EMERGENCY DEPARTMENT VISITS FOR MIGRAINE AND

ORIGINAL PAPERS
International Journal of Occupational Medicine and Environmental Health 2009;22(3):235 – 242
DOI 10.2478/v10001-009-0024-5
EMERGENCY DEPARTMENT VISITS FOR MIGRAINE
AND HEADACHE: A MULTI-CITY STUDY
MIECZYSŁAW SZYSZKOWICZ1, GILAAD G. KAPLAN2, ERIC GRAFSTEIN3, and BRIAN H. ROWE4
Health Canada, Ottawa, ON, Canada
Population Studies Division
2
University of Calgary, Calgary, AB, Canada
Departments of Medicine and Community Health Sciences
3
Providence Health Care and St. Paul’s Hospital, Vancouver, BC, Canada
Department of Emergency Medicine
4
University of Alberta, Edmonton, AB, Canada
Department of Emergency Medicine and School of Public Health
1
Abstract
Objectives: We set out to examine associations between ambient air pollution concentrations and emergency department (ED) visits for migraine/headache in a multi-city study. Materials and Methods: We designed a time-series study
of 64 839 ED visits for migraine (ICD-9: 346) and of 68 495 ED visits for headache (ICD-9: 784) recorded at hospitals
in five different cities in Canada. The data (days) were clustered according to the hierarchical structure (location, year,
month, day of week). The generalised linear mixed models technique was applied to fit the logarithm of clustered daily
counts of ED visits for migraine, and separately for headache, on the levels of air pollutants, after adjusting for meteorological conditions. The analysis was performed by sex (all, male, female) and for three different seasonal periods: whole
(January–December), warm (April–September), and cold (October–March). Results: For female ED visits for migraine,
positive associations were observed during the warm season for sulphur dioxide (SO2), and in the cold season for particulate matter (PM2.5) exposures lagged by 2-days. The percentage increase in daily visits was 4.0% (95% CI: 0.8–7.3) for SO2
mean level change of 4.6 ppb, and 4.6% (95% CI: 1.2,–8.1) for PM2.5 mean level change of 8.3 μg/m3. For male ED visits
for headache, the largest association was obtained during the warm season for nitrogen dioxide (NO2), which was 13.5%
(95% CI: 6.7–20.7) for same day exposure. Conclusions: Our findings support the associations between air pollutants and
the number of ED visits for headache.
Key words:
Air pollution, Migraine, Headache, Emergency department visit, Urban
INTRODUCTION
Migraine headache is an important cause of morbidity
in modern society. There are many self-reported triggers
including weather, [1] fatigue, stress, food, menstruation,
and infections [2]. Air quality in the home [3], office environment [4,5], and the occupational setting [6] may also
exacerbate headaches. A daily diary study of 32 headache
sufferers in Italy revealed that the severity and frequency of
headaches was related to days with higher concentrations
of carbon-monoxide and nitrogen dioxide [7]. Headaches
were more commonly reported from a neighbourhood with
a pulp mill, with higher sulphur dioxide (SO2) emissions,
compared to one without [8]. Moreover, large scale ED
studies have demonstrated an association between air
pollution and ED visits for migraine and all headaches in
Edmonton [9], Vancouver [10], and Ottawa [11,12]. The
associations were most consistent for particulate matter;
however, these results require confirmation and also further explorations using larger databases.
Received: July 8, 2009. Accepted: August 7, 2009.
Address reprint request to M. Szyszkowicz, Population Studies Division, Health Canada, 269 Laurier Avenue, Room 3-030, Ottawa, ON, K1A 0K9, Canada
(e-mail: [email protected]).
IJOMEH 2009;22(3)
235
ORIGINAL PAPERS
M. SZYSZKOWICZ ET AL.
The purpose of this study was to assess the relationship
between urban air pollution and ED visits for migraine
and headache in five cities in Canada. The study is based
on daily ED migraine and headache visits data. The study
takes into account ambient air pollution exposures. We
constructed models for single ambient air pollutant and
adjusted for temperature and relative humidity.
MATERIALS AND METHODS
This study was initiated and conducted at Health Canada
and it is the result of collaborative work with partners
from other research centers. The used data were provided
by 10 hospitals with emergency departments.
Study population
The study population consisted of the population serviced
by the hospitals in five cities in Canada: Edmonton, Halifax, Ottawa, Toronto, and Vancouver. The data for Halifax,
Ottawa and Vancouver were from a single hospital ED.
ED data for Edmonton were obtained from five area hospitals organised under the Capital Health regional health
authority. ED data from Toronto were obtained from two
hospitals (i.e. Toronto St. Michael and Sunnybrook Hospital), which were analysed separately because the Toronto
St. Michael Hospital is located in downtown, whereas
Sunnybrook is situated in the suburbs of Toronto.
ED visits were identified based on a discharge diagnosis of migraine using the International Classification for
Table 1. The number of days in study and number of ED visits
for migraine and headache by cities since start date of data
City
236
Start date
Days
Migraine Headache
Edmonton
April, 1992
3 652
56 241
48 022
Halifax
September, 1998
1 583
1 621
6 651
Ottawa
April, 1992
3 075
4 561
8 000
Toronto
July, 2000
639
461
1 884
Sunnybrook May, 1999
1 049
896
2 272
Vancouver
January, 1999
1 520
1 059
1 666
All cities
April, 1992
11 518
64 839
68 495
Warm
April, 1992
5 718
33 333
33 338
Cold
October, 1992
5 800
31 506
35 157
IJOMEH 2009;22(3)
Diseases 9th revision (ICD-9), rubric 346 [13]. In total,
the analysis is based on 64 839 ED visits for migraine
over 11 518 days for five cities. The diagnosed ED visits
for headache were retrieved by the ICD-9 code, rubric 784
and represented 68 495 ED visits (also over 11 518 days)
for five cities (Table 1).
Meteorological data
Environment Canada supplied hourly data for relative humidity, temperature (dry bulb) and atmospheric pressure
(sea level). We used the daily mean (average of 24 hourly
measurements) of these weather parameters. Because
recent findings have implicated weather fluctuations as
a major factor in migraine triggers we included these variables in our models as a confounder [14]. We incorporated temperature and relative humidity in form of natural
splines with three degrees of freedom.
Air pollution data
Hourly air pollution data were obtained from fixed monitoring stations in the considered cities. These data were
supplied by Environment Canada. For each air pollution
variable we have 24 measurements recorded at hourly
intervals. These include data on gaseous pollutants: SO2,
carbon monoxide (CO), nitrogen oxide (NO2), ozone (O3),
and particulate matter data: respirable particles PM2.5 and
inhalable PM10, particulate matter of median aerodynamic
diameter less than 2.5 and 10 microns, respectively. The
daily shared exposures of the population were expressed
as mean values among stations in city (Table 2).
Statistical methods
To relate short-term effects of air-pollution to the number
of daily ED visits for migraine we applied the generalised
linear mixed models (GLMM) methodology [15,16]. We
defined the clusters based on a 4-level nested structure
{location, year, month, day of week}. For example, one
constructed cluster groups all Mondays in April 1992 for
Edmonton’s data. We used Poisson models on the 4-level
hierarchical clusters with a random intercept and fixed
slope. We built models with one single pollutant, temperature and relative humidity. The values used in the models
EMERGENCY DEPARTMENT VISITS FOR MIGRAINE AND HEADACHE ORIGINAL PAPERS
Table 2. The average of pollutant concentrations and percentage of days with available data
City
CO
NO2
ppm
%
SO2
ppb
%
O3
ppb
%
PM10
PM2.5
ppb
%
μg/m
%
3
μg/m3
%
Edmonton
0.7
100
21.9
100
2.6
99
18.6
100
22.6
77
8.5
39
Halifax
0.5
65
17.5
63
10.0
69
22.1
69
0
0
9.8
33
Ottawa
0.9
100
18.8
100
3.9
99
17.5
99
20.1
12
6.5
31
Toronto
1.1
100
22.9
100
4.2
100
20.8
100
20.6
86
8.9
100
Sunnybrook
1.2
100
23.4
100
4.5
100
21.0
100
20.8
91
9.4
100
Vancouver
0.6
100
16.8
100
2.5
100
14.2
100
12.8
100
6.4
71
All cities
0.8
95
20.2
95
4.6
95
19.0
96
19.4
54
8.3
49
were lagged by 0, 1 and 2 days. For each lagged values
a separate model was constructed. Our analysis represents
the percentage value of changes in relative risk (%RR)
with 95% confidence intervals (CI) associated with an
increase in average of pollutant concentrations of the exposure to the study pollutants (SO2, NO2, CO, O3, PM10,
and PM2.5) after adjusting for temperature and relative
humidity. Estimates whose 95% CI do not cross 0 are considered significant. Additionally, we stratified our analysis by sex (all, male, female) and by season, defined as
warm (April to September) and cold (October to March)
seasons.
In addition, as a sensitivity analysis we applied case-crossover methodology [17] for ED visits for migraine (female,
warm season, exposure to sulfur dioxide) and for headache (male, warm season, exposure to nitrogen dioxide).
We only used data from Edmonton. We applied casecrossover analyses to compare measures of weather and
ambient air pollution on the day of presentation and control days for each patient. In case-crossover study, cases
serve as their own controls and therefore the design eliminates confounding by stable individual characteristics. We
selected control periods according to the time-stratified
approach [18]. In this approach exposure for cases is compared to exposure for controls on the three or four other
occurrences of that day of the week in a common calendar
month. We used conditional logistic regression models
to calculate adjusted odds ratios (OR) and 95% confidence intervals (CI) associated with an increment of one
interquartile range (IQR) in 24-hour mean levels of exposure; for NO2 we estimated IQR = 12.8 ppb, for SO2
we obtained IQR = 2.3 ppb. We defined 67 age groups
each of length 20 years. The groups were defined as follows: the first group was (0, 19), second (1, 20), and each
next group is constructed by shift up its preceding group
by one year. The last group was (66, 85). For such defined
sequence of age groups we calculated the corresponding ORs and 95% CIs.
RESULTS
Table 1 contains the number of ED visits for migraine
and headache by city. The number of days in study and
first date of study are also shown. During the warm season (5718 days), 33 333 ED visits for migraine were
made; 25 797 cases for females and 7475 for males. In
the cold season (5800 days), 31 506 visits for migraine
were made; 24 486 cases for females and 6933 for males.
Overall, 33 338 ED visits were made for headache in the
warm season; 18 650 cases were female and 14 448 were
males. In the cold season 35 157 ED visits were made; visits with 19 135 cases were female and 15 632 were males.
For some cases value for sex was unknown (missing), as
a consequence summary by sex are different than total.
Table 2 presents the mean values of the pollutant concentrations and the percentage of days with data. These
average values were used to calculate the excess risks for
the corresponding pollutants. The patterns of air-borne
pollutants show variability. For example, Halifax had the
lowest CO measurement, yet the highest O3 measurement.
Edmonton, on the other hand, had low CO measurements
yet high NO2 and the highest PM10 measurements. Some
IJOMEH 2009;22(3)
237
ORIGINAL PAPERS
M. SZYSZKOWICZ ET AL.
Fig. 1. The excess risk (%RR) by pollutants (lagged 0–2 days), sex, and period. The bolded symbols indicate positive results (p-value < 0.05).
other cities (Toronto, Ottawa) recorded high CO measures, while other pollutants were variable. Based on all
measurements, Vancouver tended to have the highest air
quality, while Toronto had the lowest.
Figure 1 shows the results for all considered pollutants
lagged by none (same day), 1 and 2 days. The results are
shown by sex and for three time periods: whole, warm and
238
IJOMEH 2009;22(3)
cold. The results for migraine are in the left part of Figure 1, and for headache in the right part of this figure.
The values of the percentage changes in the relative risks
(%RR) are shown. The positive and statistically significant
values are represented by bolded symbols (black). The
values were calculated for the mean values of the pollutant concentrations presented in Table 2. The statistically
EMERGENCY DEPARTMENT VISITS FOR MIGRAINE AND HEADACHE Fig. 2. The odds ratios (OR) and their 95% CIs for the
sequence of 67 age groups of length 20 years. The results for
the warm season: (top) female, migraine, SO2; (bottom) male,
headache, NO2.
ORIGINAL PAPERS
significant associations for migraine and for headache,
by pollutant, lagged values, sex and season are presented
in Table 3 and 4, respectively. For example, in the warm
season high concentration of SO2 was associated with increased presentation for migraine headaches in all individuals (%RR = 3.1; 95% CI: 0.3–6.1) and when stratified
by women (%RR = 4.0; 95% CI: 0.8–7.3). ED visits for
headaches were significantly increased on days with higher
concentrations of NO2, particularly during the warm season (%RR = 6.9; 95% CI: 2.8–11.4). In warmer months,
men presented to ED for headaches on days of high concentrations of NO2 (%RR=13.5%; 95% CI: 6.7–20.7).
Figure 2 summarises the results from the case-crossover
approach applied to the 67 age groups. The figure should
be recognised as a suggestive representation of the relations between pollutants and health outcomes (here, SO2
and migraine, and NO2 and headache) considered by age
groups. We can not order the relations among age groups
based on the obtained 95% CIs, but we have some suggestive image of the associations between pollutants
and ED visits [19]. For migraine the associations are growing with age of patients, and positive significant relations
are expected for the age range (30, 70) years. For males,
for ED visits for headache, positive significant results are
expected for the age range (0, 45) years.
Table 3. The percentage increases of the relative risk (%RR) and their 95% confidence intervals (95% CI) for ED visits for migraine
Pollutant
SO2
PM2.5
PM2.5
SO2
PM2.5
PM2.5
PM2.5
O3
PM2.5
PM2.5
SO2
PM10
PM2.5
PM2.5
PM2.5
Lagged
Same
Same
2-day
Same
Same
1-day
2-day
2-day
Same
2-day
Same
2-day
Same
1-day
2-day
Patient
All
All
All
All
All
All
All
Male
Female
Female
Female
Female
Female
Female
Female
Period
Whole
Whole
Whole
Warm
Cold
Cold
Cold
Cold
All
All
Warm
Cold
Cold
Cold
Cold
%RR
1.2
1.5
1.6
3.1
3.4
2.7
3.9
5.3
1.8
1.7
4.0
3.3
3.6
3.0
4.6
95% CI
–0.5, 2.9
–0.3, 3.3
–0.2, 3.4
0.3, 6.1
0.4, 6.5
–0.4, 5.7
0.9, 6.9
–1.5, 12.6
–0.1, 3.9
–0.3, 3.7
0.8, 7.3
0.4, 6.2
0.2, 7.1
–0.3, 6.5
1.2, 8.1
Significant
+
+
+
*
*
+
*
+
+
+
*
*
*
+
*
The symbol * marks values with p-value < 0.05.
The symbol + marks values with p-value in the interval [0.05, 0.1].
IJOMEH 2009;22(3)
239
ORIGINAL PAPERS
M. SZYSZKOWICZ ET AL.
Table 4. The percentage increases of the relative risk (%RR) and their 95% confidence intervals (95% CI) for ED visits for headache
Pollutant
Lagged
Patient
Period
%RR
95% CI
Significant
NO2
Same
All
Whole
3.9
1.7, 6.2
*
NO2
1-day
All
Whole
1.9
–0.4, 4.1
+
SO2
Same
All
Whole
1.2
–0.2, 2.6
+
PM10
Same
All
Whole
1.5
–0.3, 3.3
+
PM2.5
Same
All
Whole
3.0
1.4, 4.7
*
PM2.5
1-day
All
Whole
2.4
0.7, 4.1
*
CO
Same
All
Warm
4.4
–0.4, 9.3
+
CO
1-day
All
Warm
4.1
–0.7, 9.1
+
NO2
Same
All
Warm
6.9
PM10
Same
All
Warm
2.1
–0.3, 4.6
+
PM2.5
Same
All
Warm
2.7
0.5, 5.0
*
PM2.5
1-day
All
Warm
3.0
0.8, 5.2
*
PM2.5
Same
All
Cold
2.9
0.2, 5.6
*
CO
Same
Male
Whole
2.5
–0.6, 5.7
+
NO2
Same
Male
Whole
4.1
0.7, 7.6
*
PM10
Same
Male
Whole
2.8
0.1, 5.5
*
PM2.5
Same
Male
Whole
4.1
1.6, 6.6
*
PM2.5
1-day
Male
Whole
2.2
–0.3, 4.7
+
CO
Same
Male
Warm
11.0
3.4, 19.2
*
CO
1-day
Male
Warm
11.1
3.4, 19.5
*
CO
2-day
Male
Warm
7.5
–0.1, 15.6
+
NO2
Same
Male
Warm
13.5
6.7, 20.7
*
NO2
1-day
Male
Warm
7.2
0.8, 14.1
*
NO2
2-day
Male
Warm
5.2
–1.1, 11.9
+
PM10
Same
Male
Warm
5.1
1.3, 8,9
*
PM2.5
Same
Male
Warm
4.7
1.5, 8.0
*
PM2.5
1-day
Male
Warm
3.1
–0.2, 6.4
+
NO2
Same
Female
Whole
3.7
0.7, 6.8
*
SO2
Same
Female
Whole
1.7
–0.3, 3.7
+
PM2.5
Same
Female
Whole
2.2
0.1, 4.4
*
PM2.5
1-day
Female
Whole
2.5
0.4, 4.7
*
PM2.5
1-day
Female
Warm
2.6
–0.3, 5.5
+
PM2.5
Same
Female
Cold
3.7
0.2, 7.4
*
2.8, 11.4
*
The symbol * marks values with p-value < 0.05.
The symbol + marks values with p-value in the interval [0.05, 0.1].
240
DISCUSSION AND CONCLUSION
migraine headache associated with specific SO2 and PM2.5
This is amongst the largest study to link air quality conditions to ED presentation for headaches and migraines.
This study demonstrated an increase in daily visits for
level changes. For ED visits for headache, the largest sta-
IJOMEH 2009;22(3)
tistical significant association was obtained in the warm
period for same day NO2 exposure and CO in men. These
EMERGENCY DEPARTMENT VISITS FOR MIGRAINE AND HEADACHE findings add further evidence to the hypothesis that air
pollution exposure triggers migraines and headaches.
We had previously detected significant effects of sulphur
dioxide (SO2) and particulate matter (PM2.5) on ED visits
for migraine headaches [8]. The results for PM2.5 in the
present study were based only on 50% of days in study
with data. We detected significant effects of carbon monoxide, nitrogen dioxide, and particulates on ED visits for
headache. This result agrees with those reported in the
literature [7].
In addition, we verified the results by performing the analysis separately for some cities. We used the nested 3-level
hierarchical structure (year, month, day of week) to cluster the data records. We applied meta-analysis methodology to pool the results [20]. The obtained values (data not
shown here) were similar to the pooled results generated
on the 4-level clusters.
We can only speculate at the mechanisms by which air pollution exacerbates headaches and migraines. Neurogenic
inflammation, which can be triggered by air pollutants,
may result in headaches or migraines [21]. Alternatively,
air pollutants may impair endothelial-dependent vasodilation [22] leading to the development of migraines or
headaches. Future studies will be needed to understand
the pathogenesis of air pollution mediated headaches and
migraines.
There are several limitations of this study. First, fixedsite monitoring sites provide daily pollution exposures of
ambient air pollution and are applied to represent average population exposure. The included sites are all large
geographic areas and thus fixed site monitors will not fully
reflect variation in exposure between individuals. Second,
individual data on potentially important effect modifiers
such as medication use, socio-economic status, race and
co-morbidity were not available from this database. Third,
we have conducted numerous hypothesis tests, increasing the risk of false positive results; however, we have attempted to highlight those exhibiting greatest consistency
with other research. Fourth, pollutants exhibit associations
with one another to the extent that they originate from
common sources, making it difficult to singularly attribute
observed associations to individual pollutants. Fifth, many
ORIGINAL PAPERS
episodes of migraine and/or general headache do not result in an ED visit, thus our findings are not generalisable
to all such episodes.
Notwithstanding the above concerns there are also much
strength to the research including control of temporal
trends, which may have confounded the results such as
season, day of the week, measures of relative humidity,
and temperature. Moreover, these findings were corroborated in the first multicity study to evaluate the association
between acute air pollution exposure and the occurrence
of headaches or migraines. Future studies should evaluate the mechanisms through which air pollutants trigger
migraine and headaches.
ACKNOWLEDGEMENT
The authors express their appreciation to Health Canada
for securing these data and for funding data acquisition. Dr.
Rowe’s research has been supported by the 21st Century Canada Research Chair from the Government of Canada (Ottawa,
Ontario).
REFERENCES
1. Prince PB, Rapaport AM, Sheftell FD, Tepper SJ, Bi­
gal ME. The effect of weather on headache. Headache 2004;44:
596–602.
2. Chabriat H, Danchot J, Michel P, Joire JE, Henry P. Precipitating factors of headaches. A perspective study in a national
control-matched survey in migraineurs and non-migraineurs.
Headache 1999;39:335–8.
3. Farrow A, Taylor H, Northstone K, Golding J. Symptoms of
mothers and infants related to total volatile organic compounds
in household products. Arch Environ Health 2003;58:633–41.
4. Panzmolhave L, Kjaergaard SK. Effects on eyes and nose in
humans after experimental exposure to airborne office dust. Indoor Air 2000;10:237–45.
5. Wargocki P, Lagercrantz L, Witterseh T, Sundell J, Wyon DP,
Fanger PO. Subjective perceptions, symptom intensity and
performance: A comparison of two independent studies, both
changing similarly the pollution load in an office. Indoor
Air 2002;12:74–80.
IJOMEH 2009;22(3)
241
ORIGINAL PAPERS
M. SZYSZKOWICZ ET AL.
6. Schiffman SS, Studwell CE, Landerman LR, Berman K,
15. R Development Core Team. A language and environment for
Sundy JS. Symptomatic effects of exposure to diluted air sam-
statistical computing. R Foundation for Statistical Computing.
pled from a swine confinement atmosphere on healthy human
Vienna, Austria: R Development Core Team; 2008. Avail-
subjects. Environ Health Perspect 2005;113:567–76.
able from: URL: http://www.R-project.org.
7. Nattero G, Enrico A. Outdoor pollution and headache. Headache 1996;36:243–5.
8. Partti-Pellinen K, Marttila O, Vikka V, Jaakkola JJ, Jappinen P, Haahtela T. The South Karelia air pollution study:
Effects of low-level exposure to malodorous sulfur compounds
on symptoms. Arch Environ Health 1996;51:315–20.
9. Szyszkowicz M, Stieb D, Rowe B. Air pollution and daily
ED visits for headache in Edmonton, Canada. Am J Emerg
Med 2009;27:391–6.
10. Szyszkowicz M, Rowe B, Kaplan G. Ambient Sulphur Di-
examine the association between air pollution and health outcomes. Int J Occup Med Environ Health 2006;19(4):224–7.
DOI 10.2478/v1001-006-0032-7
17. Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epide­
miol 1991;133:144–53.
18. Janes H, Sheppard L, Lumley T. Case-crossover analyses of
air pollution exposure data: referent selection strategies and
their implications for bias. Epidemiology 2005;16:717–26.
oxide Exposure and Emergency Department Visits for Mi-
19. Gelman A, Stern H. The difference between “significant” and
graine in Vancouver, Canada. Int J Occup Med Environ
“not significant” is not itself statistically significant. The Amer-
Health 2009(1);22:1–6. DOI 10.2478/v1001-009-0006-7
ican Statistician 2006;60:328–31.
11. Villeneuve PJ, Szyszkowicz M, Stieb D, Bourque DA.
Weather and emergency room visits for migraine headaches in
Ottawa, Canada. Headache 2006;46:64–72.
12. Szyszkowicz M. Ambient Air Pollution and Daily Emergency
Department Visits for Headache in Ottawa, Canada. Headache 2008;48:1076–81.
13. World Health Organization. The International Classification
of Diseases, 9th Revision. Geneva: WHO; 1997.
14. Mukamal KJ, Wellenius GA, Suh HH, Mittleman MA.
Weather and air pollution as triggers of severe headaches.
Neurology 2009;72:922–7.
242
16. Szyszkowicz M. Use of generalized linear mixed models to
IJOMEH 2009;22(3)
20. DerSimonian R, Laird N. Meta-analysis in clinical trials.
Control Clin Trials 1986;7:177–88.
21. Meggs WJ. Neurogenic inflammation and sensitivity to environmental chemicals. Environ Health Perspect 1993;103(3):
234–8.
22. Tornqvist H, Mills NL, Gonzalez M, Miller MR, Robinson SD, Megson IL, et al. Persistent endothelial dysfunction in
humans after diesel exhaust inhalation. Am J Respir Crit Care
Med 2007;176:395–400.