Professional Documents
Culture Documents
Panic Anxiety, Under The Weather?: Originalarticle
Panic Anxiety, Under The Weather?: Originalarticle
Panic Anxiety, Under The Weather?: Originalarticle
DOI 10.1007/s00484-004-0236-0
ORIGINAL ARTICLE
Received: 6 April 2004 / Revised: 8 September 2004 / Accepted: 30 September 2004 / Published online: 18 November 2004
ISB 2004
Abstract The relationship between weather conditions were three times more common with the poniente wind
and psychiatric disorders has been a continuous subject of (hot wind), twice less often with rainfall, and one and a
speculation due to contradictory findings. This study at- half times more common in autumn than in other seasons.
tempts to further clarify this relationship by focussing on These three trends (hot wind, rainfall and autumn) were
specific conditions such as panic attacks and non-panic accumulative for panic episodes in a logistic regression
anxiety in relation to specific meteorological variables. formula. Significant reduction of episodes on weekends
All psychiatric emergencies attended at a general hospital was found only for non-panic episodes. Panic attacks,
in Barcelona (Spain) during 2002 with anxiety as main unlike other anxiety episodes, in a psychiatric emergency
complaint were classified as panic or non-panic anxiety department in Barcelona seem to show significant mete-
according to strict independent and retrospective criteria. orotropism. Assessing specific disorders instead of overall
Both groups were assessed and compared with meteoro- emergencies or other variables of a more general quality
logical data (wind speed and direction, daily rainfall, could shed new light on the relationship between weather
temperature, humidity and solar radiation). Seasons and conditions and behaviour.
weekend days were also included as independent vari-
ables. Non-parametric statistics were used throughout Keywords Weather · Wind · Rain · Panic · Anxiety
since most variables do not follow a normal distribution.
Logistic regression models were applied to predict days
with and without the clinical condition. Episodes of panic Introduction
A. Bulbena ()) · R. Acea · C. Garcia-Ribera The relationship between meteorological variables and
Department of Psychiatry, Institut d’Atenci Psiquitrica: Salut human behaviour has been the subject of conjecture since
Mental i Toxicomanies, IMAS, Hospital del Mar, Hippocrates. This interaction has been held more in
Universitat Autonoma Barcelona, popular belief than in scientific verification. However,
Passeig Maritim 29, 08003 Barcelona, Spain since the mid-1900s it has been more thoroughly studied,
e-mail: abulbena@acmcb.es and we are now in a better position to test popular beliefs
Tel.: +34-93-2483175 regarding this connection.
Fax: +34-93-2483445
Numerous studies in different fields have been carried
G. Pailhez out in an attempt to assess this relationship and its im-
Centres Assistencials Dr. Emili Mira i Lpez, plications. Studies regarding meteorological variables and
Recinte Torribera, Sta Coloma de Gramanet, stroke onset (Wang et al. 2002), myocardial infarction
Barcelona, Spain (Larcan et al. 1983) and arthritic pain (Redelmeier and
J. Cunillera · A. Rius Tversky 1996) have shown significant results. Moreover,
Servei Meteorolgic de Catalunya, subjective experience from patients, such as variations in
Spain pain thresholds and mood swings when the weather
changes, points towards an association and paves the road
J. Gutirrez
towards further research.
Servei Urgncies,
Hospital del Mar, IMAS, In the psychiatric field, findings have been irregular
Barcelona, Spain and speculations about the physiological process of such
interaction unclear. However, research has posed several
C. Rojo hypotheses as to how the weather influences behaviour. A
Centre de Salut, few syndromes have been proposed in the literature. The
Port de la Selva, Girona, Spain
239
“serotonin irritation syndrome” is an anxiety state oc- intolerance, etc. (Gazit et al. 2003). Due to findings
curring in the presence of elevated levels of atmospheric confirming that joint hypermobility patients and panic
or ambient cations, and is associated with elevated central disorder patients tend to overlap (Bulbena et al. 2004), it
and peripheral serotonin levels. It is characterized by, is plausible that their autonomic nervous system might be
among other symptoms, anxiety, insomnia, migraines, over-reactive to some environmental stimuli, such as the
vomiting, breathing difficulty and rheumatoid pain (Gi- weather. This evidence has partly motivated the need to
annini 1978; Giannini et al. 1983). Although based mainly assess the association of meteorological variables with
on animal evidence, Sulman (1976) described the “ex- anxiety disorders and panic attacks separately and
haustion syndrome” associated with warm winds and heat specifically.
waves, which is caused by a deficiency of cate- Moreover, results from studies assessing overall psy-
cholamines; symptoms include hypotension, fatigue, ap- chiatric emergency consultations and weather are not
athy, lack of concentration and episodes of hypogly- clear because significant interactions in some subgroups
caemia. Although appealing, a word of caution should be might be hidden within huge samples. Examining psy-
taken into account as no conclusive confirmation data for chiatric emergencies according to syndromic profiles such
humans is available. Moreover, exposure to cold tem- as panic attacks can minimise this bias and help verify the
peratures could be responsible for inducing physiological specificity of each.
stresses, including sympathetic autonomic system acti- This study aims to determine the relationship between
vation (Zvan et al. 1998; Douglas et al. 1995). The core meteorological variables and anxiety emergency consul-
interaction between meteorological variables and physi- tations grouped as panic and non-panic episodes in Bar-
ological variables is still being sought, but current studies celona (Spain) during the 365 days of the year 2002.
are elucidating this issue.
Most studies regarding psychiatric disorders and
weather have assessed general variables such as overall Materials and methods
emergency consultations (Garca-Carretero et al. 1989;
Maes et al. 1994; Makie et al. 2002), number of telephone All emergency consultations with anxiety as main complaint at-
calls (Driscoll and Stillman 2002; Hribersek et al. 1987; tended at Hospital del Mar, Barcelona, during 2002 were selected
for the study. Cases were classified into panic and non-panic epi-
Noble 1996), or suicide rates (Barker et al. 1994; Garca sodes by review of patient’s history records and retrospective cri-
et al. 1991; Linkowski et al. 1992; Maes et al. 1993; Salib teria. The clinical information provided by the psychiatric emer-
1997; Salib and Gray 1997; Wang et al. 1997). Results gency department included main complaint, clinical description,
have been more suggestive when studying specific ill- results of clinical severity scales and diagnosis based on DSM IV
criteria (Diagnostic and statistical manual of mental disorders, 4th
nesses (schizophrenia, mania, etc.) rather than complex edn; American Psychiatric Association) enabling a strict retro-
dimensions like service utilisation (Albert et al. 1991; spective distinction between panic attacks and other anxiety states.
Peck 1990; Salib and Sharp 2002; Summers and Shur The selected cases could be associated with other conditions (i.e.
1992). San Gil, a psychiatrist and meteorologist, pub- depression) provided that anxiety manifestations were predominant
lished results confirming an association between clinical and had appeared in the previous 12 h.
Daily meteorological data were obtained from the closest ob-
cases treated at emergency services and weather condi- servatory to the hospital area, supplied by the Meteorological
tions (San Gil et al. 1988, 1994). Service of Catalonia. Variables included were: (1) maximum,
There are few studies designed to specifically assess minimum and average temperatures ( C), (2) average relative hu-
the association between meteorological variables and midity (%), (3) average barometric pressure (hPa), (4) average wind
speed (m/s), (5) predominant wind direction [according to the wind
clearly defined psychiatric disorders. In this paper, the rose in the area and therefore not analysed as a numerical variable:
relationship between the weather and anxiety and panic tramontana (winds from the north), gregal (from the northeast),
attacks is addressed specifically. Anxiety disorders are a levante (from the east), xaloc (from the southeast), migjorn (from
clinically heterogeneous group that should be evaluated the south), garb (from the southwest), poniente (from the west),
and mestral (from the nothwest)], (6) daily precipitation (mm
by differentiating panic and non-panic anxiety states due equivalent to l/m2), and (7) solar radiation (MJ/m2). Barcelona has
to their different clinical features. Panic disorder has been a Mediterranean climate, with warm summers with little rainfall
associated with an alteration in collagen fibres (Bulbena and mild winters. Mean values of meteorological variables men-
et al. 1993, 1988; Gago 1992; Gratacs et al. 2001), which tioned above for the period 1993–2002 are in Table 1. The seasons
is not seen in generalised anxiety disorder. In patients and number of days with predominance of each of the major winds
were registered. Rainfall as a categorical variable (presence/ab-
with panic disorder, the benign joint hypermobility syn- sence) was also examined. Finally, the day of the week was in-
drome (a heritable collagen disorder) is almost 17 times troduced in the analysis to control a frequent bias in emergency
more common than in any other group of patients or the services derived from the lower incidence of consultations during
general population (Martn-Santos et al. 1998). Due to the weekend.
Statistical analysis was performed using the JMP program
this association, the physical variables tend to be more (version 5.01; SAS 2003). Nonparametric correlation was applied
relevant, as seen in pain threshold variations present in to assess simple association between meteorological variables and
both joint hypermobility (Grahame 2000) and panic dis- number of anxiety episodes, both collected on a daily basis. The
order (Schmidt et al. 2002). More recently, joint hyper- non-parametric Spearman’s correlation coefficient (rho) is com-
puted on the ranks of the data. On the other hand, to test whether
mobility patients have been found to suffer more often group means or medians differ across groups, the nonparametric
from dysautonomia, which includes, among other symp- Wilcoxon rank test (two independent groups compared) and
toms, disregulation of blood pressure, heart rate, heat Kruskal-Wallis test (more than two independent groups compared)
240
Table 1 Mean values of meteorological variables during the period Results are presented as successive analyses between
1993–2002 in Barcelona anxiety episodes, or cases, per day (overall anxiety epi-
Annual mean temperature 16.7 C sodes, panic attacks, and non-panic episodes) and each
Annual mean of maximum temperatures 19.4 C meteorological variable by comparing each group by
Annual mean of minimum temperatures 14.0 C season, wind, precipitation, day before rainfall and, fi-
Annual average precipitation 523.3 mm
Annual average of days with precipitation 71 days nally, day of the week.
Annual mean relative humidity 72% The total number of anxiety episodes, panic attacks
Annual mean barometric pressure 1,016 hPa and non-panic episodes was correlated with eight mete-
Annual mean wind speed 2.9 m/s orological variables. Correlations did not reach signifi-
Annual predominant wind direction Northeast (gregal) cance, except for a negative correlation between rainfall
Annual mean of daily solar radiation 14.7 MJ/m2
and panic attacks. In other words, panic attacks seem to
be less frequent with rainfall (Table 2). Autumn had the
highest number of anxiety episodes and spring the fewest,
were used because most of the studied variables did not follow a both overall and separately. However, the differences did
normal distribution. Finally, logistic regression models were ap- not reach statistical significance (Table 3). As for the
plied to predict days with anxiety. Dichotomous (0/1) dependent
variables chosen were: days with anxiety episodes (any), days with predominant wind, no significant differences were found
panic anxiety episodes, and days with non-panic anxiety episodes. for total anxiety episodes and for non-panic episodes.
The corresponding odds ratio (OR), with 95% confidence intervals However panic attacks were significantly more common
(CI), were obtained for every selected variable. Significance is in days with poniente wind (P=0.006; Table 4).
reached when the 95% confidence intervals do not include 1.
The total number of anxiety episodes, panic episodes
and non-panic episodes per day were compared according
to the presence or absence of rainfall on that day. No
Results significant differences were found for total anxiety epi-
sodes and for non-panic episodes. However, panic attacks
A total of 368 psychiatric emergencies with anxiety as were significantly less frequent in days with rain
main complaint were registered during 2002 in Hospital (P=0.008; Table 5).
del Mar. This results in an average of around one case per Total number of anxiety episodes, panic episodes and
day even though there were days with none. There were non-panic episodes per day were also compared according
223 (60.6%) cases of non-panic anxiety (94 males and to their appearance during the week (Monday–Friday) and
129 females) and 145 (39.4%) cases of panic attacks (48 weekends. No significant differences were found for
males and 97 females). The mean age of all patients was panic attacks. However, non-panic episodes and total
37.8 years. number of anxiety episodes were found significantly less
Cases with anxiety symptoms were attended on often during the weekend (P=0.025; Table 6).
212 days of the year (57.1%). Panic attacks were present
on 118 days of the year (32.3%) and non-panic anxiety on
155 days (42.5%). Both conditions coincided on 57 days Predictive model of days with anxiety
(15.6%).
The diagnostic distribution of non-panic anxiety was: A logistic regression model was developed taking the
anxious-depressive features (mixed anxiety and depres- presence or absence of anxiety episodes (total, panic and
sion symptoms) in 23.5% of cases, adjustment disorder non-panic) as the dependent variable. The independent
with anxiety (symptoms of anxiety appearing after a clear variables included the three significant variables in the
life event) in 21.5%, anxiety disorder due to general unvariate model: precipitation, poniente wind and week-
medical condition or substance-induced in 16.7%, non- end days. Autumn was also included as a variable because
specified anxiety disorder in 10%, generalised anxiety it was over-represented in all types of anxiety episodes
disorder in 9 and 19.3% of others. despite not being significant. The OR and 95% CI were
calculated for each selected variable.
Table 2 Non-parametric correlations (Spearman’s rho) between number of anxiety episodes and meteorological variables
Variable Total anxiety episodes Panic anxiety episodes Non-panic anxiety episodes
rho P rho P rho P
Maximum temperature 0.09 0.08 0.06 0.27 0.10 0.06
Minimum temperature 0.09 0.07 0.06 0.26 0.10 0.07
Average temperature 0.09 0.09 0.06 0.24 0.09 0.08
Relative humidity
0.07 0.21
0.04 0.47
0.05 0.30
Barometric pressure 0.01 0.90 0.04 0.43
0.04 0.47
Wind speed 0.05 0.35 0.03 0.62 0.05 0.38
Precipitation
0.08 0.15
0.11 0.03 0.00 0.94
Solar radiation 0.05 0.37 0.04 0.39 0.03 0.62
241
Table 3 Anxiety episodes by Season N Total anxiety episodes Panic anxiety episodes Non-panic anxiety episodes
season. P values obtained by per day per day per day
Kruskal-Wallis test. SE Stan-
dard error Daily average SE Daily average SE Daily average SE
Spring 92 0.84 0.11 0.35 0.06 0.48 0.08
Summer 92 1.12 0.11 0.39 0.06 0.72 0.08
Autumn 90 1.19 0.11 0.45 0.07 0.73 0.08
Winter 91 0.89 0.11 0.39 0.07 0.49 0.08
P=0.14 P=0.40 P=0.16
Table 4 Anxiety episodes by Kind of Number Total anxiety Panic anxiety Non-panic anxiety
predominant wind. P values wind episodes episodes episodes
obtained by Kruskal-Wallis test
Mean SE Mean SE Mean SE
Tramontana 55 1.13 0.15 0.44 0.08 0.69 0.11
Gregal 40 1.13 0.17 0.40 0.10 0.73 0.13
Levante 29 1.03 0.20 0.38 0.12 0.66 0.16
Xaloc 54 0.91 0.15 0.35 0.09 0.56 0.12
Migjorn 46 0.72 0.16 0.33 0.09 0.39 0.13
Garbi 37 0.97 0.18 0.24 0.10 0.73 0.14
Poniente 47 1.28 0.16 0.74 0.09 0.53 0.12
Mestral 57 0.93 0.14 0.28 0.08 0.65 0.11
P=0.21 P=0.006 P=0.53
Table 5 Anxiety episodes by Rain conditions N Total anxiety Panic anxiety Non-panic anxiety
days with rain. P values ob- episodes episodes episodes
tained by Wilcoxon rank test
Average SE Average SE Average SE
Days with rain 95 0.86 0.11 0.26 0.06 0.60 0.08
Days with no rain 270 1.06 0.06 0.44 0.04 0.61 0.05
P = 0.14 P=0.008 P=0.98
Table 6 Anxiety episodes by Days N Total anxiety Panic anxiety Non-panic anxiety
weekend. P values obtained by episodes episodes episodes
Wilcoxon rank test
Mean SE Mean SE Mean SE
Monday-Friday 261 1.099 0.067 0.409 0.03 0.689 0.05
Saturday-Sunday 104 0.778 0.106 0.365 0.06 0.413 0.08
P=0.025 P=0.76 P=0.006
Table 7 Odds ratio (OR) for Variable Any anxiety days Panic anxiety days Non-panic anxiety days
days with anxiety (all, panic and
non-panic), through logistic re- OR (95% CI) OR (95% CI) OR (95% CI)
gression models. CI Confidence
interval Poniente wind 1.23 0.66 2.36 3.32 1.76 6.34 0.60 0.30 1.15
Saturday–Sunday 0.69 0.43 1.09 0.92 0.55 1.52 0.55 0.34 0.89
Autumn 1.63 0.99 2.72 1.67 1.00 2.77 1.40 0.86 2.27
Rain 0.78 0.49 1.26 0.55 0.31 0.93 0.94 0.58 1.51
Whole model P 0.11 0.0003 0.035
When any anxiety episode was used as dependent 1.67 (CI 95% 1.0–2.77). However, weekend days did not
variable (days with anxiety episodes = 1; days without reach significant OR in the model. Finally, taking days
anxiety episodes = 0), the model was not significant with with/without non-panic anxiety as dependent variable, the
a chi square of 7.37; df=4 (P=0.11). When days with/ model was significant with a chi square of 10.32; df=4
without panic attacks was used as dependent variable, the (P=0.035). The only variable in this model that reached
model reached significance with a chi square of 21.12; significance was weekend days, which were significantly
df=4 (P<0.0003). The presence of rainfall had an OR of less likely to be associated with non-panic anxiety in the
0.55 (CI 95% 0.31–0.93). The poniente wind had an OR emergency department (Table 7).
of 3.32 (CI 95% 1.76–6.34) and autumn had an OR of
242