Article Text


Personal exposure assessment in the epidemiology of air pollutants
  1. R M Harrison1,
  2. R P Kinnersley1,
  3. R G Lawrence1,
  4. J G Ayres2,
  5. D Mark3
  1. 1Division of Environmental Health and Risk Management, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK; r.m.harrison.ipe{at}
  2. 2University of Aberdeen
  3. 3Health & Safety Laboratory

    Statistics from

    In commenting on our paper published recently in Occupational and Environnmental Medicine,1 Kromhout and van Tongeren admonish us for paying insufficient attention to the earlier literature on occupational pollutant exposures.2 While no doubt an element of their criticism is justified, we feel that the exposure situation for the general public is sufficiently different that it should not be assumed that findings in the occupational environment can necessarily be extrapolated to environmental exposures of the general public. A large component of environmental exposure arises from diffuse sources and may therefore be very spatially homogeneous at locations such as people’s homes which are often relatively remote from outdoor pollution sources.

    There has been some controversy in the literature regarding the extent to which measurements at fixed central urban background monitoring locations can reflect the exposures of large urban populations who spend much of their time indoors at locations relatively remote from the monitoring station.3,4 It has been typical to find that for an individual, daily personal exposures correlate with concentrations at the monitoring station, while if data are pooled from many individuals, the exposures appear to be uncorrelated with ambient air data.5,6 This finding suggests that the diffuse background as represented by the central urban monitor does account for a substantial proportion of variance in the exposure of an individual, and this conclusion is supportive of causality in the time series epidemiological studies, which would appear implausible if the monitoring data were unrelated to human exposures. The finding of our paper that microenvironment measurements do, in general, well represent individual personal exposures in that microenvironment (except for the personal cloud of PM10) is far from self-evident from much of the earlier literature and is a useful addition to knowledge. The fact that cigarette smokers were outliers in the regression analysis shows not unexpectedly that they generate strong local concentration gradients and would therefore need to be treated differently in any modelling of personal exposures. In the absence of such local sources of pollution, our study supports the concept that were sufficient microenvironment measurement data available, it would be perfectly feasible to model personal exposures with some degree of reliability.

    Kromhout and van Tongeren advocate the use of personal exposure measurements in environmental epidemiological studies. In doing so, they fail to acknowledge the magnitude of such studies. For example, in the large North American cohort studies, 8111 subjects were recruited in the Harvard Six Cities Study and over one million in the American Cancer Society Study. Were it possible to reconstruct the exposure environments of those individuals, even in a rather general way from time activity diaries, a considerable refinement would have been achieved. Even in panel studies, which typically recruit a far smaller number of individuals, the subjects are frequently drawn from susceptible groups and therefore not willing to be encumbered with troublesome and heavy sampling equipment. It must be remembered that concentrations in environmental samples are typically orders of magnitude lower than in occupational samples, therefore requiring higher flow rates (hence bigger pumps) and longer sampling intervals. In some instances it may be possible to use passive samplers such as diffusion tubes and badges, but these are typically rather imprecise and are available only for a very limited range of pollutants.

    In summary, therefore, while the measurement of personal exposure in environmental epidemiology is highly desirable, it is in reality very unlikely to be practicable in most studies. Thus, our study of the feasibility of reconstructing exposures from microenvironment data is well justified and has thrown useful light on the problem, for example, in illustrating the lower exposures of members of certain of our susceptible groups.


    View Abstract

    Request permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.