In our study we observed a significant association between PM10
and mortality pre-2000 being obscured by an insignificant association
post-2000. This suggest that in future time-series studies of
the effect of ambient air pollution on mortality that it may be
prudent to allow for estimated pollutant effects to change over
time. Allowing for such changes would help avoid spurious results
of no significant pollutant effect being found when it fact the underlying
relationship is, for example, one of a significant pollutant
effect for a period of time followed by an insignificant pollutant
effect. The need to allow for a changing pollution effect is perhaps
becoming more important as the length of data available to conduct
time-series investigations is continually increasing.