Acute health effects of the Tasman Spirit oil spill on residents of Karachi, Pakistan.

ral
ssBioMed CentBMC Public Health
Open AcceResearch article
Acute health effects of the Tasman Spirit oil spill on residents of
Karachi, Pakistan
Naveed Zafar Janjua*1,3, Pashtoon Murtaza Kasi1, Haq Nawaz1,
Sadia Zohra Farooqui1, Urooj Bakht Khuwaja1, Najam-ul-Hassan1,
Syed Nadim Jafri1, Shahid Ali Lutfi2, Muhammad Masood Kadir1 and
Nalini Sathiakumar3
Address: 1Department of Community Health Sciences, the Aga Khan University, Karachi, Pakistan, 2Sindh Environmental Protection Agency,
Karachi, Pakistan and 3Department of Epidemiology, School of Public Health, University of Alabama at Birmingham, Birmingham, AL, USA
Email: Naveed Zafar Janjua* - naveed@uab.edu; Pashtoon Murtaza Kasi - pashtoon.kasi@gmail.com; Haq Nawaz - haq.shahzad@gmail.com;
Sadia Zohra Farooqui - virtualmind32@hotmail.com; Urooj Bakht Khuwaja - urooj_nawaz@yahoo.com; Najam-ul-
Hassan - najamhasan@hotmail.com; Syed Nadim Jafri - nadimsj@hotmail.com; Shahid Ali Lutfi - shahid_lutfi@yahoo.com;
Muhammad Masood Kadir - masood.kadir@aku.edu; Nalini Sathiakumar - nalini@uab.edu
* Corresponding author
Abstract
Background: On July 27 2003, a ship carrying crude oil run aground near Karachi and after two weeks released
37,000 tons of its cargo into the sea. Oil on the coastal areas and fumes in air raised health concerns among
people. We assessed the immediate health impact of oil spill from the tanker Tasman Spirit on residents of the
affected coastline in Karachi, Pakistan.
Methods: We conducted a study consisting of an exposed group including adults living in houses on the affected
shoreline and two control groups (A and B) who lived at the distance of 2 km and 20 km away from the sea,
respectively. We selected households through systematic sampling and interviewed an adult male and female in
each household about symptoms relating to eyes, respiratory tract, skin and nervous system, smoking, allergies,
beliefs about the effect on their health and anxiety about the health effects. We used logistic regression
procedures to model each symptom as an outcome and the exposure status as an independent variable while
adjusting for confounders. We also used linear regression procedure to assess the relationship exposure status
with symptoms score; calculated by summation of all symptoms.
Results: Overall 400 subjects were interviewed (exposed, n = 216; group A, n = 83; and group B, n = 101). The
exposed group reported a higher occurrence of one or more symptoms compared to either of the control groups
(exposed, 96% vs. group A, 70%, group B 85%; P < 0.001). Mean summary symptom scores were higher among
the exposed group (14.5) than control group A (4.5) and control group B (3.8, P < 0.001). Logistic regression
models indicated that there were statistically significant, moderate-to-strong associations (Prevalence ORs (POR)
ranging from 2.3 to 37.0) between the exposed group and the symptoms. There was a trend of decreasing
symptom-specific PORs with increase in distance from the spill site. Multiple linear regression model revealed
strong relationship of exposure status with the symptoms score (β = 8.24, 95% CI: 6.37 – 10.12).
Conclusion: Results suggest that the occurrence of increased symptoms among the exposed group is more likely
Published: 03 April 2006
BMC Public Health 2006, 6:84 doi:10.1186/1471-2458-6-84
Received: 08 October 2005
Accepted: 03 April 2006
This article is available from: http://www.biomedcentral.com/1471-2458/6/84
© 2006 Janjua et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Page 1 of 11
(page number not for citation purposes)
to be due to exposure to the crude oil spill.

BMC Public Health 2006, 6:84 http://www.biomedcentral.com/1471-2458/6/84
Background
On 27 July 2003, the Greek tanker Tasman Spirit, carrying
67000 tons of Iranian light crude oil from Iran to Paki-
stan, ran aground before entering the harbor channel at
the Karachi port, Pakistan. About two weeks later, on the
night between 13 and 14 August 2003, the ship broke
apart and released its cargo into the sea. Strong winds and
rough sea facilitated the spread of oil to about 10 km of
residential coastline (Figure 1). After this incident, there
were two more episodes of oil spill- the latest being on 29
August 2003, resulting in a total spill of more than 35000
tons of crude oil.
The affected shoreline is a highly populated residential
and recreational area. Fumes of volatile organic com-
pounds and mist containing hydrocarbons accompanied
by a strong smell dispersed into the residential area.
Newspapers and the electronic media provided a wide
television channels showed pictures of piles of dead fish
and turtles on the oil-covered beach. Extensive media cov-
erage raised concerns about the potential of human health
effects among residents in the affected area and among
government officials.
Chemical analysis of crude oil from the tanker showed
that it contained a high quantity of aromatic hydrocar-
bons (personal communication Lutfi SA). Crude oil is
mixture of many chemicals, the major components being
hydrocarbons. Iranian Light Crude oil carried in Tasman
Spirit contains high quantity of sulfur (1.35% by wt), 14%
light naphtha, 20% heavy naphtha and 4% gasoline and
its pour point is -29°C [3]. Aromatic hydrocarbons are
probably the main airborne exposure.
Spread of oil and evaporation of volatile contents after
spill depends on the physical and chemical properties of
Map of the shoreline covered with crude oil residues after crude oil spill from Tasman Spirit in Karachi, Pakistan 2003 (Source: Sindh Environm ntal Protection Agency)Figure 1
Map of the shoreline covered with crude oil residues after crude oil spill from Tasman Spirit in Karachi, Pakistan 2003 (Source:
Sindh Environmental Protection Agency).Page 2 of 11
(page number not for citation purposes)
coverage of the incident and the potential health and envi-
ronmental impact of the oil spill[1,2]. Newspapers and
the crude oil and the prevailing weather conditions. Low
viscosity crude oil spreads rapidly. Low boiling point con-

BMC Public Health 2006, 6:84 http://www.biomedcentral.com/1471-2458/6/84
tents of crude oil evaporate within hours of the spill and
if volatile contents are high then evaporation is more.
High temperature, high speed wind and strong wave
action increase the spread and evaporation of the volatile
contents [4]. During August 10th to August 30th, 2003 the
average maximum temperature in Karachi was 31°C
while average minimum temperature was 27°C with the
average mean of 29°C [5]. July and August is a rough sea
season in Pakistan with strong waves. Moderately high
temperature, strong waves and high speed wind [2]
moved oil from sea to shoreline and winds carried vapors
from sea to land. Initial assessment suggested that about
11,000 tones of volatile organic compounds entered the
air after the spill. Assessment of the air revealed that on
August 20, VOCs in the polluted air ranged from 44 ppm
at Shireen Jinnah Colony to 179 ppm at Village Restau-
rant.
Residents living in this area were exposed to the high level
of volatile organic compounds for at least 15 to 20 days.
The level of pollutants gradually tapered off from 8 to 10
ppm of volatile organic compounds on August 30. The
pungent smell was perceptible till September 5 at the dis-
tance of about 1 km from the sea [6]. These levels suggest
high exposure of population to crude oil contents during
spill days.
Previous investigations of the Sea Empress, the Nakhodka
and the Shetland oil tankers' spills found an increased
occurrence of upper respiratory tract irritation, exacerba-
tion of asthma, vertigo, headache, and back and leg pains
and psychological ailments among persons living in
exposed areas and clean-up workers [7-9].
Studies on acute health effects of exposure to an oil spill
are few. This paper is first of such report from a developing
country. To address the concerns regarding potential
health effects, we undertook an epidemiologic investiga-
tion to determine if exposure to the oil spill had resulted
in increased acute ill health among the exposed resident
population.
Methods
Design and study base
The study began on September 1st 2003, three weeks after
first spill and three days after the last spill. The exact pop-
ulation size of the shoreline is not available; however, it is
estimated that about 700,000 people live in this beach
town. The beach in this area was covered with thick layer
of oil and there were fumes and a mist of oil in the air. The
public was prohibited from visiting the beach after spill.
However, residents were not asked to evacuate. Exposure
was assigned on the basis of geographic location. We
extended from Shirin Jinnah colony (near the harbor) to
Village restaurant (end of the residential area) and who
did not migrate after the oil spill on August 13th, 2003. We
selected two control groups: control group A living at a
distance of 2 km from the affected coastal area and control
group B living at a distance of 20 km from the affected
shoreline. The distance of 2 km and 20 km were chosen
arbitrarily to assess if symptoms depended on the distance
from the shoreline. We obtained distance measurements
from City District Government of Karachi which provided
a list of residential areas at 2 and 20 km, taking the shore-
line as a reference point.
Subject selection
Most of the residential dwellings in both the exposed and
control areas were 4 to 10-storeyed buildings with apart-
ments. No official listing of buildings was available
through which we could have developed a list of only
those buildings that were on the shoreline. There were
approximately 100–150 buildings on the shoreline. We
selected every 4th building on the shoreline facing towards
the beach starting from the Shirin Jinnah colony to the
Village restaurant, a total of 25 buildings. From each
building we randomly selected one household per floor to
assess the effect of elevation on exposure to oil fumes.
From each apartment, we selected the first adult male and
female we met in the home for interview. Both males and
females were included, because of the likelihood of differ-
ences in their daily activities and resulting exposure. Typ-
ically, men spend more time outdoors and women spend
more time indoors. We used the same sampling procedure
to select buildings (control group A, 12 building; control
group B, 12 buildings), apartment and subjects in control
groups. Of 410 persons initially approached, 400 (98%)
participated in the study. Of the 400 persons, 216 were in
the exposed groups, 83 in control group A and 101 in con-
trol group B.
Interviews and questionnaire
A research team (4 sub-teams comprising of 2 interviewers
in each) consisting of physicians and final year medical
students conducted face to face interview of participants
during household visits. We had an in-office training of
the research team on aspects of sampling, consent process,
questionnaire and conducting the interviews. We field-
tested the questionnaire before actual field work and
revised the content as deemed necessary. We developed a
questionnaire that included a symptom check list of the
known health effects of hydrocarbon exposure based on
previous investigations [7,8], adapted from the symptom
checklist developed for the Nakhodka oil spill. We specif-
ically asked respondents to report symptoms (as yes/no)
that occurred after the first major spill on August 13th,Page 3 of 11
(page number not for citation purposes)
defined the exposed group as adults resident in apart-
ments/houses on a 10 km long affected shoreline that
2003. However, we did not ask to specify the exact date of
start of a symptom nor when it became better. Major cat-

BMC Public Health 2006, 6:84 http://www.biomedcentral.com/1471-2458/6/84
egories of symptoms included ocular (sore, itchy, watery
and red eyes), dermal (itchy skin and skin irritation)
upper and lower respiratory tract (scratchy throat, sore
throat, cough, breathing difficulties, wheeze), gastrointes-
tinal (nausea/vomiting, loss of appetite), central nervous
system (headache, dizziness, irritability) and constitu-
tional symptoms (general fatigue, weakness and fever).
Besides the detailed symptom list, the questionnaire
included information on socio-demographic characteris-
tics, location of house, presence of windows towards sea,
active and passive smoking, allergies to chemicals, dust
and pollen, course of illness, care sought, and effects on
daily life. It also included questions regarding perception
about the role of oil spill in producing ill health, percep-
tion that if it had affected respondents' health and anx-
iousness about the effects of oil spill on health [See
Additional file 1]. Before the start of interview we
obtained a verbal informed consent that included infor-
mation on the purpose of study, the procedures, confiden-
tiality, risk and benefits and right to withdraw from study
any time. We gave a copy of consent from to participants
that included phone number and mailing address of prin-
cipal investigator for additional information and clarifica-
tion. Study was reviewed by Sindh Environmental
Protection Agency and the Department of Community
Health Sciences at the Aga Khan University review com-
mittee. On returning from the field every day, we assessed
the filled questionnaires for completeness. Any vague or
inconsistent responses were clarified from the respond-
ents and edited in the questionnaire.
Statistical analysis
We entered data in MS Access database and analysed it
using Statistical Package for Social Sciences (SPSS) version
11. We compared the socio-demographic characteristics
of exposed group with the two control groups by the chi-
square test for categorical variables and ANOVA for con-
tinuous variables. We computed a symptoms score for
symptoms relating to oil spill; if a symptom was present
we gave a score of "1"otherwise "0" and by summing 48
different symptoms we obtained symptoms score for each
subject. We compared symptoms scores among groups
using ANOVA. We compared the proportion of symptoms
between the exposed and control groups using the preva-
lence odds ratio (POR) with its corresponding 95% confi-
dence interval (CI). We present results for symptoms with
prevalence greater than 30% among the exposed or those
have been reported in previous studies to enhance the
informativeness of the results. We used logistic regression
procedures to adjust for other risk factors. We included a
potential risk factor other than crude oil exposure in a
logistic regression model if its P value was less than 0.20
or if it was biologically meaningful. We assessed the asso-
age, sex, education, history of chemical allergy, dust
allergy, active and passive smoking, house on the main
road as proxy for exposure to vehicle smoke and anxiety
about effect of spill on health. For some variables the
numbers in either control group A or B were small, hence
we decided not to present adjusted estimates for symp-
toms with number less than 8 in any of the cell. Goodness
of fit of the logistics regression model was assessed using
the Hosmer and Lemshow test.
We conducted a subgroup analysis on the exposed group
to evaluate the relationship between demographic charac-
teristics and symptoms through logistic regression mode-
ling. The evaluated factors included geographic locations
on the basis of residence, age, sex, presence of windows
towards sea and house on the main road. Goodness of fit
of the logistics regression models was assessed by Hosmer
and Lemshow test.
We performed multiple linear regression to assess the rela-
tionship between exposure status and the symptoms score
while adjusting for age, gender, education, smoking, dust
allergy, chemical allergy, and anxiousness about health
effect of oil spill. However, only chemical allergy and anx-
iousness about health effect of oil spill were significant in
the final model, hence model included only these two var-
iables beside exposure status. Model assumptions for
goodness of fit were assessed using residual analysis that
revealed good fit.
Results
Of the overall 400 participants, 225 (57%) were women
and 175 (43%) were men. There were statistically signifi-
cant differences in the distribution of age, years of educa-
tion, location of house, self-reported health impact and
the number of affected household members between the
three groups. The mean age of subjects in control group A
(46 years) was higher than the other two groups (exposed
group, 35 years; control group B, 31 years) due to a higher
proportion of subjects in the > 50 age group and a smaller
proportion in the 15–30 age group. Control group B sub-
jects had higher mean years of schooling (12.7 years) than
the exposed group (10.3 years) and control group A (11.0
years). A higher proportion of the exposed group (53%)
lived in houses on the main road compared to control
group A (22%) or B (18%). A higher proportion of the
exposed group (78%), a lower proportion of control
group A (53%) and a still lower proportion of control
group B (24%) reported that they were worried about the
possible effects of the oil spill on their health. A higher
proportion of exposed group (81%) believed that oil spill
had affected their health, while a lower proportion of
unexposed group A (23%) and still lower proportion ofPage 4 of 11
(page number not for citation purposes)
ciation of exposure status with selected symptoms as out-
come variable while adjusting for confounding effect of
unexposed group B (3%) believed the same. Differences
in smoking habits, passive smoking status, past history of

BMC Public Health 2006, 6:84 http://www.biomedcentral.com/1471-2458/6/84
Table 1: Selected characteristics, medical history and beliefs of study participants according to exposure category for health effect of
oil spill in Karachi, Pakistan 2003
Variables Exposed group Control group A Control group B
n % n % n % Pa
Total 216 100 83 100 101 100
Gender
Female 123 57 50 60 52 52 0.47
Male 93 43 33 40 49 48
Age group (years)
15–30 95 44 20 24 64 63 <0.001
31–50 87 40 26 31 27 27
>50 34 16 37 45 10 10
Mean (SD) 35.5 (14.3) 45.5 (16.4) 31.2 (13.3) <0.001
Education (years)
0 38 18 10 12 6 6 0.05
1–10 47 22 18 22 13 13
11–12 44 20 17 21 23 23
13–14 48 22 19 23 32 32
>14 39 18 19 23 27 27
Mean (SD) 10.3 (5.5) 12.7 (3.8) 11.0 (5.1) <0.001
Employment status
Housewife 88 40.7 35 42.2 13 12.9 <0.001
Student 28 13.0 8 9.6 26 25.7
Employed 100 46.3 40 48.2 62 61.4
House on main road
Yes 115 53 18 22 18 18 <0.001
No 101 47 65 78 83 82
Smoking
Yes 36 17 12 14 14 14 0.78
No 180 83 71 86 87 86
Passive smoking
Yes 46 21 15 18 17 17 0.60
No 170 79 68 82 84 83
Allergy to chemicals
Yes 59 27 22 26 17 17 0.12
No 157 73 61 74 84 83
Allergy to dust
Yes 78 36 32 39 26 26 0.12
No 138 64 51 61 75 74
Allergy to pollen
Yes 16 7 6 7 0 0 _
No 200 93 77 93 101 100
Ever had wheezing
Yes 18 8 9 11 5 5 0.33
No 198 92 74 89 96 95
Ever had wheezing with shortness of breath
Yes 13 6 6 7 3 3 0.40Page 5 of 11
(page number not for citation purposes)
No 203 94 77 93 98 97