h"p://www.pri.

org/stories/2012-12-03/video-
land-tobacco

Session 2: Smoking and lung cancer

Dr Clarence Tam(clarence_tam@nuhs.edu.sg)
Saw Swee Hock School of Public Health
National University of Singapore
SPH2101: Public Health and Epidemiology

Session objec+ves
Overview of lung cancer epidemiology
Epidemiological evidence for the associa+on
between smoking and lung cancer
Tobacco control strategies
Introduce epidemiological concepts:
Incidence and prevalence
Measures of associa@on
Types of epidemiological studies

BURDEN AND EPIDEMIOLOGY

LUNG CANCER

Global burden of lung cancer

Most common cause of cancer-related death
worldwide (~1.6 million in 2012)
Accounts for ~13% of all cancers
~70% of lung cancer deaths caused by tobacco
use
Lung cancer deaths expected to rise to ~2.4
million worldwide by 2030

IHME:
h"p://vizhub.healthdata.org/gbd-compare/

globalcancermap.com

In Singapore, lung cancer is the second most

common cancer in men, and the most common
cause of cancer-related mortality

(it is the third most common cancer in

women and the second most common
cause of cancer-related mortality)

Lim. Singapore Med J 2012;53(1):3

Lung Cancer (C33-C34): 2010-2011

Net Survival up to Ten Years after Diagnosis, Adults (Aged 15-99), England and Wales

If youre diagnosed with

lung cancer, your chances
of surviving 5 years are
~10%

Please include the citation provided in our Frequently Asked Questions when reproducing this chart: http://info.cancerresearchuk.org/cancerstats/faqs/#How
Prepared by Cancer Research UK
Original data sources:
Survival estimates were provided on request by the Cancer Research UK Cancer Survival Group at the London School of Hygiene and Tropical Medicine.
http://www.lshtm.ac.uk/eph/ncde/cancersurvival/

Lung Cancer (C33-C34): 1971-2011

Age-Standardised Five-Year Net Survival, England and Wales

Despite medical
advances, 5-year survival
rates have not improved
drama@cally since the
1970s

Please include the citation provided in our Frequently Asked Questions when reproducing this chart: http://info.cancerresearchuk.org/cancerstats/faqs/#How
Prepared by Cancer Research UK
Original data sources:
Survival estimates were provided on request by the Cancer Research UK Cancer Survival Group at the London School of Hygiene and Tropical Medicine.
http://www.lshtm.ac.uk/eph/ncde/cancersurvival/http://www.lshtm.ac.uk/eph/ncde/cancersurvival/

Lung cancer prevalence

In 2012, 1.89 million people worldwide were
living with lung cancer
This equates to ~0.03% of the worlds
popula+on, or ~3 in 10,000 people

We call this the prevalence of lung cancer

Prevalence
What propor@on of the popula@on
has a given disease at a given point
in @me?

Includes all those with exis@ng disease at

@me of study (whether onset was recent or
a long @me ago)

Prevalence calcula+on These people

(the numerator)

Number of people with disease at a given +me

x 100%
Total number of people in the popula+on

Expressing the number of cases

rela@ve to the popula@on size
allows us to compare popula@ons of
dierent sizes

must be included in

These people
(the denominator)

Prevalence examples
On 1 January 2010, 0.16% of women in
the USA had a history of cervical cancer

The prevalence of Hepa@@s B virus

surface an@gen among Singaporeans
aged 18 to 69 years in 2005 was 2.8%

Have you smoked at least 1 cigare_e

in the previous 12 months?
Dividing the number of Yes responses by the number of
respondents would give us an es@mate of the prevalence
of smoking over 1 year among SPH2101 students

But how good is this es@mate? Well come back to this in Tutorial 1

New or exis+ng disease?

Epidemiologists dieren@ate between new
(incident) and exis@ng (prevalent) disease

Why?

Incidence and prevalence

Many chronic diseases
do not develop un@l
years or decades ager
exposure, e.g. cancer

Many infec@ous
diseases have long
incuba@on periods
between infec@on and
clinical symptoms, e.g.
HIV/AIDS, tuberculosis

Loeb et al. Cancer Res 1984;44:5950-58

Incidence and prevalence

Prevalent (exis@ng) disease is aected by disease
trends both now and in the past
Lung cancer cases among middle-aged men today tell you
about smoking pa"erns a few decades ago
AIDS cases today tell you about HIV transmission pa"erns a
decade ago

Incident (new) disease reects recent trends

If you want to know whether recent policies to

reduce lung cancer are working, you need to look at
new cancer cases and deaths
If you want to know if HIV transmission is going
down, you need to study new HIV infec@ons, not AIDS

Which is the most successful mobile

phone company?

Samsung vs Apple

Prevalence is inuenced by incidence and disease dura@on

Rahul Patwari: h"ps://www.youtube.com/watch?v=1jzZe3ORdd8

Prevalence over this :me period

includes 2 new cases + 3 exis:ng
cases

Disease onset

TIME
Prevalence depends on incidence and disease dura@on

Prevalence over this :me period

includes 2 new cases only

Death

X
X

X
X

TIME
If disease survival is short, prevalence and incidence will be similar

Measuring incidence
Country

Singapore

New lung cancer

cases (2012)

1,590

Vietnam

19,559

India

63,759

China

597,182

USA

167,545

What does this tell you about lung cancer

incidence?
Where are you most likely to get lung
cancer?

Absolute numbers tell us only where most cases occur

To learn about the risk of contrac@ng lung cancer, we need to know the size
of the popula@on in which those deaths occur

Calcula+ng incidence rates

Country

Singapore

New lung
cancer cases
(2012)

1,974

Popula>on
(mid-2012)

5,250,000

New cases per

100,000
persons
(2012)

The lung cancer rate is the

number of newly diagnosed
cases in the popula@on
during 2012, divided by the
size of the popula@on
37.6
(usually the mid-year
24.4
es@mate)

Vietnam

21,865 89,610,656

India

70,275 1,254,910,714

5.6

China

652,842 1,360,087,500

48.0

USA

214,226 315,967,552

67.8

We ogen express this as cases per 100,000 persons:

1,974 / 5,250,000 * 100,000 = 37.6 cases per 100,000 persons in Singapore in 2012

Comparing incidence rates

We can take the ra@o of rates between countries, e.g.:
Country

Singapore

New lung
cancer cases
(2012)

Popula>on
(mid-2012)

New cases per

100,000 persons
(2012)

1,974

5,250,000

37.6

Vietnam

21,865

89,610,656

24.4

India

70,275 1,254,910,714

5.6

China

652,842 1,360,087,500

48

USA

214,226

315,967,552

37.6
= 1.54
24.4
This is the
rate ra:o

67.8

Lung cancer rate in Singapore appears to be 1.54 @mes (or 54%) higher than in Vietnam
But its not that simple! WHY?
Kimman. Asian Pacic J Cancer Prev 2012;13:411-20

Singapore has a higher propor@on of the popula@on in older age groups, who have
a higher risk of lung cancer

If we use some sta@s@cal techniques to account for dierences in the popula@on

age structure of dierent countries, the picture looks quite dierent!
We call these age-standardised rates
Country

Singapore

New lung
cancer cases
(2012)

Popula>on
(mid-2012)

New cases per

100,000 persons
(2012)

Age-standardised
cases per 100,000

1,974

5,250,000

37.6

24.9

Vietnam

21,865

89,610,656

24.4

25.2

India

70,275 1,254,910,714

5.6

6.9

China

652,842 1,360,087,500

48

36.1

USA

214,226 315,967,552

67.8

38.4

The rate ra@o is now:

24.9
= 0.98
25.2

i.e. lung cancer incidence in Singapore is about the

same as in Vietnam

So why does this happen?

Life expectancy is lower in Vietnam a
smaller frac@on of the popula@on
survive to older age when lung cancer
risk is higher
If Vietnam had the same popula@on
age structure as Singapore, lung
cancer rates would similar in the two
countries
In epidemiology, we say that age confounds the associa@on between country and lung
cancer incidence

At rst look, it seems as if Singapore has higher lung cancer incidence than Vietnam, but
this is really explained by dierences in age structure
You will learn more about confounding later in the module

A quick recap

Weve done quite a lot of

epidemiology so far
We learn how important a disease is in a popula@on
by measuring how common it is, e.g. cases, deaths:
Prevalence
Incidence

These are measures of disease frequency

Absolute numbers dont tell us much about

disease risk need to know popula@on size
(denominator)

 We can compare disease frequency between groups

or popula@ons, e.g.:
Risk ra@o
Mortality ra@o

These are measures of associa:on

We must be careful that these associa@ons are real and not

due to other dierences between popula@ons, e.g.:
Age structure
Socioeconomic status
Ethnicity

These are ogen confounders in our

analysis

WHAT CAUSES IT AND HOW DO WE KNOW?

LUNG CANCER

How do we know that smoking causes

lung cancer?

Epidemiological evidence
Secular trends in cigare_e sales and lung
cancer deaths
Ecological data
Early case-control studies
Cohort studies
Second-hand smoking

Historical data
Increases in sales of
cigare"es precede
increases in lung
cancer by about 20
years, in both males
and females

Strength of evidence:
WEAK
Many other things
could precede rises in
lung cancer death

Loeb et al. Cancer Res 1984;44:5950-58

Ecological data
We can look at
correla@ons between
the % of the
popula@on that
smokes and lung
cancer mortality for
dierent countries
This is called an
ecological study
Strength of evidence:
WEAK TO MODERATE
Many other things
could correlate with
lung cancer
mortality
www.gapminder.org

Ecological studies are a quick way to explore associa@ons between diseases and risk factors
Need only aggregate sta@s@cs on disease frequency and the risk factor of interest

BUT
This does not mean that
smoking more will make
you live longer!

Associa@on Causa@on!

Ecological data cant tell

you if its the people who
smoke who get lung
cancer
(although this seems
reasonable!)
More on ecological studies
later in the module
www.gapminder.org

What else could we do?

Compare lung cancer pa@ents with healthy individuals do lung cancer
pa@ents smoke more?
This is a case-control study

Compare smokers and non-smokers do smokers have a higher risk of lung

cancer?
This is a cohort study

Sir Richard Doll: Smoking and lung cancer

Richard Doll conducted seminal studies to

establish the link between smoking and lung
cancer

At the @me, other environmental factors were
thought to cause lung cancer, e.g. tarmac, car
fumes

His ndings convinced him to quit smoking!

We call this the outcome

In a case-control study, we
recruit people with lung cancer
(cases) and a comparable group
of people without lung cancer
(controls)

We then look back to see if

lung cancer pa@ents are
more likely to have smoked
in the past

This is the exposure or risk factor

Cases and controls should be as similar
as possible except for disease status
this is not easy!

cases

controls

A higher % of lung cancer pa@ents were smokers

But note that most people were smokers at
the @me, even among non-cancer pa@ents!

But lung cancer pa@ents also smoked more heavily

This is an example of dose response

Strength of evidence: MODERATE TO STRONG

Lung cancer pa@ents more likely to be smokers
Among smokers, lung cancer pa@ents also smoke more
Not due to other dierences between lung cancer cases and controls wrt:
age, sex, social class or place of residence

Look back at smoking

history

Smoker

LUNG CANCER
CASES

Non-smoker
Compare smoking history
between cases and controls
Are cases more likely smoke
than controls?

Iden+fy individuals
with disease (cases)
and without disease
(controls)

Non-smoker

CONTROLS
Smoker
Look back at smoking
history

TIME

POPULATION

Some issues with case-control studies

How reliably do people report life@me exposure to smoking?
Perhaps lung cancer pa@ents have thought more
about what could have caused their illness, so they
remember smoking history more accurately

This is an example of recall bias
Because we collect informa@on on lung cancer and smoking history at
the same @me, can we be sure that its smoking that causes lung cancer?
Perhaps lung cancer pa@ents are more
stressed, so they take up smoking

This is an example of reverse causality
More on these issues later in the module

Sir Richard Doll (1912 2005)

established the rst cohort study
to show deni@vely the health
eects of smoking
Credit: CJ Dub

The Bri@sh Doctors Study involved 34,439 doctors

born before 1930
They provided informa@on about smoking habits in
1951, 1957, 1966, 1971, 1978, 1991 and 2001
Compared annual risk of death between smokers and
non-smokers

Smokers aged 45+ years had

more than twice the risk of
death each year compared with
lifelong non-smokers
Smokers were also 15 @mes more likely to die from lung cancer
compared with lifelong non-smokers
Strength of evidence: STRONG
Very large study
Data on smoking and death collected over several decades
Can be sure that lung cancer occurred ager people started smoking avoids reverse causality

More on this in Tutorial 1

follow-up at pre-
determined +me points
to see who died from
lung cancer

POPULATION

Lung cancer
death

SMOKERS

Survivor
Iden+fy smokers
and non-smokers
in the popula+on

compare mortality between

groups
Is mortality higher among
smokers?
Lung cancer
death

NON-SMOKERS

follow-up at pre-
determined +me points
to see who died from
lung cancer

TIME

Survivor

Another quick recap

We can inves@gate associa@ons between exposures (risk factors)

and outcomes (disease, death) using dierent epidemiological
study designs
Historical trends in exposure and disease
Ecological studies: plot aggregate sta@s@cs of exposure and outcome
for dierent popula@ons, e.g. ci@es, countries, and look for correla@ons
Case-control studies: compare exposure history in cases
of disease and disease-free controls
Cohort studies: follow up exposed and unexposed groups over
@me and compare risk of acquiring disease (incidence)

Researchers recruited children with leukaemia and

children without leukaemia into a study. They found that
leukaemia children were more likely than non-leukaemia
children to have been exposed to X-rays in utero. What
type of study is this?

This is a case-control study we compare cases of leukaemia

with non-leukaemia controls to see if cases are more likely to
have been exposed to X-rays in utero

A study inves@gated whether air pollu@on is linked

to heart disease deaths. Researchers compared data
on average air pollu@on levels and heart disease
mortality from 25 ci@es. What type of study is this?

This is an ecological data aggregated (not individual)

data on air pollu@on and heart disease mortality are
compared between ci@es

TOBACCO PROJECTIONS

CONTROL MEASURES

Tobacco-related mortality is expected

to rise by 40-50% in low/middle-
income countries by 2030

Mathers. PLoS Medicine 2006; 3(11): e442

Deaths in low/middle-income countries

Donor funding in low/middle-income countries

WHAT WORKS?

CONTROL MEASURES

In 2005, the WHO

Framework Conven@on on
Tobacco Control (FCTC) was
ra@ed by the UN
In 2008, WHO introduced
MPOWER, a package of 6
evidence-based measures
to reduce tobacco use

Taxa+on
10% increase in tobacco price
reduced consump@on by 4% in
high-income countries

Smoking reduc@ons more
pronounced in young people

Increased tax revenue

from tobacco can be
used to directly fund
health programmes

240 Filipinos die each

day from tobacco-
related diseases
Increased tax on
tobacco raised US$1.2
billion in the rst year,
used to provide
healthcare to an
addi@onal 14 million
families

Bans on Tobacco Adver+sing,

Promo+on and Sponsorship (TAPS)
TAPS includes direct and indirect adver@sing:
Commercials
Events sponsorship
Product placement
Corporate social responsibility ac@vi@es
Branding
Price discounts
In-store displays

Since 1998, the Master Se"lement

Agreement in the US restricts product
placement deals with the movie industry

Brand appearances and

tobacco screen @me have
decreased yearly in the 100
top-grossing Hollywood
lms

Bergamini et al. JAMA Pediatrics 2013;167(7):634-9

Packaging
Restrict use of misleading
descriptors, e.g. LIGHT,
SMOOTH, GOLD
No evidence that
these are less
harmful

Packaging
Use of health warnings
Graphic pictures more
eec@ve than plain text

Packaging
Plain packaging
Reduces appeal,
par@cularly to children

Helps to make health

warnings more visible

Smoking restric+ons
Restrict smoking
in the work place
and public spaces

Reduce exposure
to second hand
smoke

Second hand
smoke kills
600,000 people a
year

h"p://www.who.int/tobacco/mpower/publica@ons/en_{i_mpower_brochure_p.pdf?ua=1

Smoking cessa+on
Integrate cessa@on advice into rou@ne
healthcare
Access to quitlines and cessa@on
treatments
Train healthcare workers in cessa@on
support

Advice from a healthcare worker
increases quit rates

Summary
Lung cancer is the most common cancer
worldwide
Tobacco accounts for ~70% of cases worldwide
Major progress in high-income countries, but
challenges ahead in low/middle-income countries
Eec+ve control measures outlined in WHO
Framework Conven+on on Tobacco Control
(FCTC) through MPOWER measures

 We measure how common a disease is using:

Prevalence: percentage of popula+on with disease

Incidence: measures occurrence of new disease

We can compare disease frequency between

popula+ons, e.g. using risk ra+os

But need to account for important dierences that

aect risk of disease, e.g. age, sex

We can study associa+ons between diseases and

exposures using dierent study designs:
Ecological
Cross-sec+onal
Case-control
Cohort