer. Further details on environmental variables and other confounders are reported in the protocol for environmental data collection, Action 3. Methods a) Methodological framework The

PM-mortality association will be investigated using Poisson regression models allowing for overdispersion (using quasi(poisson) in R). The model is of the form: log E [Yt]=β 0 + b * PM t + confounders where E[Y t ] is the expected value of the Poisson distributed variable Y t indicating the daily count of unscheduled hospital admissions for a specific disease on day t, with Var(Y t )=φE[Y t ], φ being the overdispersion parameter, PM t is the concentration of particles (for a specific fraction) on day t, and “confounders” include an extensive list of confounding factors, as described below. b, the parameter of interest, is the adjusted log(relative risk) of hospitalization for a unit increase in PM. b) Confounder list The confounders are chosen a priori based on past knowledge and preliminary exploratory analyses. They include the following (see the protocol of statistical analysis, Action 9, for further details): - Time-trend. In order to take into account long-term as well

as seasonal time-trend, a penalized regression spline of day of the death will be fitted. Natural cubic splines will be used as basis functions for the penalized regression splines. We will choose 8 effective degrees of freedom (edfs) per calendar year of available data to control for seasonality. Two sensitivity analyses will be performed: 1) one model will use the time series approach with penalised splines for seasonality control (as the main model) but with the final edfs for time trend based on the choice of the smoothing parameter for time that minimizes the absolute value of the sum of the partial autocorrelations (PACF) of the residuals from lags 1 to 30; 2) a three-way interaction between year, month and day of the week will be fitted as alternative to the spline term + day of the week adjustment, since this approach has been shown to be equal to the case-crossover design, with “time-stratified” strategy for controls selection; 4 Source: http://www.doksinet - Air

temperature. It will be adjusted for with two natural splines, in order to control for both high and low temperatures at different reference lags: lag 0 for high summer temperatures, lag 1 or lag 1-6 for low winter temperatures. The choice between lag 1 and lag 1-6 for cold temperatures adjustment will be based at the city level on the lag which minimizes AIC. For both natural splines, 3 degrees of freedom will be fixed a priori at quartiles of the city-specific distribution. - Relative humidity. It will be adjusted for with a linear term at lag 0 - Influenza epidemics. The data on influenza should preferably be daily counts ie number of cases When there are weekly or biweekly data on number of cases then daily values should be calculated by division. If the only available information is the existence of an epidemic, then a variable taking the value 1 for epidemic days and 0 otherwise should be used. In case there are no influenza data available we will use the APHEA-2 method for

influenza control, including a dummy variable taking the value of 1 when the 7-day moving average of the respiratory mortality was greater than the 90-th percentile of its city-specific distribution. In this case, since influenza definition is based on the distribution of respiratory mortality, we will include the influenza dummy variable only when analyzing non respiratory mortality. In all other influenza definitions, instead, it will be included for all study outcomes, including respiratory ones. - Population dynamics during holidays and summer. This phenomenon will be adjusted for with indicator variables for holidays and summer population decreases, as described in the protocol for environmental data collection, Action 3. c) Lag structure of exposure According to the EPIAIR protocol, for each primary outcome the following three steps will be implemented: a) cubic polynomial distributed lag models and single-lag models from lag 0 to lag 6 to visually examine the lag structure of

the association between PM exposures and health outcomes; b) three cumulative lags chosen a priori to represent immediate effects (lag 0-1), delayed effects (lag 2-5) and prolonged effects (lag 0-5); c) for each combination of exposure/outcome, choice of one of these three lags as the reference lag, based on the meta-analytic polynomial distributed lag shape, the metaanalytic estimates of the a priori cumulative lags, and the heterogeneity among city-specific estimates. This choice is relevant to identify the lagged exposure to be used for additional analyses as bi-pollutant models, sensitivity analyses on time-trend and temperature adjustment, and analysis of the secondary outcomes. In particular, for each secondary outcome the referent lag from the primary outcome of the same group will be applied (for example, if the relevant lag for PM2.5 and cardivascular mortality is 0-1, the same lag is applied to the secondary outcomes “Acute coronary events”, “Conduction disorders or

arrhythmias” and “Heart failure”). d) Two-pollutant models 5 Source: http://www.doksinet For each primary outcome, and for PM2.5 and PM25-10, we will run two-pollutant models, including NO2 and ozone in turn. Special care will be given to the inclusion of NO2 in PM25 models if there is high correlation (r>0.7) In addition, two pollutant models will be run with PM25 and PM25-10 as well. Concerning the lag chosen for two-pollutant models, the two pollutants will be put in the model at the same lag, equal to the relevant lag for the PM exposure. For example, if from the analysis of the lag structure there is evidence that the most relevant lag for PM2.5 and cardiac mortality is 0-1, the twopollutant models for PM25 will be run by adding both PM25 and the other pollutant (NO2, O3 or PM2.5-10) in the model, both at lag 0-1 e) Meta-analytical techniques Once city-specific effect estimates are obtained, they will be meta-analyzed with random-effects models. For each

meta-analytical estimate, a test for heterogeneity will be performed and corresponding p-value reported, together with the I2 statistic, which represents the proportion of total variation in effect estimates that is due to between-cities heterogeneity. Conclusions: The protocol defines the methodological steps to investigate the association between mortality outcomes and PM concentrations, with regards to different causes of mortality, different PM fractions and different modeling choices of confounding adjustment. Mortality data of resident people that died in the city will be collected for 10 cities form Italy, Greece, Spain and France The analyses are conducted within each city participating the Med-Particles project, and then pooled estimates are derived with random-effects meta-analytical procedures. 6