Running Head CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFSCRIT.docx

Running Head: CRITIQUE OF OCEAN TEMPERATURES IN
CORAL REEFS
CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFS
Madison McNeill
Introduction
Coral reef ecosystems are the most diverse marine ecosystem in
the world. They provide a home to thousands of species of
plants and animals. In the last few decades, global warming has
caused increased temperatures, resulting in ocean acidification
and increasing surface temperatures of the ocean. This can lead
to the bleaching of coral reefs as well as the death of coral reef
fishes due to their inability to acclimate to the elevated
temperature. These three papers were chosen, because they
illustrate the environmental impact higher temperatures have on
these coral reefs and the organisms that live within them.
· Dias, M., Ferreira, A., Gouveia, R., Cereja, R., & Vinagre, C.
(2018). Mortality, growth and regeneration following
fragmentation of reef-forming corals under thermal
stress. Journal of Sea Research, 141, 71-82. doi:
10.1016/j.seares.2018.08.008.
· De'ath, G., Lough, J., & Fabricius, K. (2009). Declining Coral
Calcification on the Great Barrier Reef. Science, 323(5910),
116-119. doi: 10.1126/science.1165283.
· Nilsson, G., Östlund-Nilsson, S., & Munday, P. (2010).
Effects of elevated temperature on coral reef fishes: Loss of
hypoxia tolerance and inability to acclimate. Comparative
Biochemistry and Physiology Part A: Molecular & Integrative
Physiology, 156(4), 389-393. doi: 10.1016/j.cbpa.2010.03.009.
Dias (2018) evaluated how elevated surface temperatures of the
ocean affected growth, mortality, and regeneration following
the fragmentation of nine coral species in the Indo-Pacific,
while De'ath (2009) suggested that the ability of coral in the
Great Barrier Reef may have depleted due to a decrease in the
saturation state of aragonite and rising temperature stress in this

region. The third paper evaluated, Nilsson (2010), examined
whether or not an elevated temperature decreased tolerance
levels for low-oxygen regions in two species of coral reef
fishes. This experiment used adults fishes of two species and
tested their ability to acclimate to changes in higher
temperatures, which differed from the other two studies in that
Dias and De’Ath did not study the fishes in the ecosystems,
only the coral there. Dias found that whether or not a coral had
previous injury did not impact the mortality, partial mortality,
or rate of growth of each fragment. However, the species of
coral and the ocean temperature had significant impacts on the
results for each fragment. Although the cause for coral
calcification of Great Barrier Reef corals was not determined by
the De'ath’s study, he did find that it was largely related to
increasing temperatures of oceans, which caused more thermal
stress in coral populations. This differed from the Nilsson
paper, which showed that certain species of coral reef fishes
were unable to adjust to higher ocean temperatures, a
phenomenon that has occurred due to global warming and ocean
acidification.
Analysis
Introduction
When the three articles’ introductions were evaluated, some
similarities as well as dissimilarities stood out. For example,
the titles of the articles varied in appropriateness. Nilsson's
title was too long. The paper had a title that told its audience
what the researchers hoped to get out of it, but the title seemed
long and bulky. The title, in my opinion, could have been
shortened or rephrased to one that grabbed the audience’s
attention more quickly, even a change as simple as changing the
title to, “Effects of elevated temperature on coral reef fishes.”
However, Dias’s title was accurate and concise. “Mortality,
growth and regeneration following fragmentation” was a title
that accurately explained what was being examined within the
confines of this study. De'ath had a title that matched the
contents of the paper as well.

The abstract’s statement of purpose of all three articles matched
the introductions. For the Dias article, they stressed that the
impacts of thermal stress on fragments of regenerating coral
species needed to quickly be explored, while De'ath’s abstract
was well written, telling readers how many coral colonies were
studied and what the results showed. The abstract of Nilsson’s
paper plainly stated what occurred within the first two
sentences. The abstract’s statement of purpose for this article
was to display how two species of coral reef fishes in the Great
Barrier Reef are failing to acclimate to higher sea surface
temperatures. This was plainly stated in both the abstract and
introduction of the article.
The hypotheses of the three articles varied greatly. Dias stated
that the change in the global climate has led to rising sea
surface temperatures and ocean acidification, which jeopardized
coral reef survival. With this sentence, Dias made it clear why
his study efforts were so urgent. Nilsson followed a similar
pattern when he clearly stated his concerns for the inability of
coral reef fishes to acclimate to rising water temperatures.
De'ath’s hypothesis was stated in the abstract, which said that
his study suggested that the increasing thermal stress may be
depleting the ability of Great Barrier Reef corals to deposit
calcium carbonate. Thus, the hypotheses of all three articles
were given. I also found that Dias, De'ath, and Nilsson all had
a nice way of arranging their data, which allowed the
information to build to what the experimental design included
and what the researchers were hoping to accomplish from this
experiment.
Methods
The sample selection among the three articles showed great
contrast. The Dias paper used nine reef-forming coral species,
while De'ath’s experiment studied 328 colonies of coral from
the same genus, Porites, which is a stony coral. Nilsson studied
adults of two species of coral reef fishes. For Dias’s paper, the
methods were easy to follow and seemed easy to repeat, while
De’ath’s methods were harder to follow, for the details did not

appear to all be listed. The Methods section of the Nilsson
article was both valid and delivered with enough detail that
another group could perform most of this study again. Only
most of the experiment, because although the article listed when
and where the experiment was conducted, the number of each
species of adult coral reef fish caught and analyzed was not
given in the Methods sections of the paper. This information is
crucial, because a small sample size could invalidate the data,
while a large sample size could support the data more
accurately. Furthermore, if one species of coral reef fishes had
a much larger or small number than the other species, the data
would also not be well represented in the results found by this
study. The number of samples for both of the other articles
were given.
While some articles had strong Methods sections, others were
missing key components. The experimental design for the three
articles chosen all seemed valid. De’ath’s study seemed valid
for the experiment being conducted, though I am unsure that
this study could be repeated using the paper alone. The
experimental design did make sense overall, in that Porites is
commonly chosen for sclerochronological analyses because they
have annual density bands that are widely distributed. Portites
coral also has the capability of growing for hundreds of years,
so choosing this genus of coral for a large analysis made sense.
Using the three growth parameters De’ath mentioned—skeletal
density, calcification rate, and annual extension rate—are good
parameters to look at in a genus of coral that has such a long
life span. Dias justified every step of his experimental design,
making it easy to repeat the process. For instance, Dias utilized
contrasting morphologies, because they have different
susceptibilities to thermal stress, giving the overall results more
credit. These corals were held in captivity for several years,
giving the researchers knowledge of the corals’ thermal history.
Twenty fragments were cut from each of the nine species, half
of which were used as a control. Sources of variations were
eliminated in this process by cutting only one coral from each

colony. These methods appear valid, and each is given a reason
as to why a scientist would conduct the experiment in this way,
making the overall flow of the methods logical and easy to
follow. This was similar to De’ath’s paper in that De’ath listed
the parameters used to test the samples, and he mention that
Porites has such a long lifespan, so these types of corals have
been proven to record environmental changes within their
skeletons. This statement justified why De’ath chose this coral
and explained why these particular parameters were chosen.
However, he did not specify how to conduct these analyses.
Also, although the data was collected within a two-month period
for both Dias and Nilsson’s experiments, De’ath’s experiments
was a composite collection from the years 1900-2006,
containing over sixteen thousand annuals records with corals
ranging from ten to 436 years old. Hence, the broad range of
years the specimens were collected was overwhelming, not to
mention the three growth parameters the paper mentioned but
again failed to explain. Lastly, for Nilsson’s experimental
design, I found that it was carried out well, using adults of the
two species of coral reef fishes and varying temperatures that
supported their hypothesis. However, not including the number
of each species caught negates the data to a certain degree.
Overall, I found that the Methods section of Nilsson’s paper
was logical, but it did not contain details that were pertinent to
this experiment, whereas Dias included all pertinent information
and De’ath failed to include how he performed the parameters
that were chosen.
Results
Since the concentration for each study varied, the results were
also quite different in composition. Dias et al. (2018) found
that injury—whether present or absent—had no impact on the
death or growth rate of the coral fragments studied. The
researchers determined that the true factors that impacted death
and growth rate of the corals analyzed were temperature and the
coral species itself. These results were illustrated using tables
and figures. Table 1 was difficult to follow, because some of

the columns were abbreviated using terms not explained within
the content of the article. However, the numbers in the table
coincide with the text, showing that injury did not impact the
growth or mortality rate of the coral fragments used in this
experiment. The results found in De'ath’s paper were easy to
follow, but showed that the cause of decline in coral
populations in the GBR were still not known. Within the
Results section of the articles written by Nilsson and De’ath, I
saw that the figures and tables matched the text without
repeating the same information to the audience. The figures and
tables were accurate with what the text had previously stated,
showing P values that were statistically significant, and the data
was very easy to understand. The table in Dias’s article could
have been better presented if the abbreviations used had some
type of key that denote what each header meant. I did not find
any discrepancies among the figures and text of the three
articles as far as percentages were concerned.
The results found in the three studies did test the hypothesis of
the researchers. For Dias’s paper, these results were shown in
Figure 1, which illustrated that as temperature increased, the
mortality rate of coral species also increased. As Dias
mentioned in the Abstract section, there were two coral species
that survived this experiment, Turbinaria reniformis and
Galaxea fascicularis. However, the results of the Nilsson paper
were to test the hypothesis of the researchers in that study,
which was that after a given number of days in varying
temperatures, the coral reef fishes studied would fail to
acclimate to those temperature changes. Again, in the third
paper, De’ath’s results tested the hypothesis, including 328
colonies of massive corals form 69 various reefs, which made
the results more broad.
Discussion
I found that none of the three articles repeated the same
information in both the figures and the text of the article. From
the Discussion section of the Dias paper, a reader could tell the
main points of the article, which were to show that there was

variability in the susceptibility to thermal stress of different
coral reef species. These coral reef species had the lowest
mortality, partial mortality, and levels of bleaching at 26
degrees Celsius, while their growth rate was at its zenith at this
temperature. Dias found that the regeneration rate of corals
generally increased as the temperature increased. These results
also show that the bleaching resistance capacity of most of the
corals analyzed was overcome at 32 degrees Celsius. Because
this paper is so new—published in 2018—I could not find its
interpretations to be supported by other research. However, the
article does list the direction in which the research is headed
and lists other studies similar to this one.
The findings of De'ath’s and Dias’s articles were supported by
each other, as well as many other articles over coral reef
ecology. However, I found very few articles that supported the
conclusions drawn by Nilsson (2010) about the effects
increasing temperatures had on two species of coral reef fishes,
which was a weakness for the paper. In all three journal
articles, I found that the interpretation of data was logical.
Conclusion
Summary
The three articles, overall, had both strengths and weaknesses.
For instance, all three papers were peer reviewed. Nilsson’s
paper had few other articles that backed up its findings, while
Dias and De’ath backed up each other’s paper. The article by
Dias (2018) was very recent, which made it one of the newest
published papers in its field, while the De'ath and Nilsson
papers were a few years older. Although the Dias paper had
tables and figures that were not entirely straight forward, the
content of the article itself was very easy to follow. Each
section within the article was set apart, whereas in the article by
De’ath, the sections (introduction, methods, etc.) were not
separated from each other. The Dias and De’ath papers had
appropriate titles, while Nilsson’s title was too long. The
abstracts and introductions match for all three articles. The
pace for the Dias article is great, leading the audience straight

into the hypothesis and objectives for the analysis, while De’ath
failed to separate his paper into different sections. Overall,
Dias and the other researchers took many steps to ensure
accurate results, and the Methods section of this article is
explained well enough to be repeated, which differed from
Nilsson’s paper in that Nilsson did not include his sample size.
When it came to reproducing the experiment, Dias included all
pertinent information, but De’ath failed to include how he
performed the parameters that were chosen.
De’ath had a concise paper with text that accurately related to
the figures mentioned. Overall, the article was concise in its
findings, but not as easy to follow as it could have been if the
proper sections and subsections had been utilized. The title
seemed appropriate, and readers know from the statement of
purpose and the introduction that the primary goal of this study
was to help determine what is causing the decline in corals’
ability to lay down a calcium carbonate skeleton to more
efficiently build coral reef ecosystems. The De'ath paper used a
large sample selection, which made the results seem more
inclusive as opposed to Dias’s samples size of nine species of
corals using twenty fragments of each species. Nilsson’s article
had excellent figures that were easy to interpret, in contrast to
Dias’s paper that did not explain what some of the abbreviations
meant in the tables. The strengths and weaknesses of the three
papers varied greatly.
Significance
When evaluating the role these articles play in the world,
striking similarities were found. Nilsson’s article showed
primary concerns toward two populations of fish species that
lived in the Great Barrier Reef, while De’ath and Dias wrote
papers over the reactions of different coral species to increasing
surface temperatures. De’ath’s article has practical significance
similar to the Dias paper, in that major ecosystems are dying as
a result of rising ocean surface temperatures, and these
researchers tried to find ways to explain these issues. Nilsson’s
article examined whether or not increasing temperatures

reduced the hypoxia tolerance of coral reef fishes. It has been
cited sixty-nine times, cited in papers involving hypoxia
tolerance of coral reef fishes, how temperature and hypoxia play
a role in respiratory performance of certain tropical fishes, and
many other similar studies (Nilsson et al.). De'ath et al. has
been cited twenty-four times, which sparked interest for similar
research in the Great Barrier Reef in the last decade. The Dias
et al. paper was only published in 2018, so not many other
researchers have cited this paper yet. This can be seen as a
potential problem; however, the article was peer-reviewed by
individuals who are well-educated in this particular field. The
currency of the Dias article may also be seen as a good
attribute, showing that this information was some of the newest
in its field of interest. The research among all three articles has
significance to today’s society, in that the bleaching of coral
reefs has become a growing problem, and without more research
to determine what factors are causing this issue, large hypoxic
zones in aquatic ecosystems may result.
Overall, all three of these articles illustrated environmental
significance. Human survival depends on the biodiversity of
plants and animals, and many animals live in these coral reef
ecosystems.
Works Cited
De'ath, G., Lough, J., & Fabricius, K. (2009). Declining Coral
Calcification on the Great Barrier Reef. Science, 323(5910),
116-119. doi: 10.1126/science.1165283
Dias, M., Ferreira, A., Gouveia, R., Cereja, R., & Vinagre, C.
(2018). Mortality, growth and regeneration following
fragmentation of reef-forming corals under thermal
stress. Journal of Sea Research, 141, 71-82. doi:
10.1016/j.seares.2018.08.008
Nilsson, G., Östlund-Nilsson, S., & Munday, P. (2010). Effects
of elevated temperature on coral reef fishes: Loss of hypoxia
tolerance and inability to acclimate. Comparative Biochemistry

and Physiology Part A: Molecular & Integrative
Physiology, 156(4), 389-393. doi: 10.1016/j.cbpa.2010.03.009
[Type here]
2
9
QUESTION 1
Naïve Bayesian Classifiers are among the most successful
known algorithms for learning to classify text documents and
they are used to detect fraud.
True
False
QUESTION 2
What is the most accurate statement that describe what “Leaf
Nodes” are in a decision tree?
Decisions that are often used as components in ensemble
techniques predictive models which will all vote.
The nature of the variable, you may need to include an equal to
component on one branch.
Decision at the end of the last branch on the tree. These
represent the outcome of all the prior decisions. They are the
class labels, or the segment in which all observations that

follow the path to the leaf would be placed.
None of the above.
QUESTION 3
What is confusion matrix and what are the values, rates, metrics
associated with the matrix?
ANS:
QUESTION 4
Decision Trees are a flexible method very commonly deployed
in data mining applications. There are two types of decision
trees. What are the two trees and the descriptions of both?
ANS:
QUESTION 5
Decision Trees take only categorical variables. They cannot
handle many distinct values such as the zip code in the data and
are limited to only one attribute.
True
False
QUESTION 6
Time Series Analysis is the analysis of sequential data across
equally spaced units of time. Time Series is a basic research
methodology in which data for one or more variables are
collected for many observations at different time periods. What
are the two main objectives in Time Series Analysis?
ANS:
QUESTION 7

In the Box-Jenkins model, Autoregressive (AR) models can be
coupled with moving average (MA) models to form:
The input for the model that trend and are seasonality adjusted
in time series and the output that provides an expected future
value of the time series.
The next stage to determine the p and q in the ARIMA (p, d, q)
model.
The assurance level of obtaining the highest forecasting
accuracy possible in terms of the variables on which the
forecast is based.
A general and useful class of time series models called
Autoregressive Moving Average (ARMA) models.
QUESTION 8
The Information Gain is defined as the difference between the
base entropy and the conditional entropy of the attribute.
True
False
QUESTION 9
"Pure enough" usually means that other information can be
gained by splitting on other attributes.
True
False
QUESTION 10

The key application of Time Series Analysis is in forecasting.
In regard to Time Series data, what are the 6 fields in society
that Time Series data provide useful information about the
physical, biological, social or economic systems generating the
time series?
ANS:
Overweight, obesity and risk of liver cancer: a meta-analysis of
cohort studies
SC Larsson*,1 and A Wolk1
1
Division of Nutritional Epidemiology, The National Institute of
Environmental Medicine, Karolinska Institutet, PO Box 210,
Stockholm SE-17177, Sweden
Cohort studies of excess body weight and risk of liver cancer
were identified for a meta-analysis by searching MEDLINE and
EMBASE
databases from 1966 to June 2007 and the reference lists of
retrieved articles. Results from individual studies were
combined using a
random-effects model. We identified 11 cohort studies, of which
seven on overweight (with a total of 5037 cases) and 10 on
obesity
(with 6042 cases) were suitable for meta-analysis. Compared
with persons of normal weight, the summary relative risks of
liver
cancer were 1.17 (95% confidence interval (CI): 1.02 – 1.34)
for those who were overweight and 1.89 (95% CI: 1.51 – 2.36)
for those

who were obese. This meta-analysis finds that excess body
weight is associated with an increased risk of liver cancer.
British Journal of Cancer (2007) 97, 1005 – 1008.
doi:10.1038/sj.bjc.6603932 www.bjcancer.com
Published online 14 August 2007
& 2007 Cancer Research UK
Keywords: body mass index; cohort studies; liver cancer; meta-
analysis; obesity; review
��
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�
�

�
�
�
�
�
�
�
Although relatively rare in the United States and other
developed
countries, liver cancer is the third most common cause of
death from cancer worldwide (Parkin et al, 2005). It is rarely
detected early and is often fatal within a few months of
diagnosis.
The 5-year survival rate is only about 6 – 11% (Coleman et al,
2003; Ries et al, 2006). The age-adjusted incidence and
mortality
rates of liver cancer have been increasing rapidly in the
United States since the mid-1980s (Ries et al, 2006). While
approximately half of this increase can be attributable to
hepatitis
C virus infection, a minimal or no increase has been related to
hepatitis B virus and alcoholic liver disease (El-Serag and
Mason,
2000; Hassan et al, 2002). Given that about half of the increase
in liver cancer incidence is not related to hepatitis, the major
risk factor in a significant proportion of the cases has yet to be
identified.
Coinciding with the rising incidence of liver cancer, the
prevalence of obesity has been increasing markedly over the
past
two decades worldwide (Larsson and Wolk, 2006). Obesity has
been recognised as a risk factor for several malignancies,
including
cancer of the breast (in premenopausal women), endometrium,

kidney (renal cell), colon, pancreas, gallbladder, and esophagus
(adenocarcinoma) (IARC, 2002; Larsson et al, 2007; Larsson
and
Wolk, 2007). Accumulating epidemiologic evidence also
indicates
that excess body weight may be a risk factor for liver cancer,
but
the evidence has not been quantitatively summarised. We have
therefore quantitatively assessed the associations of overweight
and obesity with liver cancer risk in a meta-analysis of cohort
studies.
MATERIALS AND METHODS
Study selection
A literature search was conducted in the MEDLINE and
EMBASE
databases for pertinent studies published in any language from
1966 to June 2007. We used the keywords ‘body mass index’,
‘BMI’,
or ‘obesity’ in combination with ‘hepatocellular carcinoma’,
‘liver
cancer’, or ‘liver neoplasm’. Moreover, we manually reviewed
the
reference lists of retrieved articles to search for more studies.
Studies were included in the meta-analysis if they fulfilled the
following criteria: (1) cohort study in which liver cancer
incidence
or mortality was an outcome; (2) the exposure of interest was
overweight and/or obesity defined by body mass index (BMI)
(the
weight in kilograms divided by the square of height in meters);
and
(3) relative risk estimates (rate ratio or standardized incidence

ratio) with corresponding 95% confidence intervals (CIs) were
reported.
Data extraction
For each study, the following information was extracted: first
author’s last name; publication year; country in which the study
was performed; sample size; method of assessing weight and
height; type of outcome (incidence or mortality); variables
adjusted for in the analysis; and the relative risks with 95% CIs
for overweight and obesity vs normal weight. From each study,
we
extracted the most fully adjusted relative risks.
Statistical analysis
The relative risks and corresponding standard errors (derived
from the CIs) from individual studies were logarithmically
transformed to stabilize variances and normalize the
distributions.
We calculated summary relative risks for overweight (defined as
Received 13 June 2007; revised 12 July 2007; accepted 17 July
2007;
published online 14 August 2007
*Correspondence: Dr SC Larsson; E-mail: [email protected]
British Journal of Cancer (2007) 97, 1005 – 1008
& 2007 Cancer Research UK All rights reserved 0007 – 0920/07
$30.00
www.bjcancer.com
E
p

id
e
m
io
lo
g
y
http://dx.doi.org/10.1038/sj.bjc.6603932
http://www.bjcancer.com
mailto:[email protected]
http://www.bjcancer.com
BMI 25 – 30 kg m�2) and obesity (BMI X30 kg m�2) vs
normal
weight (BMI 18.5 – 24.9 kg m�2). For two studies (Calle et al,
2003;
Kuriyama et al, 2005) that reported relative risks for two
categories
of BMI that fell into the category representing overweight or
obesity, we pooled the relative risks and used the pooled
estimate
in the meta-analysis. Study-specific relative risks were
combined
using the DerSimonian and Laird random-effects model
(DerSimonian and Laird, 1986). Thus, each summary relative
risk
was a weighted average of the study-specific relative risk,
where the
weight for each study is the inverse of the sum of the within-
study
variance for that study, and the between-study variance.

Statistical heterogeneity among studies was evaluated with the
Q
and I2 statistics (Higgins and Thompson, 2002). For the Q
statistic,
statistical significance was set at Po0.1. We used funnel plots
(i.e. plots of study results against precision) to assess
publication
bias, and tested the symmetry of the funnel plot using Egger’s
test
(Egger et al, 1997).
Results are presented graphically, whereby squares represent
study-specific relative risks and diamonds represent summary
relative risks. The area of each square is proportional to the
inverse
of the variance of the logarithm of the relative risk; 95% CIs for
individual studies are represented by horizontal lines and for the
summary estimates by the width of the diamonds. Statistical
analyses were performed with Stata, version 9.0 (StataCorp,
College Station, TX, USA).
Population attributable risk (PAR) for liver cancer was
estimated for individuals with excess body weight
(BMIX25 kg m�2) compared to those of normal weight
(BMIo25 kg m�2). The PAR describes the theoretic proportion
of cases that would be prevented if all individuals were moved
into
the exposure level associated with the lowest risk for that
factor.
The PAR (PAR%) was calculated as: PAR% ¼ (p � [RR – 1]/
[p � (RR – 1) þ 1]) � 100%, where p represents the prevalence
in
the population and RR the relative risk. Prevalence data were
obtained from the National Health and Nutrition Examination
Survey that assessed the prevalence of overweight and obesity

in a
representative sample of the US population (Ogden et al, 2006).
In
that survey, 39.7% of the men were overweight and 31.1% were
obese. Among women, 28.6% were overweight and 33.2% were
obese. PARs were calculated for the overweight and obese
categories using the obtained summary relative risks, and then
summarized across the two categories for men and women
separately.
RESULTS
We identified 11 eligible cohort studies (Møller et al, 1994;
Wolk
et al, 2001; Nair et al, 2002; Calle et al, 2003; Samanic et al,
2004,
2006; Batty et al, 2005; Kuriyama et al, 2005; Oh et al, 2005;
Rapp
et al, 2005; N’Kontchou et al, 2006), of which 7 on overweight
(with
a total of 5037 cases) (Calle et al, 2003; Batty et al, 2005;
Kuriyama
et al, 2005; Oh et al, 2005; Rapp et al, 2005; N’Kontchou et al,
2006;
Samanic et al, 2006) and 10 on obesity (with a total of 6042
cases)
(Møller et al, 1994; Wolk et al, 2001; Nair et al, 2002; Calle et
al,
2003; Samanic et al, 2004, 2006; Batty et al, 2005; Oh et al,
2005;
Rapp et al, 2005; N’Kontchou et al, 2006) were suitable for
meta-
analysis. Characteristics of the studies are shown in Table 1.
Seven
studies were conducted in Europe, two in the United States, and

two in Asia. Weight and height were measured in six studies
and
self-reported in two studies; in three studies, obesity was
defined
by a discharge diagnosis of obesity. The outcome was incidence
of liver cancer in all but two studies (Calle et al, 2003; Batty et
al,
2005) in which the outcome was mortality from liver cancer.
Two
studies were based on standardized incidence ratio (Møller et al,
1994; Wolk et al, 2001). Two studies consisted of patients with
cirrhosis (Nair et al, 2002; N’Kontchou et al, 2006).
Relative risks of liver cancer for overweight and obese
individuals compared to those of normal weight for individual
studies (separately for men and women wherever this data were
available) and all studies combined are shown in Figure 1.
Meta-
analysis of all studies found that compared to individuals with
normal weight, those who were overweight or obese had a 17
and
89%, respectively, increased risk of liver cancer. There was
statistically significant heterogeneity among the results of
indivi-
dual studies (Figure 1). The summary relative risk for obesity
was
statistically significantly higher (P ¼ 0.03) for men (RR: 2.42;
95%
CI: 1.83 – 3.20; n ¼ 7 studies) than for women (RR: 1.67; 95%
CI:
Table 1 Characteristics of cohort studies included in the meta-
analysis
Study Country
No. of cases

(men/women) Study participants
Assessment of
exposure Adjustments
Møller et al (1994) Denmark 22/36 Men: 14 531
Women: 29 434
Discharge diagnosis of
obesity
Age
Wolk et al (2001) Sweden 15/13 Men: 8165
Women: 19 964
obesity
Age, calendar year
Nair et al (2002) USA 659a Men and women: 19 271a Measured
Age, sex, race, diabetes
Calle et al (2003) USA 620/345 Men: 404 576
Women: 495 477
Self-reported Age, race, education, marital status, smoking,
physical activity, aspirin use, estrogen-replacement
therapy (women), alcohol, dietary factors
Samanic et al (2004) USA 322 whites/38
blacks
White men: 3 668 486
Black men: 832 214

obesity
Age, calendar year
Kuriyama et al (2005) Japan 69/31 Men: 12 485
Women: 15 054
Self-reported Age, type of health insurance, smoking, intakes of
alcohol, meat, fish, fruits, vegetables, bean-paste
soup
b
Batty et al (2005) UK 51 Men: 18 403 Measured Age,
employment grade, marital status, physical
activity, smoking, other
c
Oh et al (2005) Korea 3347 Men: 781 283 Measured Age, area
of residence, family history of cancer,
smoking, exercise, alcohol
Rapp et al (2005) Austria 57 Men: 67 447 Measured Age,
occupational group, smoking
N’Kontchou et al (2006) France 220a Men and women: 771a
Measured Age, sex, cirrhosis cause, diabetes
Samanic et al (2006) Sweden 297 Men: 362 552 Measured Age,
smoking
a
Patients with cirrhosis.
b

Odds ratios for women were further adjusted for age at
menarche, age at end of first pregnancy, and menopausal status.
c
Other factors adjusted for
include disease at entry, weight loss in the last year, height-
adjusted FEV1, triceps skinfold thickness, blood pressure-
lowering medication, blood pressure, plasma cholesterol,
glucose intolerance, and diabetes.
Overweight, obesity and gallbladder cancer
SC Larsson and A Wolk
1006
British Journal of Cancer (2007) 97(7), 1005 – 1008 & 2007
Cancer Research UK
E
p
id
e
m
io
lo
g
y
1.37 – 2.03; n ¼ 3 studies). There was no evidence for

publication
bias on the funnel plot (data not shown) or by Egger’s test
(P ¼ 0.31 for overweight and P ¼ 0.21 for obesity).
In a sensitivity analysis excluding the two studies that consisted
of patients with cirrhosis (Nair et al, 2002; N’Kontchou et al,
2006),
the summary relative risks were 1.07 (95% CI: 1.01 – 1.15) for
overweight and 1.85 (95% CI: 1.44 – 2.37) for obesity. With
stratification by assessment of obesity, the summary relative
risks
for the association between obesity and liver cancer were 2.15
(95%
CI: 1.66 – 2.77) for studies based on measured or self-reported
weight and height (Nair et al, 2002; Calle et al, 2003; Batty et
al,
2005; Oh et al, 2005; Rapp et al, 2005; N’Kontchou et al, 2006;
Samanic et al, 2006) and 1.61 (95% CI: 1.14 – 2.27) for studies
based
on a discharge diagnosis of obesity (Møller et al, 1994; Wolk et
al,
2001; Samanic et al, 2004).
The PAR for excess body weight was calculated using the
estimates of prevalence in the United States (Ogden et al, 2006)
and the obtained summary relative risks of 1.17 and 1.89 for
overweight and obesity, respectively. We estimated that 28% of
liver cancer cases among men and 27% among women could be
attributable to excess body weight (BMIX25 kg m�2).
DISCUSSION
This is the first meta-analysis on overweight and obesity in
relation
to liver cancer risk and it indicates that excess body weight is
associated with increased risk. Summary results showed that the

risk was 17 and 89% higher among persons who were
overweight
and obese, respectively, compared with those of normal weight.
The relation between obesity and liver cancer seemed to be
stronger in men than in women.
Although there was statistically significant heterogeneity among
study results, the relation between obesity and risk of liver
cancer
was consistent. Differences in the relative risk estimates were
largely in the magnitude rather than the direction of the
association. All but 1 out of the 14 relative risk estimates for
the association between obesity and liver cancer were above one
(ranging from 1.44 to 3.76), and 12 of these estimates were
statistically significant.
A potential limitation of this meta-analysis is that individual
studies may have failed to control for potential known or
unknown
confounders. The most important risk factors for the
development
of liver cancer are chronic infections with hepatitis B virus and
hepatitis C virus. Heavy, long-term alcohol consumption is also
a
risk factor (Yu and Yuan, 2004). None of the studies adjusted
for
hepatitis B or C virus infections, and only three (Calle et al,
2003;
Kuriyama et al, 2005; Oh et al, 2005) controlled for alcohol
intake.
It is unlikely, however, that these risk factors are strongly
related
to body weight and entirely explain the observed relationship
between excess body weight and liver cancer risk. Another
limitation is that we could not examine whether the association

between excess body weight and liver cancer was modified by
hepatitis virus infections and alcohol intake because the studies
included in this meta-analysis did not provide results stratified
by
these factors.
As this meta-analysis was based on published studies, possible
publication bias could have affected the results. However,
neither
funnel plots nor formal statistical tests showed evidence for
publication bias.
The observed increased risk of liver cancer associated with
excess body weight may be mediated through the development
of
non-alcoholic fatty liver disease (NAFLD), a chronic liver
disease
that occurs in non-drinkers. NAFLD is characterized by a
spectrum of liver tissue changes, ranging from accumulation of
fat in the liver to non-alcoholic steatohepatitis (NASH),
cirrhosis,
and liver cancer at the most extreme end of the spectrum. Up to
90% of obese individuals have some degree of fatty liver, and
approximately 25 – 30% have NASH (Neuschwander-Tetri and
Caldwell, 2003).
Relative risk
0.5 0.7 1.0 1.5 2.0 3.0 4.5 6.0
Relative risk
(95% CI)
Overweight
Calle et al (2003) (M) 1.13 (0.94–1.34)
Calle et al (2003) (W) 1.02 (0.80–1.31)
Kuriyama et al (2005) (M) 0.91 (0.52–1.59)

Kuriyama et al (2005) (W) 1.13 (0.57–2.27)
Batty et al (2005) (M) 0.99 (0.53–1.88)
Oh et al (2005) (M) 1.05 (0.97–1.14)
Rapp et al (2005) (M) 1.32 (0.73–2.37)
N'Kontchou et al (2006) (M/W) 2.00 (1.40–2.70)
Samanic et al (2006) (M) 1.29 (1.00–1.68)
Summary estimate 1.17 (1.02–1.34)
Obesity
Møller et al (1994) (M) 1.90 (1.20–2.90)
Møller et al (1994) (W) 1.90 (1.40–2.70)
Wolk et al (2001) (M) 3.60 (2.00–6.00)
Wolk et al (2001) (W) 1.70 (1.10–2.50)
Nair et al (2002) (M/W ) 1.65 (1.22–2.22)
Calle et al (2003) (M) 2.41 (1.92–3.01)
Calle et al (2003) (W) 1.47 (1.08–2.00)
Samanic et al (2004) (M, whites) 1.44 (1.28–1.61)
Samanic et al (2004) (M, blacks) 0.68 (0.49–0.94)
Batty et al (2005) (M) 3.76 (1.36–10.4)
Oh et al (2005) (M) 1.56 (1.15–2.12)
Rapp et al (2005) (M) 1.67 (0.75–3.72)
N'Kontchou et al (2006) (M/W) 2.80 (2.00–4.00)
Samanic et al (2006) (M) 3.62 (2.62–5.00)
Summary estimate 1.89 (1.51–2.36)
Figure 1 Relative risks of liver cancer associated with
overweight and obesity. Relative risk estimates are for
overweight and obese persons compared
with normal weight persons. Tests for heterogeneity:
overweight, Q ¼ 16.83, P ¼ 0.03; I2 ¼ 52.5%; obesity, Q ¼
88.03, Po0.001; I2 ¼ 86.4%. M ¼ men;
W ¼ women.

1007
British Journal of Cancer (2007) 97(7), 1005 – 1008& 2007
Cancer Research UK
E
p
id
e
m
io
lo
g
y
In summary, this meta-analysis supports evidence of an
increased risk of liver cancer among overweight and obese
persons. These findings indicate that liver cancer may, in part,
be prevented by maintaining a healthy body weight.
ACKNOWLEDGEMENTS
This work was supported by grants from the Swedish Cancer
Society.
REFERENCES
Batty GD, Shipley MJ, Jarrett RJ, Breeze E, Marmot MG, Smith

GD (2005)
Obesity and overweight in relation to organ-specific cancer
mortality in
London (UK): findings from the original Whitehall study. Int J
Obes
(Lond) 29: 1267 – 1274
Calle EE, Rodriguez C, Walker-Thurmond K, Thun MJ (2003)
Overweight,
obesity, and mortality from cancer in a prospectively studied
cohort of
US adults. N Engl J Med 348: 1625 – 1638
Coleman MP, Gatta G, Verdecchia A, Esteve J, Sant M, Storm
H, Allemani
C, Ciccolallo L, Santaquilani M, Berrino F (2003) EUROCARE-
3
summary: cancer survival in Europe at the end of the 20th
century.
Ann Oncol 14(Suppl 5): v128 – v149
DerSimonian R, Laird N (1986) Meta-analysis in clinical trials.
Control Clin
Trials 7: 177 – 188
Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias
in meta-
analysis detected by a simple, graphical test. BMJ 315: 629 –
634
El-Serag HB, Mason AC (2000) Risk factors for the rising rates
of
primary liver cancer in the United States. Arch Intern Med 160:
3227 – 3230
Hassan MM, Frome A, Patt YZ, El-Serag HB (2002) Rising

prevalence of
hepatitis C virus infection among patients recently diagnosed
with
hepatocellular carcinoma in the United States. J Clin
Gastroenterol 35:
266 – 269
Higgins JP, Thompson SG (2002) Quantifying heterogeneity in
a meta-
analysis. Stat Med 21: 1539 – 1558
IARC (2002) IARC handbooks of cancer prevention. Weight
control and
physical activity. Vol. 6, Lyon, France: IARC Press
Kuriyama S, Tsubono Y, Hozawa A, Shimazu T, Suzuki Y,
Koizumi Y,
Ohmori K, Nishino Y, Tsuji I (2005) Obesity and risk of cancer
in Japan.
Int J Cancer 113: 148 – 157
Larsson SC, Orsini N, Wolk A (2007) Body mass index and
pancreatic
cancer risk: A meta-analysis of prospective studies. Int J Cancer
120:
1993 – 1998
Larsson SC, Wolk A (2006) Epidemiology of obesity and
diabetes:
prevalence and trends. In Obesity and Diabetes, Mantzoros C
(ed) pp
15 – 36. Boston: Humana Press
Larsson SC, Wolk A (2007) Obesity and the risk of gallbladder
cancer: a
meta-analysis. Br J Cancer 96: 1457 – 1461

Møller H, Mellemgaard A, Lindvig K, Olsen JH (1994) Obesity
and cancer
risk: a Danish record-linkage study. Eur J Cancer 30A: 344 –
350
Nair S, Mason A, Eason J, Loss G, Perrillo RP (2002) Is obesity
an
independent risk factor for hepatocellular carcinoma in
cirrhosis?
Hepatology 36: 150 – 155
Neuschwander-Tetri BA, Caldwell SH (2003) Nonalcoholic
steatohepatitis:
summary of an AASLD Single Topic Conference. Hepatology
37:
1202 – 1219
N’Kontchou G, Paries J, Htar MT, Ganne-Carrie N, Costentin L,
Grando-
Lemaire V, Trinchet JC, Beaugrand M (2006) Risk factors for
hepatocellular carcinoma in patients with alcoholic or viral C
cirrhosis.
Clin Gastroenterol Hepatol 4: 1062 – 1068
Ogden CL, Carroll MD, Curtin LR, McDowell MA, Tabak CJ,
Flegal KM
(2006) Prevalence of overweight and obesity in the United
States, 1999 –
2004. JAMA 295: 1549 – 1555
Oh SW, Yoon YS, Shin SA (2005) Effects of excess weight on
cancer
incidences depending on cancer sites and histologic findings
among
men: Korea national health insurance corporation study. J Clin

Oncol 23:
4742 – 4754
Parkin DM, Bray F, Ferlay J, Pisani P (2005) Global cancer
statistics, 2002.
CA Cancer J Clin 55: 74 – 108
Rapp K, Schroeder J, Klenk J, Stoehr S, Ulmer H, Concin H,
Diem G,
Oberaigner W, Weiland SK (2005) Obesity and incidence of
cancer: a
large cohort study of over 145 000 adults in Austria. Br J
Cancer 93:
1062 – 1067
Ries LAG, Harkins D, Krapcho M, Mariotto A, Miller AB,
Feuer EJ, Clegg L,
Eisner MP, Horner MJ, Howlader N, Hayat M, Hankey BF,
Edwards BK
(2006) SEER cancer statistics review, 1975 – 2003. Bethesda,
MD: National
Cancer Institute, 2006 http://seer.cancer.gov/csr/1975_2003/
(accessed 30
June 2007)
Samanic C, Chow WH, Gridley G, Jarvholm B, Fraumeni JF
(2006) Relation
of body mass index to cancer risk in 362 552 Swedish men.
Cancer Causes
Control 17: 901 – 909
Samanic C, Gridley G, Chow WH, Lubin J, Hoover RN,
Fraumeni Jr JF
(2004) Obesity and cancer risk among white and black United
States
veterans. Cancer Causes Control 15: 35 – 43

Wolk A, Gridley G, Svensson M, Nyren O, McLaughlin JK,
Fraumeni JF,
Adam HO (2001) A prospective study of obesity and cancer risk
(Sweden). Cancer Causes Control 12: 13 – 21
Yu MC, Yuan JM (2004) Environmental factors and risk for
hepatocellular
carcinoma. Gastroenterology 127: S72 – S78
1008
British Journal of Cancer (2007) 97(7), 1005 – 1008 & 2007
Cancer Research UK
E
p
id
e
m
io
lo
g
y
http://seer.cancer.gov/csr/1975_2003/Overweight, obesity and
risk of liver cancer: a meta-analysis of cohort
studiesMATERIALS AND METHODSStudy selectionData
extractionStatistical analysisRESULTSDISCUSSIONFigure 1

Relative risks of liver cancer associated with overweight and
obesity.Table 1 Characteristics of cohort studies included in the
meta-analysisACKNOWLEDGEMENTSREFERENCES
Hindawi Publishing Corporation
ISRN Preventive Medicine
Volume 2013, Article ID 680536, 16 pages
http://dx.doi.org/10.5402/2013/680536
Review Article
The Association between Obesity and Cancer Risk:
A Meta-Analysis of Observational Studies from 1985 to 2011
M. Dobbins,1 K. Decorby,1 and B. C. K. Choi2
1 School of Nursing, McMaster University, 1280 Main Street
West, 3N25G, Hamilton, ON, Canada L8N 3Z5
2Department of Epidemiology and Community Medicine,
University of Ottawa, Ottawa, ON, Canada K1H 8M5
Correspondence should be addressed to M. Dobbins;
[email protected]
Received 10 January 2013; Accepted 10 February 2013
Academic Editors: C. R. González Bonilla, C. Grandjean, and F.
Mawas
Copyright © 2013 M. Dobbins et al. This is an open access
article distributed under the Creative Commons Attribution
License,
which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
Background. Cancer and cardiovascular diseases are the leading

causes of mortality and morbidity worldwide. The purpose of
this meta-analysis is to synthesize the evidence evaluating the
association between obesity and 13 cancers shown previously to
be
significantly associated with obesity. Methods. Relevant papers
from a previously conducted review were included in this paper.
In addition, database searches of Medline and Embase identified
studies published from the date of the search conducted for the
previous review (January, 2007) until May, 2011. The reference
lists of relevant studies and systematic reviews were screened to
identify additional studies. Relevance assessment, quality
assessment, and data extraction for each study were conducted
by two
reviewers independently. Meta-analysis was performed for men
and women separately using DerSimonian and Laird’s random
effectsmodel.Results. A total of 98 studies conducted in 18
countries from 1985 to 2011 were included.Data extractionwas
completed
on the 57 studies judged to be of strong and moderate
methodological quality. Results illustrated that obese men were
at higher
risk for developing colon (Risk Ratio (RR), 1.57), renal (1.57),
gallbladder (1.47), pancreatic (1.36), andmalignant melanoma
cancers
(1.26). Obese women were at higher risk for esophageal
adenocarcinoma (2.04), endometrial (1.85), gallbladder (1.82),
renal (1.72),
pancreatic (1.34), leukemia (1.32), postmenopausal breast
(1.25), and colon cancers (1.19). Conclusions. The results of
this meta-
analysis illustrate a significant, positive, and, for some cancers,
strong association between obesity and cancer incidence. Given
that
approximately 23% of Canadians are obese, a significant
proportion of cancer in Canada could be avoided if obesity was
eliminated

or significantly reduced.
1. Introduction
Chronic diseases are the leading cause of mortality and mor-
bidity and contribute significantly to the overall health expen-
ditures from both a societal perspective as well as an individ-
ual one [1]. Common chronic diseases include heart disease,
stroke, cancer, emphysema, diabetes, and osteoporosis. Fur-
thermore, cancer and cardiovascular diseases are the leading
causes of mortality andmorbidity worldwide [2], with cancer
expected to result in 75,000 deaths per year in Canada [3] and
571,950 in the US [4].
While the leading risk factor for cancer continues to be
tobacco use, evidence shows that obesemen andwomen have
a greater likelihood of developing and dying from cancer than
those who are not [2, 5–7]. Obesity is defined as a Body Mass
Index (BMI) of 30 kg/m2 or greater. Approximately a quarter
of Canadianmen and women are considered obese [9].There
are many contributing factors to obesity such as physical
inactivity, unhealthy diet, genetics, and others such as meta-
bolic, environmental, social, economic, and psychological
factors. It is estimated that globally, every year, three to four
million cases of cancer could be prevented by eating healthier
and being more physically active [2].
With respect to morbidity, research demonstrates that
obesity increases the risk of cancers of the esophagus, breast
(postmenopausal), endometrium, colon and rectum, kidney,
pancreas, thyroid, gallbladder, and possibly other cancers
as well [8]. Other evidence suggests that obesity leads to
an increased risk for thirteen cancers including esophageal
adenocarcinoma, thyroid, colon, rectal, renal, endometrial,
pancreatic, gallbladder, postmenopausal breast, malignant

melanoma, multiple myeloma, leukemia, and non-Hodgkin
2 ISRN Preventive Medicine
lymphoma [6].The cost of obesity to the health care system in
Canada is estimated to be 4.3 billion per year [10]. Elsewhere
the cost of obesity to the health care system has been found
to represent 2.3% of annual hospital care costs [11]. Physical
inactivity, which contributes to obesity, has been associated
with significant health care expenditures for numerous
chronic diseases including cancer. Canadian research esti-
mates the economic burden of obesity as $2.1 billion in both
direct and indirect costs [12]. Data from that same study sug-
gests that a 10% decrease in inactivity would result in health
savings of $150 million per annum [13].
The purpose of this meta-analysis is to synthesize the evi-
dence evaluating the association between obesity and cancer.
Specifically the association between obesity and thirteen can-
cers shown previously [6] to be significantly associated with
obesity is the focus of this meta-analysis. In addition, among
cancers for which a statistically significant positive associa-
tion with obesity is observed, population-attributable risk for
the Canadian population is calculated and reported.
2. Methods
2.1. Search Strategy. Several activities were included in the
search strategy to identify primary studies for this paper. An
overview of the review process is depicted in Figure 1. First,
the primary studies included in Renehan et al., for which
significant associations between obesity and cancer were
reported, were identified and retrieved in full document
version. The search strategy employed by Renehan et al.

identified studies published between 1985 and 2007, indexed
in Medline and Embase. Three additional studies, noted by
Renehan et al., but published after their search was con-
ducted, and therefore not included in their review, were also
retrieved.
Second, a search for studies, conducted since the Renehan
et al. search for primary studies was completed, was con-
ducted using the following search strategy. The electronic
searches performed by Renehan et al. were replicated from
January 2007 to May 2011 in Medline and Embase through
OVID. Titles and abstractswere screened for relevance by two
independent reviewers. All references chosen by one or both
of the reviewers as being potentially relevant were selected for
further review and imported into Systematic Review Software
(SRS) from the Centre for Evidence-Based Medicine. All
potentially relevant studies were retrieved in PDF version.
Finally, the reference lists of relevant studies were screened
to identify additional potentially relevant studies, as well as
the reference lists of published systematic reviews or meta-
analyses on this topic.
2.2. Relevance Assessment. Two reviewers independently
screened all retrieved articles using an existing relevance
assessment tool. The following four criteria were used to
assess relevance to the research question: (1) is the article a
primary study; (2) is the focus of the study to explore the rela-
tionship between obesity and cancer incidence in adults aged
18 years and older; (3) does the study report on any one
or more of the following 13 cancers: esophageal adenocarci-
noma, thyroid, colon, renal, endometrial, gallbladder, rectal,
malignant melanoma, postmenopausal breast, pancreatic,
leukemia, multiple myeloma, and non-Hodgkin lymphoma;
(4) is data on the risk ratio or odds ratio between obesity and
incidence of any one or more of the 13 cancers in adults aged

18 years and over provided. Both reviewers independently
assessed each study for relevance and met to resolve discrep-
ancies through discussion. All articles referring to the same
study were considered as one study.
2.3. Methodological Quality Assessment. All studies judged to
be relevant were assessed for methodological quality by two
independent reviewers using an existing quality assessment
tool, based on the work of the Evidence-based Medicine
group at McMaster University. The assessment criteria con-
sisted of the following components: (1) research design; (2)
identification of comparison groups, (3) comparison groups
compared on important confounders at baseline, (4) out-
comes and exposures measured in the same way in all groups
being compared, (5) data collection tools shown to be valid,
(6) data collection tools shown to be reliable, (7) follow
up sufficiently long for the outcome(s) of interest; (8) com-
pleteness of followup, (9) temporality (exposure is known to
precede outcome), (10) dose-response gradient, (11) signifi-
cant baseline differences controlled for in the analysis, (12)
appropriate statistical tests for the research design, (13) pre-
cision of estimate of effect, and (14) sufficient detail describ-
ing study participants.
Points were assigned to each criterion according to an
a priori scale. Studies were given an overall score out of 20
possible points and were then classified into three categories:
strong, moderate, and weak. Studies receiving an overall
rating of 16 ormore points were rated as strong.Those obtain-
ing a score of 11–15 points received a rating of moderate, and
those obtaining a score of 10 or less were rated as weak.
Reviewers independently rated each study andmet to resolve
discrepancies in overall ratings through discussion. Stud-
ies deemed as being of weak methodological quality were
excluded from further analysis as the validity of the results
was questionable given themany limitations inherent in these

studies.
2.4. Data Extraction. Data on the population, studymethods,
and outcomeswere extracted for each study independently by
two reviewers. Discrepancies were resolved through discus-
sion.
2.5. Data Analysis. Estimates of association were measured
as a risk ratio and meta-analysis performed as a weighted
average of the log risk ratios and 95% confidence intervals.
In instances where odds ratios were reported for primary
studies (e.g., case-control studies) these were first converted
into risk ratios as suggested by Renehan et al., and then the
log risk ratios and coinciding 95% confidence intervals were
calculated.
Tests of heterogeneity were conducted among studies
using a Chi square procedure, where� < 0.05was considered
ISRN Preventive Medicine 3
Search strategy development
Electronic database (1723)Original lancet review (141) Review
reference lists (69)
Relevance assessment of full documents (202) = 101 unique
studies
3 studies excluded: outcomes reported
on subgroups
Quality assessment of relevant studies (98 projects)

Weak project accounts removed based
on quality assessment (41)
Colon (26)
Endometrial (17)
Esophageal (7)
Gallbladder (5)
Leukemia (4)
Malignant melanoma (7)
Non-hodgkin
lymphoma (14)
Rectal (19)
Thyroid (7)Postmenopausal
breast (27)
Renal (7)Pancreatic (23)
Multiple
myeloma (7)
Data extraction from relevant
project accounts (57) (1 strong; 56 moderate)
Figure 1: Overview of review process.
an indication of heterogeneity. Risk ratios were pooled using
the DerSimonian and Laird random effects model, given that
there was considerable variation in how independent and

dependent variables were measured across studies, and
because in most cases statistically significant heterogeneity
across study results was observed. Risk ratios/odds ratios
and 95% confidence intervals extracted from each primary
study, where those that adjusted for the greatest number of
confounding variables including behavioural factors. Anal-
yses were performed and reported separately by sex. In all
instances reference body mass index (BMI) was 18.5–24.99
which was compared to the BMI category of 30 or more.
2.6. Population-Attributable Risk. Population-attributable
risk (PAR) is the portion of the incidence of a disease in the
population that is due to exposure. PAR% is the percent of
the incidence of a disease in the population that is due to
exposure. It is the percent of the incidence of the disease
in the population that would be eliminated if exposure was
eliminated. PAR% = [P(RR − 1)]/[1 + P(RR − 1)], where, P
is the population prevalence of obesity, and RR is the pooled
risk ratio [12]. The most current obesity rates for men and
women available in Canada [9] were used to represent
population prevalence. The same formula was used to com-
pute the 95% confidence interval of PAR%.
3. Results
3.1. Search Results. In total 141 articles were identified from
Renehan et al., 1723 from the database searches and 69 from
reference lists; the total is 1933 papers. Of the 141 articles
included in Renehan et al., 94 articles were relevant to the
13 cancers included in this paper. The remaining 47 articles
from Renehan et al’s review explored the association between

obesity and cancers other than the thirteen of interest in
this paper, and therefore were excluded. Of the 1723 articles
identified in the database searches, 75 were judged to be rele-
vant. Finally, of the 69 articles identified from the reference
lists of relevant studies, 33 were deemed relevant. A total
of 202 articles were deemed relevant for this paper. When
papers were grouped according to independent studies, a
total of 101 unique studies were relevant. The Kappa score for
agreement between reviewers on relevance assessment was
0.835, indicating high agreement. Reasons studies found to be
not relevant were data not reported on the 13 cancers, and/or
the association between obesity and cancer in adults aged 18
years and older was not the focus of the study.
3.2. Methodological Quality Assessment. One hundred and
one studies were assessed for methodological quality. It
was identified during quality assessment that three studies
(judged to be of moderate quality) reported data in a way that
was inconsistent with the other studies, therefore could not
be aggregated with other studies.These studies were excluded
from the meta-analysis. Of the remaining 98 studies one was
assessed as being of strong methodological quality, 56 were
rated as moderate and 41 as weak. Data was extracted on
the 57 strong and moderate studies. The following criteria
distinguished strong andmoderate studies fromweak studies:
strong andmoderate studies tended to usemeasurement tools
with proven validity and reliability, demonstrated that obesity
preceded cancer incidence, and established a dose-response
gradient. Studies of weakmethodological quality rated poorly
on these criteria as well as research design. A summary of
the quality assessment of the 57 studies included in the meta-
analysis is presented in Figure 1. Quality assessment of the
weak studies not included in this publication can be requested
from the primary author.

3.3. Study Characteristics. The study designs included 43
cohort and 14 case-control studies published between 1985
and 2011. The majority of studies were conducted in the
United States (19), followed by Sweden (7), Norway (4), and
Japan (5). Most studies had follow-up rates of 80% or greater.
Education level ranged from primary school to postsec-
ondary. Among the cohort studies participants were followed
up between 6 and 39 years.
3.4. Association between Obesity and Cancer Incidence
3.4.1. Colon Cancer. Sixteen studies were included in the
meta-analysis assessing the association between obesity and
colon cancer among men (Figure 2(a)) and 13 studies among
women (Figure 2(b)). The pooled risk ratio illustrated that
obese men had an increased risk of colon cancer compared to
men of normal weight (RR 1.57; 95% confidence interval 1.48
to 1.65).There was no significant heterogeneity across studies
(heterogeneity� = 0.68).Thepooled risk ratio illustrated that
obesewomen had an increased risk of colon cancer compared
to women of normal weight (RR 1.19; 95% confidence interval
1.04 to 1.36). There was significant heterogeneity observed
across studies (heterogeneity � = 0.08).
3.4.2. Endometrial Cancer. The pooled risk ratio from 16
studies illustrated that obese women had an increased risk
of endometrial cancer compared to women of normal weight
(RR 1.85; 95% confidence interval 1.3 to 2.65) (Figure 3).There
was significant heterogeneity across studies (heterogeneity
� = 0.00001).
3.4.3. Esophageal Adenocarcinoma. There were 3 studies that
assessed the association between obesity and esophageal
adenocarcinoma in men (Figure 4(a)) and 4 studies among
women (Figure 4(b)). The pooled risk ratio demonstrated
that obese men did not have an increased risk for esophageal

adenocarcinoma compared to men of normal weight (RR
1.23; 95% confidence interval 0.58 to 2.60). There was sig-
nificant heterogeneity across studies (heterogeneity � =
0.02). Among obese women, however, the pooled risk ratio
indicated a sizable increased risk of esophageal adenocar-
cinoma compared to women of normal weight (RR 2.04;
95% confidence interval 1.18 to 3.55). There was significant
heterogeneity across studies (heterogeneity � = 0.003).
3.4.4. Gallbladder Cancer. Therewere 3 studies each assessing
the association between obesity and gallbladder cancer in
men (Figure 5(a)) and women (Figure 5(b)). The pooled risk
ratio illustrated that obese men had an increased risk of
gallbladder cancer compared to normal weight men (RR 1.47;
95% confidence interval 1.17 to 1.85).There was no significant
heterogeneity across studies (heterogeneity � = 0.77). Simi-
larly, among obesewomen the pooled risk ratio demonstrated
an increased risk of gallbladder cancer compared towomenof
normal weight (RR 1.82; 95% confidence interval 1.32 to 2.50).
There was no significant heterogeneity across studies (hetero-
geneity � = 0.15).
3.4.5. Leukemia. There were 2 studies that assessed the
association between obesity and leukemia among men
(Figure 6(a)) and women (Figure 6(b)). The pooled risk ratio
illustrated that obese men had no increased risk of leukemia
compared to men of normal weight (RR 1.16; 95% confidence
interval 0.88 to 1.52). For obese women, however, the pooled
risk ratio illustrated an increased risk of leukemia compared
to women of normal weight (RR 1.32; 95% confidence interval
1.08 to 1.60). There was no significant heterogeneity across
studies for either men or women.
3.4.6.MalignantMelanoma. Four studies assessed the associ-
ation between obesity and malignant melanoma among men
(Figure 7(a)) and 3 among women (Figure 7(b)). The pooled

risk ratio illustrated that obese men had an increased risk
of malignant melanoma compared to men of normal weight
(RR 1.26; 95% confidence interval 1.07 to 1.48). There was no
significant heterogeneity across studies (heterogeneity � =
0.28). The pooled risk ratio for obese women showed no
increased risk of malignant melanoma compared to women
of normal weight (RR 0.95; 95% confidence interval 0.84 to
1.07). There was no significant heterogeneity across studies
(heterogeneity � = 0.83).
Study or subgroup Log (risk ratio)
0.79299
0.712949
0.42527
0.39878
1.08181
0.4947
0.262364
0.47
0.412
0.48858
0.65233
0.33646
0.44469
0.55389
0.57661
0.35066
SE

0.274
0.13539
0.1105
0.036
0.55413
0.294
0.72371
0.2233
0.21
0.31799
0.264
0.354
0.198
0.081
0.1834
0.169
Weight
1.1%
4.3%
6.5%
61.4%
0.3%
0.9%
0.2%
1.6%
1.8%
0.8%

1.1%
0.6%
2.0%
12.1%
2.4%
2.8%
IV, random, 95% CI
2.21 [1.29, 3.78]
2.04 [1.56, 2.66]
1.53 [1.23, 1.90]
1.49 [1.39, 1.60]
2.95 [1.00, 8.74]
1.64 [0.92, 2.92]
1.30 [0.31, 5.37]
1.60 [1.03, 2.48]
1.51 [1.00, 2.28]
1.63 [0.87, 3.04]
1.92 [1.14, 3.22]
1.40 [0.70, 2.80]
1.56 [1.06, 2.30]
1.74 [1.48, 2.04]
1.78 [1.24, 2.55]
1.42 [1.02, 1.98]
Risk ratio Risk ratio
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
Test for overall effect: � = 15.88 (� < 0.00001)
Total (95% CI) 100.0% 1.57 [1.48, 1.65]
Heterogeneity: �2 = 0.00; �2 = 12.01, df = 15 (� = 0.68); �2 =
0%

Batty et al. (2009)
Campbell et al. (2007)
Chang et al. (2006, 2007)
Engeland et al. (2004)
Ford (1999)
Gaard et al. (1997)
Kuriyama et al. (2005)
Larsson et al. (2005)
MacInnis et al. (2004, 2006)
Nock et al. (2008)
Oh et al. (2005)
Otani et al. (2005)
Rapp et al. (2005)
Samanic et al. (2006)
Sun et al. (2008)
Stolzenberg-Solomon et al. (2002, 2008)
(a)
0.343589
0.2469
0.06766
1.0079
0.0198
0.8109
0.5481
0
0.4121
0.0099

0.418
SE
0.2339
0.142
0.023
0.496
0.336
0.44
0.254
0.186
0.285
0.525
0.174
0.213
0.176
Weight
6.4%
12.4%
26.0%
1.8%
3.6%
2.2%
5.7%
8.9%
4.7%
1.6%
9.7%
7.4%
9.6%
IV, random, 95% CI

1.41 [0.89, 2.23]
1.28 [0.97, 1.69]
1.07 [1.02, 1.12]
2.74 [1.04, 7.24]
1.02 [0.53, 1.97]
2.25 [0.95, 5.33]
1.73 [1.05, 2.85]
1.00 [0.69, 1.44]
1.51 [0.86, 2.64]
0.50 [0.18, 1.40]
1.01 [0.72, 1.42]
1.52 [1.00, 2.31]
0.95 [0.67, 1.34]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.6931
−0.0513
Test for overall effect: � = 2.54 (� = 0.01)
Total (95% CI) 100.0% 1.19 [1.04, 1.36]
39%
Bostwik et al. (1994)
Chang et al. (2006, 2007)
Ford (1999)
Gaard et al. (1997)

Lin et al. (2004)
Nock et al. (2008)
Otani et al. (2005)
Sun et al. (2008)
Suzuki et al. (2006)
Terry et al. (2001, 2002)
(b)
Figure 2: (a) Obesity and colon cancer in men. (b) Obesity and
colon cancer in women.
3.4.7. Multiple Myeloma. Only one study for men
(Figure 8(a)) and two studies for women (Figure 8(b))
assessed the association between obesity and multiple
myeloma.The results reported by Samanic et al., 2006 showed
obese men had significantly lower risk of multiple myeloma
compared tomen of normal weight (RR 0.58; 95% confidence
interval 0.36 to 0.93). The pooled ratio for women showed
an increased risk of multiple myeloma compared to women
of normal weight, although the 95% confidence interval was
just short of reaching statistical significance (RR 1.20; 95%
confidence interval 0.99 to 1.45). There was no significant
heterogeneity across the two studies (heterogeneity� = 0.39).
3.4.8. Non-Hodgkin Lymphoma. Four studies in men
(Figure 9(a)) and six in women (Figure 9(b)) assessed the
Study or subgroup

Bostwik et al. (1994) 0.912
1.10856
0.92028
1.1848
0.9439
0.833
1.396
0.519
0.5068
1.264
1.0043
1.2238
SE
0.187
0.099
0.0296
0.182
0.238
0.229
0.6471
0.03
0.121
0.227
0.137
0.0342
0.08977

0.124
0.059
0.0593
Weight
6.2%
6.5%
6.7%
6.3%
6.0%
6.0%
3.6%
6.7%
6.5%
6.1%
6.4%
6.7%
6.6%
6.5%
6.6%
6.6%
IV, random, 95% CI
2.49 [1.73, 3.59]
3.03 [2.50, 3.68]
2.51 [2.37, 2.66]
3.27 [2.29, 4.67]
2.57 [1.61, 4.10]
2.30 [1.47, 3.60]
4.04 [1.14, 14.36]
0.80 [0.76, 0.85]
1.68 [1.33, 2.13]

1.66 [1.06, 2.59]
3.54 [2.71, 4.63]
2.73 [2.55, 2.92]
0.68 [0.57, 0.81]
3.40 [2.67, 4.34]
0.70 [0.62, 0.78]
0.60 [0.53, 0.67]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.219
−0.3895
−0.361
−0.5108
Log (risk ratio)
Total (95% CI) 100.0% 1.85 [1.30, 2.65]
Heterogeneity: �2 = 0.50; �2 = 16.9827, df = 15 (� <
0.00001); �2 = 99%
Chang et al. (2006, 2007)
Gaard et al. (1997)
Feigelson et al. (2004)

Kawai et al. (2011)
Maso et al. (2011)
Morimoto et al. (2002)
Nilsen and Vatten (2000)
Park et al. (2010)
Reeves et al. (2007, 2011)
Terry et al. (2001, 2002)
Thomas et al. (2009)
Yong et al. (1996)
Salazar-Martinez et al. (2000)
Figure 3: Obesity and endometrial cancer.
−0.6349
0.04879
1.0006
SE
0.113
0.3639
0.582
Weight
44.9%

32.8%
22.4%
IV, random, 95% CI
1.05 [0.84, 1.31]
2.72 [1.33, 5.55]
0.53 [0.17, 1.66]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
Total (95% CI) 100.0% 1.23 [0.58, 2.60]
75%
Sun et al. (2008)
(a)
0.248

1.3083
0.9322
0.89199
SE
0.126
0.617
0.15
0.8
Weight
38.9%
13.9%
37.7%
9.6%
IV, random, 95% CI
1.28 [1.00, 1.64]
3.70 [1.10, 12.40]
2.54 [1.89, 3.41]
2.44 [0.51, 11.70]

IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
Total (95% CI) 100.0% 2.04 [1.18, 3.55]
78%
Reeves et al. (2007, 2011)
Sun et al. (2008)
(b)
Figure 4: (a) Obesity and esophageal adenocarcinoma in men.
(b) Obesity and esophageal adenocarcinoma in women.
0.32208
0.33647
0.50077

SE
0.16
0.335
0.2
Weight
53.5%
12.2%
34.3%
IV, random, 95% CI
1.38 [1.01, 1.89]
1.40 [0.73, 2.70]
1.65 [1.11, 2.44]
IV, random, 95% CI
0.1 0.2 0.5 2 5 10
Total (95% CI) 100.0% 1.47 [1.17, 1.85]
0%

Sun et al. (2008)
(a)
0.63217
1.4929
0.3646
SE
0.083
0.593
0.197
Weight
58.9%
6.8%
34.3%
IV, random, 95% CI
1.88 [1.60, 2.21]
4.45 [1.39, 14.23]

1.44 [0.98, 2.12]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
Total (95% CI) 100.0% 1.82 [1.32, 2.50]
48%
Sun et al. (2008)
(b)
Figure 5: (a) Obesity and gallbladder cancer in men. (b) Obesity
and gallbladder cancer in women.
association between obesity and non-Hodgkin lymphoma.
The pooled risk ratio indicated there was no increased risk
of non-Hodgkin lymphoma among obese men compared to
men of normal weight (RR 1.09; 95% confidence interval 0.98
to 1.21). There was no significant heterogeneity across studies
(heterogeneity � = 0.46). Among obese women, however,
the pooled risk ratio showed a reduced risk of non-Hodgkin
lymphoma compared to women of normal weight (RR 0.91;

3.4.9. Pancreatic Cancer. Nine and ten studies, respectively,
assessed the association between obesity and pancreatic
cancer in men (Figure 10(a)) and women (Figure 10(b)). The
pooled risk ratio illustrated that obese men had an increased
risk of pancreatic cancer compared to men of normal weight
(RR 1.36; 95% confidence interval 1.07 to 1.73). There was
0.01). Among obese women the pooled risk ratio also showed
an increased risk of pancreatic cancer compared to women
of normal weight (RR 1.34; 95% confidence interval 1.22 to
1.46). There was no significant heterogeneity across studies
(heterogeneity � = 0.81).
3.4.10. Postmenopausal Breast Cancer. Eleven studies assess-
ed the association between obesity and postmenopausal
breast cancer (Figure 11). The pooled risk ratio demonstrated
that obese women had an increased risk of postmenopausal
breast cancer compared towomenof normalweight (RR, 1.25;
3.4.11. Rectal Cancer. There were 11 studies for men
(Figure 12(a)) and nine for women (Figure 12(b)) that assess-
ed the association between obesity and rectal cancer. The
pooled risk ratio illustrated that obese men had no increased
risk of rectal cancer compared to men of normal weight
(RR 1.22; 95% confidence interval 0.91 to 1.64). There was
0.00001). Similarly among obese women the pooled risk ratio
illustrated no increased risk of rectal cancer compared to
women of normal weight (RR 1.03; 95% confidence interval
0.74 to 1.44). There was significant heterogeneity across
studies (heterogeneity � = 0.00001).
3.4.12. Renal Cancer. There were 3 studies each for men

(Figure 13(a)) and women (Figure 13(b)) that assessed the
association between obesity and renal cancer. The pooled
risk ratio illustrated that obese men had an increased risk of
renal cancer compared to men of normal weight (RR 1.57;
95% confidence interval 1.38 to 1.77).There was no significant
heterogeneity across studies (� = 0.76). Similarly, among
Study or subgroup
Total (95% CI)
Log (risk ratio)
0.70803
0.11333
SE
0.589
0.142
Weight
5.5%
94.5%
100.0%
IV, random, 95% CI

2.03 [0.64, 6.44]
1.12 [0.85, 1.48]
1.16 [0.88, 1.52]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
0%
Oh et al. (2005)
(a)
Study or subgroup
Total (95% CI)
Log (risk ratio)
0.47
0.22314
SE
0.20687
0.086

Weight
21.0%
79.0%
100.0%
IV, random, 95% CI
1.60 [1.07, 2.40]
1.25 [1.06, 1.48]
1.32 [1.08, 1.60]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
18%
Reeves et al. (2007, 2011)
(b)
Figure 6: (a) Obesity and leukemia in men. (b) Obesity and
leukemia in women.
−0.5276
Study or subgroup
Total (95% CI)

0.3365
0.2397
0.3001
SE
0.54821
0.057
0.407
0.127
Weight
2.1%
65.4%
3.8%
28.7%
100.0%
IV, random, 95% CI
1.40 [0.48, 4.10]
1.27 [1.14, 1.42]
0.59 [0.27, 1.31]
1.35 [1.05, 1.73]
1.26 [1.07, 1.48]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10

Log (risk ratio)
21%
Freedman et al. (2003)
Odenbro et al. (2007)
Rapp et al. (2005)
(a)
Study or subgroup
Total (95% CI)
Log (risk ratio)
0.049
SE
0.225
0.185
0.066
Weight
7.1%
10.5%
82.4%
100.0%

IV, random, 95% CI
0.90 [0.58, 1.40]
1.05 [0.73, 1.51]
0.94 [0.83, 1.07]
0.95 [0.84, 1.07]
Risk ratio Risk Ratio
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.10536
−0.061875
0%
Huang et al. (1997)
Reeves et al. (2007, 2011)
(b)
Figure 7: (a) Obesity and malignant melanoma in men. (b)
Obesity and malignant melanoma in women.

Study or subgroup
Total (95% CI)
Log (risk ratio) SE
0.241
Weight
100.0%
100.0%
0.58 [0.36, 0.93]
0.58 [0.36, 0.93]
IV, fixed, 95% CI
0.1 0.2 0.5 1 2 5 10
IV, fixed, 95% CI
−0.54473
Heterogeneity: not applicable
(a)
Study or subgroup

Reeves et al. (2007, 2011)
Total (95% CI)
Log (risk ratio)
0.4054
0.14842
SE
0.2806
0.104
Weight
12.1%
87.9%
100.0%
IV, random, 95% CI
1.50 [0.87, 2.60]
1.16 [0.95, 1.42]
1.20 [0.99, 1.45]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10

0%
(b)
Figure 8: (a) Obesity and multiple myeloma in men. (b) Obesity
and multiple myeloma in women.
Table 1: Summary of results of meta-analysis of associations
bet-
ween obesity and cancer risk in 57 studies from 18 countries,
1985–
2011.
Cancer Males Females
RR and 95% CI RR and 95% CI
Colon 1.57 (1.48, 1.65)∗ 1.19 (1.04, 1.36)∗
Endometrial NA 1.85 (1.30, 2.65)∗
Esophageal 1.23 (0.58, 2.60) 2.04 (1.18, 3.55)∗
Gallbladder 1.47 (1.17, 1.85)∗ 1.82 (1.32, 2.50)∗
Leukemia 1.16 (0.88, 1.52) 1.32 (1.08, 1.60)∗
Malignant melanoma 1.26 (1.07, 1.48)∗ 0.95 (0.84, 1.07)
Multiple myeloma 0.58 (0.36, 0.93)∗ 1.20 (0.99, 1.45)
Non-Hodgkins lymphoma 1.09 (0.98, 1.21) 0.91 (0.86, 0.97)∗
Pancreatic 1.36 (1.07, 1.73)∗ 1.34 (1.22, 1.46)∗
Postmenopausal breast NA 1.25 (1.07, 1.46)∗
Rectal 1.22 (0.91, 1.64) 1.03 (0.74, 1.44)

Renal 1.57 (1.38, 1.77)∗ 1.72 (1.58, 1.88)∗
Thyroid 1.12 (0.72, 1,72) 1.03 (0.87, 1.23)
∗ Statistically significant at � < 0.05.
obese women the pooled risk ratio illustrated an increased
risk of renal cancer compared to women of normal weight
(RR 1.72; 95% confidence interval 1.58 to 1.88). There was no
0.22).
3.4.13. Thyroid Cancer. There were 6 studies included in the
meta-analysis assessing the association between obesity and
thyroid cancer among men (Figure 14(a)) and 4 studies
among women (Figure 14(b)). The pooled risk ratio illus-
trated that obese men had no increased risk of thyroid cancer
compared to men of normal weight (RR 1.12; 95% confidence
interval 0.72 to 1.72). There was significant heterogeneity
across studies (� < 0.00001). Likewise, among obese women
the pooled risk ratio demonstrated no increased risk of thy-
roid cancer compared to women of normal weight (RR 1.03;
heterogeneity across studies (� < 0.004).
3.4.14. Results Summary. Table 1 summarizes the results of
the observed associations between obesity and the 13 cancers
included in this meta-analysis. Among men a statistically
significant increased risk of cancer was observed for the
following 5 cancers: colon, gallbladder, malignantmelanoma,
pancreatic, and renal cancer. In addition, obese men had
significantly lower risk of multiple myeloma in comparison
to men of normal weight. Among women a statistically signi-
ficant increased risk of cancer was observed for the follow-
ing 8 cancers: colon, endometrial, esophageal, gallbladder,
leukemia, pancreatic, postmenopausal breast, and renal.

Obese women also had significantly lower risk of non-Hodg-
kin lymphoma compared to women of normal weight.
3.4.15. Population-Attributable Risk. As an example to illus-
trate the use of meta-analysis results reported in this study,
the PAR% for men and women in Canada was calculated
on cancers for which a statistically significant association
Study or subgroup
Fernberg et al. (2007)
Rapp et al. (2005)
Log (risk ratio)
0.1484
0.0198
SE
0.066
0.1234
0.4

0.128
Weight
63.3%
18.1%
1.7%
16.8%
IV, fixed, 95% CI
1.16 [1.02, 1.32]
0.95 [0.75, 1.21]
0.91 [0.42, 1.99]
1.02 [0.79, 1.31]
IV, fixed, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.0513
−0.0943
Total (95% CI) 100.0% 1.09 [0.98, 1.21]
Heterogeneity: �2 = 2.60, df = 3 (� = 0.46); �2 = 0%

(a)
Study or subgroup
Fernberg et al. (2007)
Lu (2009)
Reeves et al. (2007, 2011)
Wu et al. (2007)
Log (risk ratio)
0
0.06766
0.4574
0.174
0.17395
SE
0.171
0.0499
0.449
0.125
0.061
0.065
Weight
3.4%
39.8%
0.5%
6.3%

26.6%
23.4%
IV, fixed, 95% CI
1.00 [0.72, 1.40]
1.07 [0.97, 1.18]
1.58 [0.66, 3.81]
1.19 [0.93, 1.52]
1.19 [1.06, 1.34]
0.47 [0.41, 0.53]
IV, fixed, 95% CI
0.1 0.2 0.5 1 2 5 10
Favours experimental Favours control
−0.75311
Total (95% CI) 100.0% 0.91 [0.86, 0.97]
Heterogeneity: �2 = 139.00, df = 5 (� < 0.00001); �2 = 96%
(b)
Figure 9: (a) Obesity and non-Hodgkin lymphoma in men. (b)
Obesity and non-Hodgkin lymphoma in women.
between obesity and cancer incidencewas observed (Table 2).
The population prevalence of obesity in Canada in 2002 was
22.9% in men and 23.2% in women. Among men PAR%
ranged from a low of 5.6% to a high of 11.5%. The PAR% for
obesity among men was greatest for colon and renal cancer
(11.5%), gallbladder cancer (9.7%), pancreatic (7.6%), and

malignant melanoma (5.6%). When the 95% confidence
intervals for PAR% for men are considered the results are
particularly noteworthy. For example for colon cancer PAR%
is as low as 9.9% but at the upper limit of the confidence inter-
val PAR% is 13.1%. Likewise, for renal cancer the 95% con-
fidence interval for PAR% ranged from 8.0% to 15%, while
for pancreatic cancer, it ranged from 1.6% to 17.5%, and 1.6%
to 9.9% for malignant melanoma.
Among women PAR% ranged from a low of 4.2% to a
high of 19.4%. The PAR% for obesity among women was
greatest for esophageal adenocarcinoma (19.4%), endometrial
(16.5%), gallbladder (16%), and renal cancer (14.3). When the
95%confidence intervals for PAR% forwomen are considered
the results are also noteworthy. For example, at the upper
limit of the confidence interval, PAR%was as high as 37.2% for
esophageal adenocarcinoma, 27.7% for endometrial cancer,
and 25.8% for gallbladder cancer. However, at the lower
limit the PAR% was as low as 0.9% for colon cancer, 1.6%
for postmenopausal breast cancer, and 1.8% for malignant
melanoma.
4. Discussion
The results of this meta-analysis demonstrate a significant
association between obesity and some of the 13 cancers
for men and many of the 13 cancers for women. Obesity
was significantly and positively associated with 5 of the
possible 11 cancers relevant to men including, colon, gall-
bladder, malignant melanoma, pancreatic, and renal. A sig-
nificant association was not observed among obese men for
esophageal adenocarcinoma, leukemia, multiple myeloma,
non-Hodgkin lymphoma, rectal, and thyroid cancer. Among
women obesity was significantly and positively associated
with 8 of the 13 cancers including, colon, endometrial,

esophageal adenocarcinoma, gallbladder, leukemia, pancre-
atic, postmenopausal breast, and renal cancer. A significant
association was not observed among women for malignant
melanoma,multiplemyeloma, non-Hodgkin lymphoma, rec-
tal cancer, and thyroid cancer.
Study or subgroup
Batty et al. (2009)
Chang et al. (2006, 2007)
Oh et al. (2005)
Otani et al. (2005)
Rapp et al. (2005)
Sun et al. (2008)
Total (95% CI)
Log (risk ratio)
0.1655
0.039
0.8671
0.73237
0.03922
0.85015
0.14842

0.29267
0.73237
SE
0.201
0.159
0.236
0.365
0.507
0.275
0.351
0.141
0.293
0.365
Weight
12.8%
14.8%
11.4%
7.2%
4.5%
9.9%
7.5%
15.6%
9.2%
7.2%
100.0%
IV, random, 95% CI
1.18 [0.80, 1.75]
1.04 [0.76, 1.42]

2.38 [1.50, 3.78]
2.08 [1.02, 4.25]
1.04 [0.39, 2.81]
0.70 [0.41, 1.20]
2.34 [1.18, 4.66]
1.16 [0.88, 1.53]
1.34 [0.75, 2.38]
2.08 [1.02, 4.25]
1.36 [1.07, 1.73]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.35667
57%
(a)
Study or subgroup
Chang et al. (2006, 2007)
Otani et al. (2005)
Reeves et al. (2007, 2011)
Stevens et al. (2009)
Sun et al. (2008)

Total (95% CI)
Log (risk ratio)
0.13103
0.285
0.54812
0.39204
0.0953
0.18232
0.3148
0.293
0.5878
0.39204
SE
0.179
0.201
0.267
0.456
0.15825
0.234
0.079
0.081
0.236
0.456
Weight
6.7%

5.3%
3.0%
1.0%
8.5%
3.9%
34.2%
32.5%
3.8%
1.0%
100.0%
IV, random, 95% CI
1.14 [0.80, 1.62]
1.33 [0.90, 1.97]
1.73 [1.03, 2.92]
1.48 [0.61, 3.62]
1.10 [0.81, 1.50]
1.20 [0.76, 1.90]
1.37 [1.17, 1.60]
1.34 [1.14, 1.57]
1.80 [1.13, 2.86]
1.48 [0.61, 3.62]
1.34 [1.22, 1.46]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
0%

(b)
Figure 10: (a) Obesity and pancreatic cancer in men. (b) Obesity
and pancreatic cancer in women.
The magnitude of the associations between obesity and
cancer incidence in men and women, including the 95%
confidence intervals, is particularly noteworthy. For example,
obese men have an increased risk of developing colon and
renal (RR both 1.57), gallbladder (1.47), pancreatic cancer
(1.36), and malignant melanoma (1.26) in comparison to
nonobese men. When the 95% confidence intervals are
considered the results become even more compelling. For
example the risk ratio is as high as 1.77 for renal cancer, 1.73
for pancreatic cancer and 1.65 for colon cancer. Even at the
low end of the 95% confidence interval, cancer risk is still
noteworthy at 1.48 for colon cancer, and 1.38 for renal cancer.
The results for obese women are equally concerning.
Obese women are not only at higher risk for developing
cancer at more sites than men, the risk associated with some
of the cancers is significantly higher. For example, obese
women have an increased risk for developing esophageal
adenocarcinoma (RR 2.04), endometrial (1.85), gallbladder
(1.82), and renal cancer (1.72). Where statistically significant
associations are observed only one risk ratio is less than 20%
(colon 1.19), while the remaining risk ratios range between
1.25 (postmenopausal breast), 1.32 (leukemia), and 1.34 (pan-
creatic). The 95% confidence intervals are also alarming. For
example, at the upper limit of the confidence interval the risk
ratio is as high as 3.55 for esophageal adenocarcinoma, 2.65
for endometrial, and 2.50 for gallbladder cancer.
The results also clearly demonstrate that a significant

proportion of cancer incidence could be avoided, in Canada,
by reducing the percentage of obese adults. The population-
attributable risk percent for obesity in Canada was highest
Study or subgroup
Chang et al. (2006, 2007)
Huang et al. (1997)
Lin et al. (2004)
Lundqvist et al. (2007)
Montazeri et al. (2008)
Otani et al. (2005)
Reeves et al. (2007, 2011)
Tehard and Clavel-Chapelon (2006)
Terry et al. (2001, 2002)
Total (95% CI)
Log (risk ratio)
0.095
0.1222
0.98208
0.68813
0.26236
0.33647
0.824175

0.25464
0.36464
0.23111
SE
0.069
0.131
0.4859
0.244
0.137
0.158
0.08142
0.452
0.027
0.165
0.143
Weight
13.3%
10.5%
2.2%
6.0%
10.2%
9.2%
12.8%
2.5%
14.6%
8.9%
9.9%

100.0%
IV, random, 95% CI
1.10 [0.96, 1.26]
1.13 [0.87, 1.46]
2.67 [1.03, 6.92]
1.99 [1.23, 3.21]
1.30 [0.99, 1.70]
1.40 [1.03, 1.91]
0.76 [0.65, 0.89]
2.28 [0.94, 5.53]
1.29 [1.22, 1.36]
1.44 [1.04, 1.99]
1.26 [0.95, 1.67]
1.25 [1.07, 1.46]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
Favours experimental Favours control
−0.27119
Heterogeneity: �2 = 0.04; �2 = 50.72, df = 10 (� < 0.00001);
�2 = 80%
Figure 11: Obesity and postmenopausal breast cancer.
Table 2: Population-attributable risk percent in Canada for
obesity and cancer.

Cancer
Males Males Females Females
RR and 95% CI Population-attributable risk
Percent
RR and 95% CI Population-attributable risk
Percent
Colon 1.57 (1.48, 1.65)∗ 11.5% (9.9, 13.1)∗ 1.19 (1.04, 1.36)∗
4.2% (0.9, 7.7)
Endometrial NA NA 1.85 (1.3, 2.65)∗ 16.5% (6.5, 27.7)∗
Esophageal 1.23 (0.58, 2.60) — 2.04 (1.18, 3.55)∗ 19.4% (4.0,
37.2)∗
Gallbladder 1.47 (1.17, 1.85)∗ 9.7% (3.7, 16.3)∗ 1.82 (1.32,
2.50)∗ 16% (6.9, 25.8)∗
Leukemia 1.16 (0.88, 1.52) — 1.32 (1.08, 1.60)∗ 6.9% (1.8,
12.2)∗
Malignant melanoma 1.26 (1.07, 1.48)∗ 5.6% (1.6, 9.9)∗ 0.95
(0.84, 1.07) —
Multiple myeloma 0.58 (0.36, 0.93)∗ — 1.20 (0.99, 1.45) —
Non-Hodgkin lymphoma 1.09 (0.98, 1.21) — 0.91 (0.86, 0.97)∗
Pancreatic 1.36 (1.07, 1.73)∗ 7.6% (1.6, 17.5)∗ 1.34 (1.22,
1.46)∗ 7.3% (4.9, 9.6)∗
Postmenopausal breast NA NA 1.25 (1.07, 1.46)∗ 5.5% (1.6,
9.6)∗
Rectal 1.22 (0.91, 1.64) — 1.03 (0.74, 1.44) —
Renal 1.57 (1.38, 1.77)∗ 11.5% (8.0, 15.0)∗ 1.72 (1.58, 1.88)∗

14.3% (12.0, 16.9)∗
Thyroid 1.12 (0.72, 172) — 1.03 (0.87, 1.23) —
∗ Statistically significant at � < 0.05.
in men for five cancers: colon, renal, gallbladder, pancreatic,
and malignant melanoma. Significant reductions in obesity
therefore would dramatically decrease the incidence of those
cancers. For women the greatest benefits from obesity reduc-
tion and prevention efforts would be observed for esophageal
adenocarcinoma, endometrial, gallbladder, renal, pancreatic,
leukemia, postmenopausal breast, and colon cancer. Signif-
icant reductions in these cancers for men and women in
Canada would result in decreased health care expenditures
and improved quality of life for many men and women in
Canada.
When the results of this meta-analysis are compared to
those of Renehan et al. [6], who reviewed similar but earlier
literature on obesity and cancer, a number of important simi-
larities and differences are observed. First, there were some
differences in the associations found between the two meta-
analyses for men. Renehan et al. reported a significant and
positive association for the following five cancers that was not
Study or subgroup
Chang et al. (2006, 2007)

Oh et al. (2005)
Otani et al. (2005)
Rapp et al. (2005)
Total (95% CI)
Log (risk ratio)
0.239
0.8796
0.36464
0.262
0.07696
0.47
0.50682
0.30748
0.41211
SE
0.062
0.165
0.042
0.6025
0.303
0.3128
0.338
0.454

0.253
0.102
0.213
Weight
12.2%
10.8%
12.3%
4.1%
8.2%
8.1%
7.6%
5.8%
9.2%
11.8%
9.9%
100.0%
IV, random, 95% CI
0.59 [0.52, 0.66]
0.94 [0.68, 1.30]
1.27 [1.17, 1.38]
2.41 [0.74, 7.85]
1.30 [0.70, 2.40]
1.08 [0.56, 2.09]
1.60 [0.66, 3.90]
1.66 [1.01, 2.73]
1.36 [1.11, 1.66]
1.51 [0.99, 2.29]
1.22 [0.91, 1.64]

IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.53604
−0.061875
1.44 [0.80,2.61]
Stolzenberg-Solomon et al. (2002, 2008)
�2 = 92%
(a)
Study or subgroup
Chang et al. (2006, 2007)
Lin et al. (2004)
Otani et al. (2005)
Sun et al. (2008)
Terry et al. (2001, 2002)
Total (95% CI)
Log (risk ratio)

0.0862
0.0392
0.1906
0.43825
0.09531
0.26236
0.13103
0.3001
SE
0.07712
0.242
0.033
0.73798
0.455
0.2788
0.443
0.198
0.219
Weight
15.7%
12.1%
16.1%
3.9%
7.4%
11.2%
7.6%
13.2%

12.7%
100.0%
IV, random, 95% CI
0.50 [0.43, 0.58]
1.09 [0.68, 1.75]
1.04 [0.97, 1.11]
1.21 [0.28, 5.14]
1.55 [0.64, 3.78]
1.10 [0.64, 1.90]
1.30 [0.55, 3.10]
1.14 [0.77, 1.68]
1.35 [0.88, 2.07]
1.03 [0.74, 1.44]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
−0.69315
�2 = 90%
(b)
Figure 12: (a) Obesity and rectal cancer in men. (b) Obesity and
rectal cancer in women.
reported in this meta-analysis: leukemia, multiple myeloma,
non-Hodgkin lymphoma, rectal, and thyroid cancer. Further-

more, Renehan et al. did not report a significant and positive
association for gallbladder and pancreatic cancer whereas a
significant andpositive associationwas reported in this paper.
There was greater similarity between the results of this
paper and Renehan et al. for women. For example, there were
only two cancers, multiple myeloma and thyroid cancer
where Renehan et al. reported a positive and significant asso-
ciation with obesity for women, and this paper did not.There
were no differences for the remaining eleven cancers on the
relationship between obesity and cancer in women.
The second important difference between the two meta-
analyses concerns the observed magnitude of the association
and the coinciding 95% confidence intervals. Generally the
results of our meta-analysis demonstrated higher magnitude
of associations and confidence intervals. The most notable
differences occurred for four cancers among men: renal,
colon, pancreatic, and gallbladder cancer. For example, we
reported a pooled risk ratio and 95% confidence interval of
1.57 (1.38, 1.77) for renal cancer, while Renehan et al. re-
ported 1.24 (1.15, 1.34). Likewise, for colon cancer, we
reported
1.57 (1.48, 1.65), while Renehan et al. reported 1.24 (1.20,
1.28),
and for pancreatic cancer, we reported 1.36 (1.07, 1.73) and
Renehan et al. reported 1.07 (0.93, 1.23). Finally, for gallblad-
der cancer, we reported 1.47 (1.17, 1.85) while Renehan et al.
reported 1.09 (0.99, 1.21).
Similar differences existed for obese women with the
greatest differences in the pooled risk ratios and 95%
confidence intervals being for three cancers: endometrial,
renal, and gallbladder. For these cancers the magnitude of
associations we reported is larger than those reported by

Renehan et al. For endometrial cancer we reported a risk
ratio of 1.85 (1.30, 2.65), while Renehan et al. reported 1.59
Study or subgroup
Kanda et al. (2010)
Oh et al. (2005)
Total (95% CI)
Log (risk ratio)
0.43825
0.688
0.4967
SE
0.065
0.331
0.415
Weight
94.1%
3.6%
2.3%
100.0%

IV, random, 95% CI
1.55 [1.36, 1.76]
1.99 [1.04, 3.81]
1.64 [0.73, 3.71]
1.57 [1.38, 1.77]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
0%
(a)
Study or subgroup
Kanda et al. (2010)
Reeves et al. (2007, 2011)
Total (95% CI)
Log (risk ratio)
0.6152
0.438
0.47

0.4187
SE
0.055
0.367
0.2069
0.078
Weight
62.9%
1.4%
4.4%
31.3%
100.0%
IV, fixed, 95% CI
1.85 [1.66, 2.06]
1.55 [0.75, 3.18]
1.60 [1.07, 2.40]
1.52 [1.30, 1.77]
1.72 [1.58, 1.88]
IV, fixed, 95% CI
0.1 0.2 0.5 1 2 5 10
Heterogeneity: �2 = 4.47, df = 3 (� = 0.22); �2 = 33%
(b)

Figure 13: (a) Obesity and renal cancer in men. (b) Obesity and
renal cancer in women.
(1.5, 1.68). For renal cancer we reported 1.72 (1.58, 1.88) while
Renehan reported 1.34 (1.25, 1.43), and for gallbladder cancer
we reported 1.82 (1.32, 2.50) and Renehan, 1.59 (1.02, 2.47).
There was only one cancer where Renehan et al. reported a
larger risk ratio than we did, which was for non-Hodgkin
lymphoma.
There are some plausible explanations for these differ-
ences. Likely the most significant factor contributing to the
differences in the results of the two meta-analyses are the
different approaches to the statistical methods used to aggre-
gate the data across studies. The approach used by Renehan
assessed the risk in developing cancer as BMI increased every
5 points, from the lowest BMI category up to the highest
category. This strategy incorporates the risk of those who
have a BMI slightly above normal, to above normal to those
identified as obese. Given the risk of developing cancer at
lower BMI levels is generally less than that of obese persons
which provides some explanation why Renehan et al’s results
generally illustrate lower overall risk and more narrow confi-
dence intervals.
In this paper we compared the risk of developing cancer
between those in the lowest BMI category (normal healthy
weight) and those considered obese (BMI 30+) category. Our
meta-analysis therefore did not incorporate into the pooled
risk ratio and 95% confidence intervals, the risk associated
with developing cancer among those with slightly above
normal and above normal BMI. Given the primary objective
of this paper was to assess the relationship between obesity
and cancer incidencewe believe this was themost appropriate

statistical analysis to answer the research question. The
question Renehan et al. answered was broader in that they
were interested in assessing across the spectrum of BMI from
normal to obese, the relationship between BMI and cancer
risk.
A second major difference between the two reviews was
our exclusion of studies judged to be of weak methodological
quality. The results of studies of weak methodological quality
are less trustworthy. The exclusion of these studies likely
contributed to the higher pooled risk ratios observed in this
meta-analysis in comparison to Renehan et al.The combined
effect of the different statistical approaches and the exclusion
of studies of weak methodological quality likely explains
much of the observed differences between the reviews.
Finally, itmay be that studies published since theRenehan
et al. meta-analysis, report higher relative risks than had
been reported previously. Given several studies published
since 2007 were added to this paper, this may further explain
the higher pooled risk ratios and 95% confidence intervals
observed in this paper.
It is important to note that the results from some of the
studies included in this paper may not be relevant to a Cana-
dian context. For example, studies conducted in South East
Asian countries may have limited applicability in Canada.
However, a rigorous and systematic meta-analysis process
Study or subgroup

Brindel et al. (2009)
Chang et al. (2006, 2007)
Oh et al. (2005)
Total (95% CI)
Log (risk ratio)
0.637
0.131
0.647
0.11333
SE
0.143
0.229
0.16
0.444
0.2449
0.353
Weight
19.9%
17.7%
19.5%
11.6%
17.2%
14.0%

100.0%
IV, random, 95% CI
0.54 [0.41, 0.72]
1.89 [1.21, 2.96]
1.14 [0.83, 1.56]
1.91 [0.80, 4.56]
1.12 [0.69, 1.81]
0.98 [0.49, 1.96]
1.12 [0.72, 1.72]
IV, random, 95% CI
0.1 0.2 0.5 1 2 5 10
Favours experimental Favours controlTest for overall effect: �
= 0.49 (� = 0.62)
−0.0202
−0.607
Heterogeneity: �2 = 0.23; �2 = 28.37, df = 5 (� < 0.0001); �2
= 82%
(a)
Study or subgroup
Brindel et al. (2009)
Chang et al. (2006, 2007)
Tehard and Clavel-Chapelon (2006)

Total (95% CI)
Log (risk ratio)
0.095
0.344
0.565
SE
0.131
0.194
0.217
0.23
Weight
46.6%
21.3%
17.0%
15.1%
100.0%
IV, fixed, 95% CI
0.76 [0.58, 0.98]
1.10 [0.75, 1.61]
1.41 [0.92, 2.16]
1.76 [1.12, 2.76]
1.03 [0.87, 1.23]
IV, fixed, 95% CI

Running Head CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFSCRIT.docx

Running Head CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFSCRIT.docx

Recommended

Recommended

More Related Content

Similar to Running Head CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFSCRIT.docx

Similar to Running Head CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFSCRIT.docx (20)

More from healdkathaleen

More from healdkathaleen (20)

Recently uploaded

Recently uploaded (20)

Running Head CRITIQUE OF OCEAN TEMPERATURES IN CORAL REEFSCRIT.docx