1.2. FACTORES ANTINUTRICIONALES DE LA QUINUA
1.2.1. SAPONINAS
Once the core model has been built up and potential confounding variables have been
adjusted for, the air pollution indicator is added to complete the statistical model. As
mentioned, Poisson regression, allowing for overdispersion and autocorrelation, is
used to determine relative risks associated with a change in each independent variable.
In the case o f the pollutants the risk is expressed as the percentage change (% change)
in number o f consultations for a given change in pollutant.
The analytical strategy was as follows. First, core models, one for each age group of
each of the five diagnostic categories, were determined. The diagnostic categories
were asthma, lower respiratory disease (LRD) excluding asthma, upper respiratory
disease (URD) excluding allergic rhinitis, allergic rhinitis and cardiovascular disease.
The core model contained all explanatory variables excluding air pollution indicators.
In turn, various measures and lags o f the air pollution variables were added to the
model for analysis. A predetermined set of daily pollution measures were tested in
turn in each statistical model: for SO2, NO2, BS, CO and PM 10 the 24 hour average
was used, and in addition for NO2 the maximum one hour measure was considered.
For O3, the maximum eight hour running average and maximum one hour measure
was used. For the sake of completeness, all pollutants were tested on all diagnostic
categories, even though an association was not expected to be found between certain
diseases and pollutants, e.g. allergic rhinitis and CO. Pollutant levels on the current
single-day lag 0, lag 1 etc.), as well as the averaged value o f lags 0 and lag 1, lags 0 to
2, and lags 0 to 3 (referred to here as ‘cumulative’ measures) were each used in the
model as indicators of air pollution levels. Each lag was tested separately and one at a
time. Where a pattern of increasing association with increase in lag was detected, lags
greater than 3 days were also investigated.
Possible differences between seasons in associations between pollutants and outcomes
were investigated using a dummy variable to indicate the season (warm season
defined as April to September, cold season as October to March). Where appropriate,
models with two pollutant measures were also examined to try and determine whether
one pollutant was more important than another. Where a strong single pollutant
association was found, exposure-response plots of the relative risk of consulting
against the level of pollutant were investigated to identify the nature of the
relationship.
All results of the single-pollutant, two-pollutant and seasonal analyses are expressed
as the percentage change in the number o f consultations associated with an increase in
the relevant pollution variable. The pollutant increases that were considered were a
1 0 unit increase in pollutant, to allow for comparisons between cities from other studies, and a percentile change for within-city comparisons. For example,
the association between consultations for asthma in the all ages group and SO2 levels
on the previous day was reported (using the non-parametric method) as a 2.5% (95%
Cl: -0.8 to 5.9) change in number of consultations should the 24 hour average levels
of SO2 on the previous day increase from 13 to 31 pg/m^, (these two values
the three year period). A reduction in consultation numbers associated with an
increase in a given measure o f pollution is reported as a negative change in risk.
These percentage changes in number of consultations are determined from the relative
risk using the following formula:
% change in number of consultations = (relative risk -1) * 1 0 0
The relative risk for a given change in the air pollution measure is determined from
the regression coefficient as follows:
relative risk = exp(regression coefficient * change in pollutant level) where exp is the
Table 3,1 : Summary o f pollutant measures and locations.
Pollutant Monitoring Sites Measurement technique Daily Measures
NO2
O3 SO2 BS COPMio
Bridge Place, Bloomsbury, West London
Bridge Place, Bloomsbury
City, Enfield, Ilford, Acton, Croydon
City, Enfield, Ilford, Acton, Croydon
Bridge Place, Bloomsbury, West London
Bloomsbury Chemiluminescence UV absorption Acid titration Reflectance method IR absorption Tapered element Oscillating microbalance 24 hr avg/max hr Max 8hr av/max hr 24 hr avg 24 hr avg 24 hr avg 24 hr avg 78
Chapter 4 : Results, Descriptive analyses
4.1 Presentation of results
The results are divided up into 4 separate chapters. This first results chapter presents
a description and summary statistics of the GP data and air pollution and
meteorological variables. The next chapter (Chapter 5) presents all non-parametric
regression analyses conducted and so contains the bulk of the results. Chapter 6
summarises the results obtained in the sensitivity analysis using the alternative
parametric methodology introduced in the methods. The final results chapter (Chapter
7) deals with a separate analysis of a thunderstorm episode that occurred in south-east
England in June o f 1994.
4.2 Descriptive Data
During the 3 year period analysed (1992-1994), the minimum monthly number of
registered patients was 268,718 and the maximum was 295,740. Of all patients
registered on the database, about 16% were children (aged 0-14 years), 70% were
adults (15-64 years) and 14% were in the elderly age group (65 years or over).
Table 4.1 shows the total numbers, by age group, of persons consulting in
participating practices in London between 1992 and 1994. This shows that the
Consultation rates for asthma and upper respiratory disease were particularly high in
children. As would be expected, few consultations for cardiovascular disease were
made by the children; and consultation numbers for asthma and allergic rhinitis in the
elderly age group were also very low.
Table 4.2 shows, by season, the mean, standard deviation (SD) and 10* and 90*
percentiles of the air pollution and meteorological variables and the daily number of
consultations by diagnostic category and age-group. As would be expected, the mean
number of consultations for all conditions except allergic rhinitis was higher in the
cool months (October - March). O3 levels were much higher in the warm period
(April - September).
Figures 4.1 to 4.5 show the daily counts of consultations for each disease (for the all
ages group only) between January 1992 and 31®^ December 1994. Time-series of
consultations by individual age-groupings followed similar patterns to the all-ages
group and so are not shown. The plots for asthma, LRD and URD suggest a rise in
the daily number of consultations in the winter months. A feature o f the asthma time
series was the dip around July and August each year; this was also evident for other
respiratory complaints. The decline appeared to be strongest in children and could, in
part, reflect the time of school summer holidays.
As would be expected with the allergic rhinitis series, visits were mainly restricted to
the spring and summer months. The numbers included two peaks each year, one
Visits for cardiovascular complaints suggested little yearly pattern, except for a slight
rise in the winter period. A slight decline over the 3 year period seemed to be in
evidence.
Figure 4.6 shows the average asthma consultation rate (per 10,000 patients) in the all
ages group by season and day of the week. Irrespective of season, Mondays were the
busiest days for asthma consultation and Thursdays the quietest weekday. The low
Sunday and Saturday rates reflect the unavailability of most practitioners at weekends.
The difference between Monday and the other weekdays was more pronounced in the
cool season.
Figures 4.7-4.12 show the daily measures of each pollutant variable between January
1992 and December 1994. Figures 4.13-4.14 show the mean temperature and relative
humidity measures over the same time period. These figures provide an indication of
the seasonal pattern, or lack of it, in the pollutants and meteorological measures as
well as the variability and range of values in London. Unsurprisingly, a summer peak
in O3 levels is in evidence. BS and CO are also clearly ‘seasonal’ pollutants, with
higher levels in the winter months. NO2 and CO levels were found to be high at the
end of December 1994, but no associated rise in GP consultation numbers was
observed.
Table 4.3 gives the Pearson correlation coefficients between the air pollution and
meteorological variables for the study period. Generally the correlation coefficients
between NO2, SO2, BS, CO and PMio were high and positive. This applied to both
with other pollutants, particularly in the cool season; in the warm season the
correlations were generally small and positive. Correlations between present day
pollution levels and levels on previous days were also investigated, but these were
generally found to be similar to the values presented in Table 4.3.
For each pollutant a single composite measure for London was calculated from data
from the available sites (see Table 3.1 in Methods Chapter for a list o f sites used for
each pollutant). Correlations between sites for NO2, O3 and CO were all positive and
high, at 0.7-0.96. However, correlations between the five sites providing SO2 and BS
data were smaller, in the range 0.5-0.8 for the five BS monitors, and 0.2-0.5 for four
of the five SO2 monitors, with the Enfield site not being in accordance with the other
Table 4.1 : Total numbers o f respiratory and cardiovascular consultations recorded in practices participating in the GPRD in London, 1992-1994, by age group.
Young (0-14 years) Adults (15-64 years) Elderly (65+ years) All ages
Complaint No. % No. % No. % No. %
Asthma 15,338 10.9 19,436 8.6 3,940 4.3 38,714 8.5
LRD exc asthma 43,495 31.0 80,882 35.9 45,384 49.8 169,761 37.2
URD exc allergic rhinitis 75,658 54.0 88,697 39.4 15,830 17.4 180,185 39.5 Allergic rhinitis 5,315 3.8 16,799 7.5 1,127 1.2 23,241 5.1 Cardiovascular disease 408 0.3 19,197 8.5 24,850 27.3 44,455 9.7 TOTAL 140,214 100 225,011 100 91,131 100 456,356 100 L R D ; L o w e r R e sp ir a to r y D is e a s e U R D : U p p e r R esp ir a to ry D is e a s e
Table 4.2 : Summary statistics, by season, for each diagnostic group (daily numbers of people consulting) and for air pollution and meteorological variables.
W ARM SE A SO N