Marzuk PM, Tardiff K, Leon AC, Hirsch CS, Portera L, Iqbal MI, Nock MK, Hartwell N. Ambient Temperature and Mortality From Unintentional Cocaine Overdose. JAMA. 1998;279(22):1795-1800. doi:10.1001/jama.279.22.1795
From the Section of Epidemiology, Department of Psychiatry (Drs Marzuk, Tardiff, and Leon, Mss Portera and Hartwell, and Messrs Iqbal and Nock), and Department of Public Health (Dr Tardiff), Cornell University Medical College, New York, NY; and Office of Chief Medical Examiner, New York, and Department of Forensic Medicine, New York University School of Medicine, New York (Dr Hirsch).
Context.— Hot weather taxes cardiovascular function and is associated with increased
deaths from heart disease. Cocaine can cause hypertension, tachycardia, coronary
vasospasm, arrhythmias, and increased core temperature.
Objective.— To determine the association between mortality from cocaine overdose
and hot weather.
Setting.— New York, NY.
Design.— Retrospective review of medical examiner cases from 1990 through 1995.
Subjects.— All fatal unintentional cocaine overdoses from 1990 through 1992 (n=1382)
and all hyperthermia deaths of cocaine users (n=10) were used to identify
a maximum daily temperature threshold above which mortality from cocaine intoxication
increased. The study population consisted of all fatal unintentional cocaine
overdoses from 1993 through 1995 (n=2008) and 4 contemporaneous comparison
groups that included fatal unintentional opiate overdoses (n=793), all other
fatal unintentional overdoses (n=85), and a subset of homicides (n=4638) and
fatalities from motor vehicle crashes (n=815).
Main Outcome Measures.— The number of overdose deaths and the proportion of homicides and traffic
fatalities with a positive cocaine toxicology test result on days with a maximum
temperature above or below the temperature threshold.
Results.— A threshold temperature of 31.1°C (88°F) was identified, above
which the mean daily number of fatal cocaine overdoses increased steadily.
On days with a maximum daily temperature of 31.1°C (88°F) or higher
("hot days"), the mean daily number of cocaine overdose deaths was 2.34 (SD=1.68),
which was 33% higher than the mean on days with a maximum temperature of less
than 31.1°C (88°F) (mean=1.76 [SD=1.37] (P<.001).
In contrast, the mean number of opiate overdose deaths per day was 0.81 (SD=0.94)
on hot days and 0.71 (SD=0.86) on other days (P=.28).
For other drug overdose deaths, the mean number of deaths per day was 0.08
(SD=0.28) on hot days and 0.08 (SD=0.28) on other days (P=.69). Among homicides, the proportion with a positive cocaine toxicology
test result was 18.9% on hot days and 19.5% on other days (P=.69), and among traffic fatalities, the proportions with positive
cocaine toxicology test results were 9.5% on hot days and 10.3% on other days
Conclusions.— High ambient temperature is associated with a significant increase in
mortality from cocaine overdose. Based on our comparison groups, the increase
is not explained by changes in cocaine use among the general population. Although
cocaine use is dangerous on all days, it appears to be even more dangerous
on hot days.
COCAINE-RELATED morbidity and mortality constitute a significant public
health problem.1 Approximately 8000 drug abuse
deaths occur in the United States each year, about half of which involve cocaine
use.2 In addition, there are more than 400000
emergency department visits for drug abuse, many of which involve cocaine.3
Cocaine use can result in serious toxic effects and death, in part because
of hyperthermia.4- 7
However, even relatively low doses of cocaine can result in elevated core
temperatures that are below those seen in extreme hyperthermia, but nonetheless
tax cardiac reserve.5,8 In addition,
as a sympathomimetic agent, cocaine can produce tachycardia, hypertension,
coronary vasospasm, and arrhythmias.7- 9
The association between cardiovascular deaths and hot weather is well documented.10- 13 Thus,
the number of deaths from cocaine intoxication might be greater on days with
high ambient temperature because of the effects of cocaine on cardiovascular
function or its thermogenic properties.
To investigate this hypothesis, we conducted a medical examiner surveillance
study in New York City to assess the association of hot weather with mortality
from unintentional cocaine overdoses. To rule out the possibility that a generalized
increase of overdose deaths occurred on hot days, we also assessed the association
of ambient temperature with fatal overdoses due to drugs other than cocaine,
such as opiates, that do not have sympathomimetic or thermogenic properties.
The exact number of individuals who used cocaine on specific dates is
unknown. Excess mortality from cocaine overdoses may occur on hot days, not
because of the cardiovascular or thermogenic effects of the drug, but because
more cocaine users may take the drug on such days. To examine this possibility,
we also assessed the relationship between temperature and the presence of
cocaine in 2 comparison groups, homicides and traffic fatalities.
All cases of fatal unintentional drug overdoses in New York City from
1990 through 1995 were identified through manual review of all medical files
at the Office of Chief Medical Examiner of New York. This office has the responsibility
for assessing all cases of persons believed to have died in an unnatural manner,
including drug overdoses. To be classified as a subject, a person had to be
certified as having died of an unintentional overdose caused by 1 or more
drugs. Cases certified as suicidal or homicidal overdoses were excluded.
In the attribution of the cause and manner of death, the medical examiner
uses the decedent's medical history, the circumstances and environment of
the fatality, autopsy findings, and supporting laboratory data. A diagnosis
of death caused by intoxication by 1 or more drugs requires that the toxicological
data be within the range customarily encountered in such fatalities, that
the history and circumstances be consistent with a fatal intoxication, and
that the autopsy fail to disclose a disease or physical injury that has an
extent or severity inconsistent with continued life. For instance, in the
findings of a disease inconsistent with life, a concurrent intoxication would
be regarded as incidental and unrelated to the cause of death.
In deaths determined to be caused by drug intoxication, when toxicology
results reveal the presence of more than 1 drug in concentrations greater
than trace amounts, it is customary to include all of the identified drugs
in the cause of death. An exception occurs when 1 drug is present in extremely
high concentrations and the others are present in concentrations commonly
encountered in persons who die from other unequivocal causes, such as firearm
For purposes of data analysis, all fatal unintentional overdoses were
divided into the following 3 hierarchical, mutually exclusive groups: (1)
those in whom cocaine was, on the basis of history, circumstances, autopsy,
and toxicological testing, causative of death; (2) those in whom opiates were
causative of death, but in whom cocaine or its metabolite either were not
detected or were detected in clinically insignificant amounts at toxicological
testing; and (3) the remaining overdoses caused by substances other than cocaine
For each case of fatal unintentional drug overdoses, demographic data,
time, date, location of injury and death, cause and manner of death, and toxicology
data were abstracted from the files. In a practical sense, the time of injury
and of death differ by minutes to hours in almost all such cases.
We used 2 comparison groups of fatal injuries that occurred in the city
from 1993 through 1995. Homicides were chosen as a comparison group because,
on the whole, they have similar demographic characteristics to those dying
of unintentional overdoses in New York City. Moreover, persons dying of homicide
or unintentional drug overdoses generally reside in the same communities of
the city that have high rates of drug use.14
We also used a group composed of persons between the ages of 15 and 54 years
who had died in a motor vehicle crash (as a driver, a passenger, or a pedestrian).
This sample was restricted to this age group because 95% of cocaine overdoses
occur among persons in this age range. In the same manner as in unintentional
overdoses, for each homicide and traffic fatality, demographic data, time,
date, location, cause and manner of death, and toxicology data were abstracted
from the files.
Urine and blood specimens collected at autopsy were stored at 4°C
until they were assayed. A single toxicologic laboratory at the medical examiner's
office performed all assays. Benzoylecgonine, the principal metabolite of
cocaine, was initially screened for in urine by enzyme immunoassay. A specimen
was considered positive if the concentration of benzoylecgonine was at least
0.3 mg/L. If urine was not readily available at autopsy, benzoylecgonine was
screened for in blood by radioimmunoassay. A blood specimen was considered
positive if the concentration of benzoylecgonine was at least 0.1 mg/L. Positive
immunoassay findings were substantiated by radioimmunoassay in another tissue.
Blood was screened for cocaine by gas chromatography involving a nitrogen-phosphorus
detector. A specimen was considered positive if the concentration of cocaine
was equal to or exceeded 0.1 mg/L. All positive findings were confirmed by
gas chromatography and mass spectrometry.
Opiates were screened for in urine by enzyme immunoassay or, if urine
was not available, by radioimmunoassay in blood. Specimens that contained
a concentration of at least 0.3 mg/L in urine or 0.1 mg in blood were considered
positive, and were substantiated by radioimmunoassay in another tissue.
Head space chromatography was used to perform ethanol analysis. Other
drugs were screened for and their presence confirmed by a variety of analytic
methods, including enzyme immunoassay, thin-layer and high-performance liquid
chromatography, and gas chromatography–mass spectrometry.
We obtained hourly temperatures from the National Weather Service for
Central Park Station in Manhattan, New York City, for each day from January
1, 1990, through December 31, 1995, and determined the daily maximum temperature
in a 24-hour period. We hypothesized that the number of fatal overdoses involving
cocaine would be higher on those days on which the ambient temperature exceeded
an as yet undefined threshold temperature. Because we are unaware of any literature
that has identified a temperature threshold for cocaine-related mortality,
we sought to identify a threshold using our own data. As there is no single
conventional method for determining such a threshold, we used the following
We identified all medical examiner cases from 1990 through 1995 that
the medical examiner certified as due to hyperthermia (n=48), but were not
classified as unintentional drug overdoses. From this group, we identified
all such cases in which benzoylecgonine was detected at autopsy (n=10). The
daily maximum temperatures for the 10 dates of death ranged from 17.2°C
(63°F) to 39.4°C (103°F) (mean=29.6°C [85.3°F], SD=6.1°C
[10.9°F], median=30.9°C [87.5°F]).
Analyses of heat-related deaths often use a 1-day lag phase, which involves
the comparison of mortality on days that exceed the threshold and the next day, with mortality on all other days.15
For instance, a person who had died of an overdose may not have been found
dead until the next day, or the date of pronounced death might have been recorded
as the next day, especially if the death occurred just before midnight. We
then examined the bivariate plot of the maximum daily temperature (x-axis) by the mean number of cocaine overdose deaths (y-axis) that occurred on days when the temperature was greater than
or equal to a given temperature (Figure 1). For each temperature value, we determined the mean number of
cocaine deaths occurring on days that had a maximum temperature that was greater
than or equal to this specified temperature value. We plotted these data for
the 1096 days from January 1, 1990, through December 31, 1992, which we refer
to as the "index sample" and included all 1382 cases of fatal cocaine overdoses
during this period (n=1382). The plot shows a relatively flat relationship
until 27.8°C (82°F), then a gradual increase, which accelerates more
markedly at 31.1°C (88°F) (Figure
Based on the median of the maximum temperature for the dates of deaths
from the sample of hyperthermia cases involving cocaine and the plot of the
index sample of fatal cocaine overdoses, we chose 31.1°C (88°F) as
the threshold temperature. For purposes of analysis, we considered a day "hot"
if the maximum hourly temperature was greater than or equal to 31.1°C
(88°F), or if the maximum hourly temperature of the preceding day was
greater than or equal to 31.1°C (88°F).
We considered reporting the results of analyses based on the heat index,
which is a function of the temperature in degrees Fahrenheit and the relative
humidity. However, for most persons the temperature is a more readily interpretable
variable. Because temperature is a fundamental component of the heat index,
the Pearson correlation between the daily maximum heat index and the daily
maximum temperature in New York City during the study period was 0.986 (P<.001).
All subsequent statistical analyses were performed using fatal unintentional
overdoses that occurred during the next 3-year period, January 1, 1993, through
December 31, 1995 (ie, the "cross-validation sample"). This split-sample approach
to data analysis was used so that our hypothesis could be tested on a sample
that was distinct from the index sample (1990-1992), which had been used to
establish the threshold temperature.
Separate analyses were conducted for the 3 types of overdoses described
earlier, deaths caused by cocaine; deaths caused by opiates, but not cocaine;
and deaths caused by other drugs. The Mann-Whitney test was used to compare
"hot" and "other" days with regard to the number of deaths from a particular
type of overdose. Day was the unit of analysis. We also assessed the Pearson
correlation coefficient for autocorrelation, which incorporated our 1-day
We further examined whether cocaine use was increased on hot days among
other causes of death during the same period, 1993 through 1995. For that
reason, we compared the proportion of homicides and traffic-related fatalities
with a positive cocaine toxicology on "hot" days with the respective proportions
with a positive cocaine toxicology test result on "other" days. Using χ2 tests, person was the unit of analysis.
For all analyses, 2-tailed α levels of .05 were considered statistically
During the 6-year period there were 48 deaths in New York City attributed
to hyperthermia, including 34 males (70.8%) and 14 females (29.2%). Three
(6.3%) of these deaths were among infants younger than 1 year, 29 (60.4%)
were among those aged 15 through 54 years, and 16 (33.3%) were among those
older than 54 years.
Among the 48 deaths, 40 (83.3%) were screened for benzoylecgonine and
cocaine. Among these 40, benzoylecgonine was detected in 10 (25% of those
screened, 21% of the total sample) and cocaine was detected in 7 (17.5% of
those screened, 14.6% of the sample). In contrast, opiates were detected in
only 1 case (2.5%) of the 40 screened. Ethanol was detected in only 5 (12.2%)
of 41 cases screened.
When the sample of hyperthermia deaths was restricted to the age group
that is most likely to have used cocaine (ie, those aged 15-54 years, n=27),
the percentages that were positive for benzoylecgonine and cocaine were 37.0%
and 25.9%, respectively.
There were 1095 days during the cross-validation period (1993-1995).
Of these, 146 days (13.3%) were considered hot and 30 days were considered
extremely hot, with maximum temperatures reaching 35°C (95°F) or higher.
The hottest days of the period were July 15 and 16, 1995, when the maximum
temperature was 39.4°C (103°F).
During this period, there were 2008 fatal unintentional cocaine overdoses.
Complete autopsies were performed on 93% of these cases. The mean age of this
group was 39.0 years (SD=8.8 years), and 94.9% were aged 15 to 54 years. There
were 1569 men (78.1%) and 439 women (21.9%), and 505 non-Hispanic whites (25.2%),
884 non-Hispanic blacks (44.0%), 605 Hispanics (30.1%), and 14 Asian or others
Significantly more deaths were due to cocaine overdoses on hot days
(n=146 hot days, mean=2.34, SD=1.68), than other days (n=949 other days, mean=1.76,
SD=1.37) (by Mann-Whitney, z=3.98, P<.001) (Table 1). This
difference represents a 33% increase in mean mortality on hot days. Mean daily
mortality began to increase when the maximum temperature equalled or exceeded
31.1°C (88°F) (Figure 2).
The 2 days with the highest number of cocaine overdose deaths (n=9) were July
10 and 13, 1993, when the maximum temperatures were 37.2°C (99°F)
and 33.3°C (92°F), respectively. There were no significant differences
by age, race, gender, or location of death between subjects who had died on
a hot day and those who had died on another day. The autocorrelation coefficient
for cocaine overdose deaths was 0.12 (P<.001),
which indicates that the number of cocaine overdoses on a prior day accounted
for only 1.4% of the variance in cocaine overdose deaths.
Of the 341 subjects who died of cocaine overdoses on hot days, the proportion
that had positive toxicology test results for ethanol was 42.9%, compared
with 41.4% for the 1667 subjects who died on other days. Likewise, the proportions
with positive toxicology test results for benzodiazepines, barbiturates, or
salicylates on hot days were 10.3%, 1.2%, and 0.9%, respectively, compared
with similar proportions as those who died on other days, which were 11.5%
(benzodiazepines), 1.4% (barbiturates), and 1.3% (salicylates). Opiates were
detected in 57.1% of cocaine overdose cases on hot days compared with 61.7%
of such cases on other days.
Some drugs, including tricyclic antidepressants, antipsychotic medications,
or psychostimulants, can affect heat regulation by central mechanisms, by
increasing heat production through muscular activity, or by impairing heat
dissipation. Many of these drugs have anticholinergic or sympathomimetic effects.
Overall, only 20% of the individuals who died of cocaine overdoses had a positive
toxicology test result for at least 1 of these drugs (Table 2), and the proportion of cocaine overdose deaths that had
a positive toxicology test result for these agents on hot days was nearly
identical to the proportion with a positive toxicology test result of those
who died on other days (Table 2).
During the cross-validation period, there were 793 fatal unintentional
overdoses of opiates and 85 fatal overdoses that tested positive for neither
cocaine nor opiates. There were no significant differences in the number of
deaths on hot days and the number on other days for either of these drug-death
groups (Table 1). Figure 3 and Figure 4
show no discernible change in the mean number of such deaths at any temperature.
During the cross-validation period, there were 4793 homicides in New
York City, of which 4638 (96.8%) were assessed for the presence of cocaine
in blood. Among these 4638 homicides, 902 homicides had a positive toxicology
test result for cocaine. The demographic characteristics of the homicides
that were positive for cocaine were similar to the cocaine overdose group
(mean [SD] age=32.9 [9.0] years; 97.5% between the ages of 15 and 54 years;
83.6% men; and 8.0% non-Hispanic white, 53.4% non-Hispanic black, 37.9% Hispanic,
and 0.6% Asian or others). Of the 719 homicides that occurred on "hot" days,
136 (18.9%) had a positive toxicology test result for cocaine. Of the 3919
homicides that occurred on "other" days, 766 (19.5%) had a positive toxicology
test result for cocaine (χ2=0.11; df=1, P=.69).
During the cross-validation period, there were 921 traffic-related fatalities
among persons between the ages of 15 and 54 years in New York City, of which
815 (88.5%) were assessed for cocaine in blood. Among these 815 fatalities,
83 had a positive toxicology test result for cocaine. The demographic characteristics
of these 83 cases were similar to the cocaine overdose group (mean [SD] age=34.4
[8.5] years; 79.5% men; and 14.5% non-Hispanic white, 44.5% non-Hispanic black,
39.8% Hispanic, and 1.2% Asian or others). Of the 126 traffic fatalities that
occurred on "hot" days, 12 (9.5%) had a positive toxicology test result for
cocaine. Of the 689 traffic fatalities that occurred on "other" days, 71 (10.3%)
had a positive toxicology test result for cocaine (χ2=0.01, df=1, P=.91).
This study demonstrates several main findings. First, a quarter of all
individuals younger than 55 years who died from hyperthermia in New York City
had used cocaine immediately prior to their deaths. Second, the mean number
of deaths from cocaine overdoses on days in which the temperature was equal
to or greater than 31.1°C (88°F) was 33% higher than the mean on days
with lower maximum temperatures. On days with temperatures higher than 31.1°C
(88°F), daily mortality from cocaine overdoses increased steadily. Third,
the association of ambient temperature with drug overdose mortality appears
specific to cocaine, and not to other major drugs of abuse, such as opiates.
Although the number of deaths from cocaine overdose was increased on hot days,
only 10% of days in a year on average reach a temperature of 31.1°C (88°F)
in New York City. Thus, the temperature effect we found does not have a major
influence on the annual mortality from cocaine overdoses.
The exact reason for an association between high ambient temperature
with mortality from cocaine overdose is not known. Increased temperature places
considerable demands on the cardiovascular system to increase cardiac output
and to decrease systemic vascular resistance.16
The immediate effects of cocaine use include increases in mean arterial pressure,
heart rate, and cardiac output.17 Moreover,
cocaine use has been associated with cardiac abnormalities including coronary
artery occlusion, malignant arrhythmias, and myocarditis.9
Thus, cocaine use during hot weather may further tax cardiovascular capacity
and increase the risk of mortality.
Another explanation is that the excess of such deaths on hot days may,
in some cases, be due to the thermogenic effects of cocaine. These effects
have been postulated to result from the propensity of cocaine to cause increased
muscular activity and peripheral vasoconstriction, or its direct effect on
dopamine-modulated, heat-regulatory centers in the hypothalamus.18- 20
For instance, several case reports have documented fatal hyperthermia with
and without rhabdomyolosis in cocaine users.21- 23
Dogs administered intravenous cocaine experienced a significant increase in
core temperatures, which proved fatal.24,25
Moreover, above an ambient temperature of 11.5°C (52.7°F) the increase
in core body temperature in canines was positively correlated with the ambient
temperature at which the cocaine had been administered.25
Hyperthermia is a clinical diagnosis, rather than a diagnosis established
by pathologic findings. Therefore, the number of deaths classified with this
cause is quite limited. Persons who overdose on cocaine may not be discovered
for hours after death, long after it would have been possible to determine
if a clinical hyperthermia syndrome had preceded their deaths. Although we
cannot determine from our database whether many of the fatal cocaine overdoses
on hot days were attributable to hyperthermia, at least 1 in 4 deaths in New
York known to have resulted from hyperthermia had a positive cocaine toxicology
test result. This prevalence is considerably greater than the 1.3% prevalence
of recent cocaine use (in the past 30 days) reported in general population
household surveys of drug use in New York.26
Moreover, most cocaine-related hyperthermia deaths occurred on warm or hot
days. It is possible that some of the deaths from acute cocaine intoxication
may have been cases of undetected hyperthermia.
Drug overdose deaths that involve a single agent are uncommon. Many
cocaine users concurrently administer opiates, ethanol, and other sedative
drugs. However, only about 20% of cocaine overdose cases had a positive toxicology
test result for other drugs that are commonly associated with a risk of hyperthermia.
Moreover, the proportions of cocaine overdose deaths with a toxicology test
result positive for at least 1 of these drugs was virtually the same for those
who died on hot days vs other days. Thus, our findings cannot be attributed
exclusively to the presence of other drugs that impair heat regulation. It
is probable that individuals who had administered cocaine along with anticholinergic
or sympathomimetic agents may have enhanced their risk of heat-related mortality.
Our findings cannot be explained by a generalized increase in the number
of overdose deaths on hot days as we observed no association of ambient temperature
with overdoses caused by opiates or by other drugs. It is also unlikely that
our findings can be explained by an increase in cocaine use in the New York
City population on hot days. Among homicides, which have similar demographic
characteristics and reside in similar neighborhoods as cocaine overdose cases,
and among persons involved in motor vehicle crashes, we found that cocaine
was detected in virtually identical proportions among those dying on hot days
or on other days. Rates of screening for drugs also are not influenced by
hot weather in New York City.
It is likely that our findings from New York would apply to other cities
in the United States. Ambient temperatures in cities are often higher than
other regions because of increased heat production from crowds, cars, and
factories; retention of heat by buildings and pavements; and diminished heat
loss because of low wind speed.27,28
Socioeconomic factors also may contribute to heat-related mortality, as city
residents are likely to be older and poorer, and more likely to live in substandard
housing without air conditioning, showers, or baths, than suburban or rural
residents.29 In an earlier study, we found
that many drug overdose deaths occur in poor neighborhoods,14
which are likely to include few homes with air conditioners.
Cities in the Northeast and on the Pacific Coast have the lowest threshold
temperatures above which general mortality increases.30
People who reside in higher latitudes or near oceans experience fewer hot
days, but have a harder time adjusting to sudden changes in temperature, particularly
in early summer.10,27 Thus, the
threshold of 31.1°C (88°F) that we found in New York may vary somewhat
in other locations.
Typically, periods of high humidity and warm nighttime temperatures,
which reflect persistent, high heat exposure and diminished cooling from evaporative
loss, have been associated with increased general mortality, rather than transient
daytime peaks in temperature.31 However, the
models that best account for variability in general mortality in New York
City involve dry, hot conditions (ie, a threshold of 33.3°C [92°F],
a low dew point, and a 1-day lag phase in mortality).30
Our analyses using the heat index yielded similar results, which is not surprising
given that during the study period the maximum daily heat index was so highly
correlated with the maximum daily temperature.
Several methodologic issues in our study warrant comment. First, it
is not possible to know the dose of cocaine that an individual had used prior
to death. Thus, the relationship between dose of cocaine and ambient temperature
is unknown. Second, we do not know the exact ambient temperature to which
the subjects were exposed when they used cocaine. However, even nighttime
low temperatures on very hot days in New York remain in the 70s. Third, the
threshold temperature of 31.1°C (88°F) was established for a group
and not for individuals. Some individuals may have higher or lower thresholds
at which their mortality risk from cocaine use changes. Fourth, we do not
know whether the individuals who died were occasional cocaine users or long-term,
Although it is not possible to prove that high ambient temperature results
in a direct, increased risk of death from cocaine use, our data suggest that
the increased mortality from cocaine overdoses on hot days is not explained
by changes in cocaine use in the general population on such days. Cocaine
use is dangerous on all days. Our findings suggest that the risk of death
associated with cocaine increases further during hot weather.