Atmospheric Environment 37 (2003) 993–1002 Longitudinal variability in outdoor, indoor, and personal PM 2.5 exposure in healthy non-smoking adults J.L. Adgate a, *, G. Ramachandran a , G.C. Pratt b , L.A. Waller c , K. Sexton a a Division of Environmental & Occupational Health, School of Public Health, University of Minnesota, Room 1260 Mayo, MMC 807, 420 Delaware St. SE, Minneapolis, MN 55455, USA b Minnesota Pollution Control Agency, 520 Lafayette Road, St. Paul, MN 55155, USA c Department of Biostatistics, Emory University Rollins School of Public Health, 1518 Clifton Rd, Atlanta, GA 30322, USA Received 11 July 2002; received in revised form 7 November 2002; accepted 7 November 2002 Abstract Multiple 24-h average outdoor, indoor and personal PM 2.5 measurements were made in a population of healthy non- smoking adults from the Minneapolis-St. Paul metropolitan area between April and November 1999. Personal (P) PM 2.5 concentrations were higher than indoor (I ) concentrations, which were higher than outdoor (O) concentrations. For 28 adults with a median of 9 (range 5–11) measurements per person, the distribution of longitudinal (i.e., within- person) correlation coefficients between P and I was moderate (median r ¼ 0:45). The distribution of longitudinal correlation coefficients between I and O concentrations showed that these variables were less strongly associated (median r ¼ 0:25; 28 residences; measurement median n=10 per residence, range 7–13), and the distribution of P and O correlation coefficients (median r ¼ 0:02; 29 subjects; measurement median n=11 per subject, range 7–15) showed little statistical relation between these two variables for a majority of participants. A sensitivity analysis indicated that correlations did not increase if days with exposure to environmental tobacco smoke or occupational exposures were excluded. On average these adults spent 91% of their time indoors, and the mean of the average PM 2.5 ‘‘personal cloud’’ was 15.3 mg/m 3 . Participants who had the largest personal cloud values tended to work outside the home and spent more time outdoors than subjects who did not work outside the home. In this population of healthy non-smoking adults, personal exposure to PM 2.5 does not correlate strongly with outdoor central site PM 2.5 concentrations. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: Ambient particulate matter; Healthy adults; Urban; Longitudinal correlation 1. Introduction Time-series epidemiological studies have shown a statistical association between mortality and morbidity and day-to-day variability in outdoor particulate matter o10 mm (PM 10 ) measured at central sites (Pope et al., 1992; Dockery et al., 1993; USEPA, 1996; Vedal, 1997; Samet et al., 2000). Scientists have hypothesized that fine particles o2.5 mm (PM 2.5 ) may be more closely linked with health effects, especially in children and adults with chronic lung disease. Statistical associations exist between PM 10 (and PM 10 in concert with other pollutants) and hospital admissions for chronic lung diseases for the elderly in the Minneapolis-St. Paul Metropolitan area (Schwartz, 1994; Moolgavkar et al., 1997), which has relatively low ambient PM 2.5 concen- trations compared to other major urban areas in the United States (Adgate et al., 2002). Epidemiological studies examining the statistical associations between PM exposure and health outcomes assume that central site outdoor monitors provide a reasonable estimate of personal exposure across the population. These studies hypothesize that particles of ambient origin penetrate indoors where people spend a *Corresponding author. Tel.: +1-612-624-2601; fax: +1- 612-626-0650. E-mail address: jadgate@umn.edu (J.L. Adgate). 1352-2310/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S1352-2310(02)00978-0