Patient characteristics
Sixty-four non-smoking panel members were recruited for repeated measurements of personal
exposure to PNC and parallel ECG recording. Table 1 describes the baseline characteristics of the 32 individuals with confirmed diagnosis
of type 2 diabetes and 32 individuals with IGT recruited based on the KORA F4 study
18],19]. No differences were observed between the type 2 diabetes patients and the individuals
with IGT concerning their age, gender, body mass index or disease history. Glycosylated
hemoglobin A1c (HbA1c) concentrations above 6.5% were more frequently observed in
individuals with diabetes than those with IGT. Diabetes prescriptions were taken by
more than half of the participants with diabetes and one participant with IGT. More
than 14,000 repeated 5-minute ECG measures and more than 1,200 1-hour ECG measures
were available (Table 1). Patients with diabetes had lower HR and HRV on a 5-minute basis. This different
was no longer apparent for HRV based on 1-hour ECG recordings.
Table 1. Description of study participants and 5-minute ECG measures
Personal exposures to particle number concentrations
Table 2 describes the distribution of the personal PNC measurements and the distribution
of particle concentrations at the central monitoring site. Substantially higher variation
in personal PNC was observed during personal monitoring compared to the background
level (Table 2). Figure 1 describes an example indicating that elevated levels of PNC may occur during times
spent in traffic, while indoor concentrations may be substantially lower in the absence
of indoor sources. Elevated personal PNC were observed when individuals spent time
in traffic (median?=?17,884Â cm?3, N?=?3,523), when cooking (median?=?43,612Â cm?3, N?=?285) or exposed to environmental tobacco smoke (ETS) (median?=?21,929Â cm?3, N?=?148). In contrast, personal PNC concentrations were lower during times spent
at home without cooking or ETS exposure (median?=?8,833Â cm?3, N?=?6,930). By design of the study, participants were commuting within the urban
area of Augsburg in the morning and midday hours. Thereby, personal exposures were
impacted by the morning rush-hour as well as by lower traffic volumes during midday
and were there deviating from concentrations measured at an urban background monitoring
site within the city center. Subject-specific Spearman correlation coefficients between
1-hour personal PNC concentrations and 1-hour ambient ultrafine particles (UFP) had
a median of 0.35 and ranged from ?0.60 at the 10th percentile to 0.90 at the 90th percentile. Personally measured PNC characterise the exposure to mobile source emissions
or other sources of freshly emitted particles and are determined by the personal activities
as well as meteorological influences in the region of Augsburg, Germany 20],21].
Table 2. Description of personal 5-minute particle measurements from 191 study visits and 1Â hour
of ambient particle measurements and meteorology recorded between March 2007 and December
2008
Figure 1. Example of personal measurements of PNC. Data was collected starting and ending at the KORA Study Center on November 27th 2007.
Ambient UFP were only moderately correlated with PM10 and PM2.5 measured at the same central monitoring site (spearman correlation coefficients of
0.49 and 0.42, respectively). In contrast, accumulation mode particles (ACP) were
highly correlated to 1-hour PM10, PM2.5 and UFP (Spearman correlation coefficients of 0.79, 0.75 and 0.70, respectively).
Changes in heart rate variability in response to particle exposure
Table 3 shows the associations between 5-minute personal exposures to PNC and HR and HRV
assessing concurrent and exposures lagged up to 15Â minutes. It shows a slightly delayed
response of HR and an immediate decrease in SDNN. Different responses of HR and SDNN
to PNC may be reasonable given the fact that correlation between HR and SDNN differed
substantially between individuals with a median Spearman correlation of ?0.10 and
a range between ?0.53 and 0.55.
Table 3. Associations between personal measurements of 5-minute average particle number concentrations
and 5-minute ECG-measures
Associations between PNC and SDNN appear to be more pronounced in individuals with
diabetes than in individuals with IGT (Figure 2). Exploratory analyses extending the time-lag between 5-minute personal exposure
to PNC and HR, SDNN or RMSSD up to one hour showed no consistent pattern beyond 15Â minutes.
Figure 2. Effects of personally measured 5-minute PNC on SDNN based on 5-minute ECG recordings
in patients with diabetes or impaired glucose tolerance. Effect estimates are shown for an increase of 16,000 particles cm?3.
We had previously shown associations between 1-hour ambient air pollution concentrations
and cardiac function occurring up to a lag of 4Â hours 17]. We had chosen one hour intervals of exposure and ECG recordings a priori as we considered this the minimal time scale for a central monitoring site in an
urban background location to represent population average exposures. In Table 4 we compare the association between 1-hour averages of personal PNC and ambient UFP,
ACP, PM10 and PM2.5 and concurrent measures of HR and HRV over 1-hour. No consistent associations between
personal or ambient particles number concentrations (PNC, UFP, ACP) and HR were observed.
In contrast, PM10 and PM2.5 were associated both with SDNN and RMSSD as reported previously 17]. The association between PM2.5 and HRV was stronger in individuals with IGT than those with type 2 diabetes, but
the differences did not achieve statistical significance. In line with our results,
Chan and colleagues observed significant decreases in SDNN and RMSSD in association
with an increase of 10,000 particles/cm3 in personally measured particles in the size range between 20Â nm and 1Â ?m in a prospective
panel study 22]. Adverse changes in HR and HRV were also observed in association with ambient UFP
in panel or cross-over studies 23]-28] and with concentrated UFP in controlled chamber studies 29],30] albeit some associations were not significant. However, some studies reported no
or even positive associations between HRV and UFP 31]-33].
Table 4. Associations between ambient 1–hour average air pollution concentrations at the central monitoring site and 1–hour average ECG-measures
Effect estimates were larger for the 1-hour PM2.5 than for personal PNC and associations between 1-hour PM2.5 concentrations and 5-minute HRV strengthened when adjusting for personal PNC (Figure 3). PM2.5 measured at an urban background monitoring site quantifies the overall particulate
matter level predominantly determined by the meteorological conditions. In the present
study, we demonstrate therefore that particle exposures determined by personal proximity
to sources and by urban background levels both are associated with changes in cardiac
function on a very immediate time scale.
Figure 3. Two pollutant models for 5-minute personal PNC and 1-hour ambient PM2.5on 5-minute HR and HRV parameters. in patients with diabetes or impaired glucose tolerance. Effect estimates are shown for an increase of 16,000 particles cm?3 and 12 ?g m?3 PM2.5.
Earlier studies have observed associations between hourly concentrations of PM2.5 and the onset of myocardial infarction in Boston, MA 5] and Rochester, NY 6]. Moreover, times spent in traffic were associated with the onset of myocardial infarction
7],8] and controlled exposure studies suggest that effects of diesel exposures might be
enhanced by exercise 34]. Previous studies have in many instances indicated that personal exposures to PM2.5 or to gaseous pollutants are associated with changes in HRV 26],35]-51]. The study participants ranged from healthy adults to patients with cardiovascular
diseases or asthma and were studied in different settings around the world. We had
chosen individuals with impaired glucose metabolism because individuals with type
2 diabetes had been shown to be susceptible to air pollution 2]-4]. A study of controlled human exposures to concentrated ultrafine particles showed
immediate effects on subjects with metabolic syndrome, however, did not observe changes
in HRV one hour after the exposure 30]. In contrast, in a study in subjects with type 2 diabetes indicated a decrease in
the high frequency component of heart rate variability and increased heart rates persisting
up to 48Â hours 16]. Furthermore, there is an emerging body of evidence linking ambient air quality as
one of the risk factors to type 2 diabetes 52]. Data from controlled animal experiments 53] as well as analyses in prospective population-based cohort studies 54]-58] support this association. Systemic inflammation, activation of innate immunity in
the lung and an imbalance of the autonomic nervous system induced by air pollution
exposures jointly potentially provide the link to insulin resistance and diabetes
exacerbation 52]. Sudden changes in cardiac function may predispose susceptible individuals to sudden
cardiac deaths during episodes with elevated particle concentrations 59]. Most likely, different underlying intrinsic mechanisms are activated by 5-minute
PNC and 1-hour PM2.5. We hypothesize that shortly elevated PNC may activate irritant receptors and lead
thereby to changes in the autonomic control 60]. In contrast, we hypothesize that the changes in HRV observed in association with
PM2.5 are associated with an activation of host defense on an alveolar level, which may
involve translocation of particle components, immediate systemic oxidative stress
response and an activation of leukocytes 52].
Sensitivity analyses
Associations were robust in sensitivity analyses and a summary is given in Figure 4 for the association between personally measured personal PNC and SDNN. No statistically
significant difference was observed in individuals without beta-blockers intake or
statin use. By selecting individuals with impaired glucose tolerance, we intended
to study the impact of particles in individuals who were not heavily treated by beta-blockers
or statins as these medications may obliterate the effects of particle exposures 61],62].
Figure 4. Sensitivity analyses of the association between concurrent exposure to personally
measured PNC and SDNN. *Regression coefficient as reported in Table 3.
Excluding time periods when the participants recorded ETS exposures or cooking rendered
consistent results, but suggested that indoor sources contributed to the observed
associations. We employed two different ways to adjust for physical activity. Neither
adjusting for the diary entries of physical activity nor for heart rate did change
the effect estimates. Models including personal noise exposure showed stronger associations
with personal PNC (Figure 3) and increased 5-minute SDNN (3.35% [95% CI: 2.95% ; 4.11%] per 5 db[A]) as reported
previously 63]. These analyses suggested that the associations of PNC and noise with ECG-parameters
were potentially confounding each other. To further test the impact of the model choices,
we conducted sensitivity analyses for the immediate effect of PNC on SDNN. Including
a time trend within the measurements or including the previous segments of SDNN as
a predictor did not change the effect estimates substantially (5-minute SDNN: ?0.56%
[?0.98%;-0.13%] or ?0.42% [?0.77%;-0.06%] per 16,000Â cm?3 PNC, respectively).
Limitations
The study assessed personal measurements of PNC which is a novel marker for personal
exposure to fresh combustion particles. The study thereby overcomes one large limitation
of previous panel studies. By employing direct measurements of PNC it also provides
different and novel information compared to studies of personal PM2.5 or gaseous pollutants 26],35]-49]. However, the measurement devices are usually operated by technical personnel to
measure indoor and outdoor particle concentrations and were not designed for study
participants. As a consequence we were only able to achieve 80% of the planned hourly
measurements albeit stringent examiner training, review of the instruction sessions
by audiotape, and written instructions for the participants. The missing measurements
had no certain pattern and were related to diligence in following the instructions
by the study participants. Diaries were kept by the participants, but no geographic
positioning system data was acquired. ECG data and personal PNC data were processed
independently. While the examiners and the participants were aware of the study hypotheses,
information on their HR was not available and levels of PNC were not discussed with
respect to limit or guideline values as these do not exist.
Timing of the measurements were based on recorded times from the instruments and the
study protocols. Discrepant times were checked individually, discussed with the study
nurses and corrected wherever possible.
Each day’s measurement provided control data for the individual and correlation within
the day and the individual was considered. Analyses proved to be relatively robust
against other assumptions of the covariance structure. Confounding by physical activity,
a potentially important individual time-varying factor was considered but did not
prove to be strong and resulted in changes of the effect estimates of less than 10%.
There were no statistically significant differences with respect to age, body mass
index, HbA1c concentrations, history of cardiovascular disease and medication use
when comparing the study participants to all individuals with either diabetes or IGT
in the underlying sample of the KORA cohort study. Participants of the panel study
were more likely to be unemployed, many of them already retired. In addition, the
proportion of ex-smokers was higher in the present study than in the overall sample.
As this study is assessing short-term impacts of urban area ambient particulate matter,
it does not address the question, whether long-term exposure to particulate matter
is associated with an increased risk for incident diabetes as recently shown 54]-58]. However, the data reported here provides evidence that short-term exposure to ambient
particulate matter may contribute to cardiovascular disease exacerbation in individuals
with impaired glucose metabolism or diabetes.