You are on page 1of 2

Perspectives | Correspondence All EHP content is accessible to individuals with disabilities.

Fully accessible
(Section 508compliant) HTML versions of these articles are available at
http://dx.doi.org/10.1289/ehp.1307403 and http://dx.doi.org/10.1289/ehp.1307403 R .

The correspondence section is a public forum and, as such, is not peer-reviewed. EHP is not responsible Kawada notes that the small number
for the accuracy, currency, or reliability of personal opinion expressed herein; it is the sole responsibility of of particip ants with low anklebrachial
the authors. EHP neither endorses nor disputes their published commentary.
index (ABI) provided low power to detect
an association; this is a limitation that we
acknowledged in our article (Rivera etal.
The Association between (PM2.5), and sulfur dioxide (SO2) as indica- 2013). Kawada also points out that the num-
Air Pollution and Subclinical tors of air pollution, and they used pulse ber of covariates in the model exceeds the
Atherosclerosis wave velocity and augmentation index in common rule of thumb of having at least
http://dx.doi.org/10.1289/ehp.1307403
addition to IMT as indicators of vascular 10events per variable. We recognize that,
damage. Traffic intensity and proximity of with a small number of cases, one is inevita-
Rivera etal. (2013) investigated the associa residence to roads were also used as indica- bly faced with the trade-off between includ-
tion between air pollution and subclinical tors of air pollution. Lenters etal. found ing all relevant confounders and keeping
atherosclerosis by using carotid intima media significant associations only between NO2 the number of covariates to a minimum.
thickness (IMT), anklebrachial index and pulse wave velocity and augmentation However, in our article we provided results
(ABI), and several indicators of air pollution. index and between SO 2 and pulse wave from a minimally adjusted model includ-
As a main outcome, air pollution was posi- velocity. Because contradictory results for ing only five confounders and an interaction
tively associated with an ABI of >1.3, and this association have been reported, further term. In this model, the number of events
also with changes in IMT. In contrast, they longitudinal studies are needed to assess this per variable was >10, and the results were
observed no significant association between association. not significantly different from those of
air pollution and an ABI of <0.9. The author declares he has no actual or potential the fully adjusted model. Kawada suggests
I have some concerns about their study competing financial interests. selecting a higher cut-off value of ABI, such
(Rivera etal. 2013). First, the study included Tomoyuki Kawada as 1.0 for low ABI given that an ABI of
a small number of subjects with ABIs <0.9 Department of Hygiene and Public Health, Nippon
0.91.0 is also associated with cardiovascular
and >1.3 (56 and 116, respectively). The Medical School, Tokyo, Japan risk (Ono etal. 2003). The findings of Ono
authors used multinomial logistic regres- E-mail: kawada@nms.ac.jp etal. were for patients on hemodialysis due
sion analysis; for the full-adjustment model, to end-stage renal disease and thus cannot be
>16air pollution variables were used. There References extrapolated to healthy population samples
is a limitation in the number of indepen- such as the one considered in our study (i.e.,
dent variables appropriate for multiple logis- Bauer M, Moebus S, Mhlenkamp S, Dragano N, with no history or current signs of cardio
NonnemacherM, Fuchsluger M, etal. 2010. Urban particu
tic regression analysis (Novikov etal. 2010; late matter air pollution is associated with subclinical vascular disease). We selected a cut-off value
Peduzzi etal. 1996), and enough events atherosclerosis: results from the HNR (Heinz Nixdorf of 0.9 for low ABI because of the strong evi-
should be included to maintain statistical Recall) study. J Am Coll Cardiol 56:18031808. dence of increased risks of incident cardio
Diez Roux AV, Auchincloss AH, Franklin TG, RaghunathanT,
power for multivariate analysis. According to Barr RG, Kaufman J, etal. 2008. Long-term exposure vascular disease, morbidity, and mortality
the criteria that at least 10events per variable to ambient particulate matter and prevalence of sub- in individuals with ABI <0.9 (Allison etal.
are required to keep stable estimates (Peduzzi clinical atherosclerosis in the Multi-Ethnic Study of 2008; Ankle Brachial Index Collaboration
Atherosclerosis. Am J Epidemiol 167:667675.
etal. 1996), Rivera etal. (2013) needed Lenters V, Uiterwaal CS, Beelen R, Bots ML, Fischer P,
2008; Lee etal. 2004; McDermott etal.
170 events with an ABI <0.9 or >1.3 for BrunekreefB, etal. 2010. Long-term exposure to air pollu- 2005). A cut-off value of 0.9 is also more
their analysis. tion and vascular damage in young adults. Epidemiology specific (Lee etal. 2004) and much more
21:512520.
Second, Rivera etal. (2013) used systolic Novikov I, Fund N, Freedman LS. 2010. A modified approach
common in the literature, which allows
and diastolic blood pressure to adjust for the to estimating sample size for simple logistic regression comparison with other studies.
relationship between air pollution and indi- with one continuous covariate. Stat Med 29:97107. Kawadas second argument involves the
Ono K, Tsuchida A, Kawai H, Matsuo H, Wakamatsu R,
cators of atherosclerosis. But multicollinearity MaezawaA, etal. 2003. Ankle-brachial blood pressure
consideration of multicollinearity in the fully
among independent variables should have index predicts all-cause and cardiovascular mortality in adjusted model (model2), in which systolic
been considered in the analysis (York 2012). hemodialysis patients. J Am Soc Nephrol 14:15911598. and diastolic blood pressure were included.
Finally, Rivera etal. (2013) could not Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. Systolic and diastolic blood pressure were
1996. A simulation study of the number of events per
clarify the lack of association between ABI variable in logistic regression analysis. J Clin Epidemiol only moderately correlated (correlation coef-
<0.9 and indicators of air pollution. An ABI 49:13731379. ficient, 0.62), and according to the variance
of 0.91.0 is also associated with cardio Rivera M, Basagaa X, Aguilera I, Foraster M, Agis D, inflation factor (VIF), there were no multi
deGrootE, etal. 2013. Association between long-term
vascular risk (Ono etal. 2003). Therefore, exposure to traffic-related air pollution and subclinical collinearity problems (VIF was 2.45 for sys-
the association between air pollution and sub atherosclerosis: the REGICOR study. Environ Health tolic blood pressure and 1.98 for diastolic
clinical atherosclerosis should be evaluated by Perspect 121:223230; doi:10.1289/ ehp.1205146. blood pressure).
York R. 2012. Residualization is not the answer: rethinking how
selecting a higher cut-off value of ABI, such as to address multicollinearity. Soc Sci Res 41:13791386.
Kawada points at contradictory results
1.0. This procedure will increase the number in the association between air pollution
of events for multivariate analysis. and carotid intima media thickness (IMT),
Other researchers have reported a signifi- The Association between mainly based on the null and weak associa-
cant association between air pollution and Air Pollution and Subclinical tions found by Lenters etal. (2010) between
IMT (Bauer etal. 2010; DiezRoux etal. several markers of air pollution [nitrogen
2008). Lenters etal. (2010) also examined
Atherosclerosis: Rivera etal. dioxide, sulfur dioxide, PM2.5 (particulate
the association between air pollutants and Respond matter 2.5m in aerodynamic diameter),
indicators of vascular damage but observed http://dx.doi.org/10.1289/ehp.1307403R black smoke, and traffic intensity] and three
no association between air pollution and We appreciate Kawadas interest in our study indicators of vascular damage (IMT, pulse
IMT. Lenters etal. (2010) used nitrogen (Rivera etal. 2013). His letter focuses on wave velocity, and augmentation index). We
dioxide (NO 2), black smoke, particulate some limitations of our study, which were consider, however, that these do not con-
matter 2.5m in aerodynamic diameter already discussed in our article. stitute results contradictory to the positive

A8 volume 122 | number 1 | January 2014 Environmental Health Perspectives


Correspondence

association between air pollution and sub Salud Pblica (CIBERESP), Spain; 5Universitat Pompeu particulate matter air pollution and systemic inflammatory
clinical markers of atherosclerosis found in Fabra, Barcelona, Spain; 6Department of Vascular markers. Environ Health Perspect 117:13021308;
Medicine, Academic Medical Centre, Amsterdam, doi:10.1289/ehp.0800362.
the six studies thoroughly discussed in our the Netherlands; 7Swiss Tropical and Public Health Hoffmann B, Moebus S, Mhlenkamp S, Stang A, LehmannN,
article (Bauer etal. 2010; Diez Roux etal. Institute, Basel, Switzerland; 8University of Basel, DraganoN, et al. 2007. Residential exposure to traffic is
2008; Hoffmann etal. 2007, 2009; Knzli Basel, Switzerland; 9University ofNorth Carolina associated with coronary atherosclerosis. Circulation
116:489496.
etal. 2005, 2010) as well as by Wilker etal. atChapel Hill, Chapel Hill, North Carolina, USA;
10 4sfera Innova, Girona, Spain; 11Research Unit, Knzli N, Jerrett M, Garcia-Esteban R, Basagaa X,
(2013). The study by Lenters etal. (2010) BeckermannB, Gilliland F, et al. 2010. Ambient air pollution
Family Medicine, Jordi Gol Institute for Primary
involved a cohort of young adults, on aver- Care Research (IDIAP Jordi Gol), Catalan Institute of
and the progression of atherosclerosis in adults. PLoS ONE
5(2):e9096; doi:10.1371/journal.pone.0009096.
age 28years of age. Exposure to air pollution Health, Catalunya, Spain; 12Department of Medical Knzli N, Jerrett M, Mack WJ, Beckerman B, LaBree L,
was estimated at the current address only. Sciences, School of Medicine, University of Girona, GillilandF, et al. 2005. Ambient air pollution and athero
As the authors acknowledged, the young Girona, Spain; 13Servicio de Cardiologa, Hospital sclerosis in Los Angeles. Environ Health Perspect
Universitari Josep Trueta, Institut Catal de la Salut, 113:201206; doi:10.1289/ehp.7523.
age of participants and the exposure mis Girona, Spain; 14Grupo de Epidemiologa y Gentica Lee AJ, Price JF, Russell MJ, Smith FB, Van Wijk MCW,
classification, which resulted from exposure Cardiovascular, Hospital del Mar Research Institute FowkesFGR. 2004. Improved prediction of fatal myo
estimated at the current address only, are (IMIM), Barcelona, Spain cardial infarction using the ankle brachial index in addi-
E-mail: marcelaepi@gmail.com tion to conventional risk factors: the Edinburgh Artery
likely explanations for their mixed results. Study. Circulation 110:30753080.
Finally, we agree with Kawadas closing Lenters V, Uiterwaal CS, Beelen R, Bots ML, Fischer P,
remark on the need for longitudinal studies, References BrunekreefB, etal. 2010. Long-term exposure to air pol-
as we concluded in our article. lution and vascular damage in young adults. Epidemiology
Allison MA, Hiatt WR, Hirsch AT, Coll JR, Criqui MH. 2008. A 21:512520
The authors declare they have no actual or high ankle-brachial index is associated with increased McDermott MM, Liu K, Criqui MH, Ruth K, Goff D, SaadMF,
potential competing financial interests. cardiovascular disease morbidity and lower quality of etal. 2005. Ankle-brachial index and subclinical car-
life. JAm Coll Cardiol 51:12921298. diac and carotid disease: the Multi-Ethnic Study of
MarcelaRivera,1 XavierBasagaa,2,3 Ankle Brachial Index Collaboration. 2008. Ankle brachial Atherosclerosis. Am J Epidemiol 162:3341.
InmaculadaAguilera,2,3,4 index combined with Framingham Risk Score to predict Ono K, Tsuchida A, Kawai H, Matsuo H, Wakamatsu R,
MariaForaster,2,3,4,5 DavidAgis,2,3,4 cardiovascular events and mortality. JAMA 300:197208. MaezawaA, etal. 2003. Ankle-brachial blood pressure
Bauer M, Moebus S, Mhlenkamp S, Dragano N, index predicts all-cause and cardiovascular mortality in
EricdeGroot,6 LauraPerez,7,8
NonnemacherM, Fuchsluger M, etal. 2010. Urban particu hemodialysis patients. J Am Soc Nephrol 14:15911598.
MichelleA.Mendez,9 LauraBouso,2,3,4 late matter air pollution is associated with subclinical Rivera M, Basagaa X, Aguilera I, Foraster M, Agis D,
JaumeTarga,10 RafaelRamos,11,12 atherosclerosis: results from the HNR (Heinz Nixdorf deGrootE, etal. 2013. Association between long-term
JoanSala,13 JaumeMarrugat,14 Recall) study. J Am Coll Cardiol 56:18031808. exposure to traffic-related air pollution and subclinical
RobertoElosua,4,14 and NinoKnzli 7,8 Diez Roux AV, Auchincloss AH, Franklin TG, RaghunathanT, atheros clerosis: the REGICOR study. Environ Health
1Universityof Montreal Hospital Research Centre Barr RG, Kaufman J, etal. 2008. Long-term exposure Perspect 121:223230; doi:10.1289/ ehp.1205146.
to ambient particulate matter and prevalence of sub- Wilker EH, Mittleman MA, Coull BA, Gryparis A, Bots ML,
(CRCHUM), Montreal, Quebec, Canada; 2Centre for
clinical atherosclerosis in the Multi-Ethnic Study of SchwartzJ, etal. 2013. Long-term exposure to black car-
Research in Environmental Epidemiology (CREAL),
Atherosclerosis. Am J Epidemiol 167:667675. bon and carotid intima-media thickness: the Normative
Barcelona, Spain; 3Hospital del Mar Research Institute Hoffmann B, Moebus S, Dragano N, Stang A, MhlenkampS, Aging Study. Environ Health Perspect 121:10611067;
(IMIM), Barcelona, Spain; 4CIBER Epidemiologa y SchmermundA, et al. 2009. Chronic residential exposure to doi:10.1289/ehp.1104845.

Environmental Health Perspectives volume 122 | number 1 | January 2014 A9