2. Epidemiology of Cancer Flashcards
what was the statement that received a lot of backlash from epidemiologists? why?
some scientists stated that most cancers are due to bad luck, i.e. from random mutations arising in normal cells
but epidemiologists know that most cancers are preventable
are rates of cancers constant over time?
no –> they change over time, can’t explain this with bad luck
can we explain the differences in cancer risk between different tissues? how?
YES - differences in cancer risk can be explained by the total number of stem cell divisions in those tissues
(i.e. correlation btwn lifetime risk and # of total stem cell divisions in a tissue)
can we explain differences in cancer risk between different people and between different populations? what does this mean?
NO - cancer is not due to bad luck bc cancer risk is due to population
what is R^2?
proportion of variation on Y axis that can be explained by variation on X axis
what is epidemiology?
study of patterns and causes of disease in a population
what is cancer surveillance?
burden of disease, incidence, and mortality trends
what is cancer risk?
assessing candidate etiologic factors
what is cancer prevention?
assessing efficacy and impact of screening, chemoprevention, etc.
what is cancer survival?
assessing prognostic factors, determinants of quality of life
2 ways to measure cancer occurence
- Number of cases
- Rate of cases
why is measuring the number of cases helpful?
look at number of new cases, new deaths –> for health system planning
what are the 2 types of rates we can measure?
- INCIDENCE RATE - new cancer cases in a population per person-years
- MORTALITY RATE - new cancer deaths in a population per person-years
what are the units of cancer rates?
cases per person-time
why is measuring cancer rates helpful?
accounts for population size and time frame so helpful for measuring risk and causality
how does cancer incidence change with age? why?
higher cancer cases in 65-69 year olds than 85-89 year olds but lower incidence rate
more 65-69 year olds than 85-89 year olds in general –> more diagnoses
how do incidence rates of cancer change in higher income countries?
higher incidence rates in higher income countries
how does age structure vary?
- varies over SPACE –> ppl live longer in higher income countries
- varies over TIME –> ppl are getting older
why do we need direct age standardization?
because age structure varies
how do we do direct age standardization?
- define a standard population age distribution
- calculate what cancer incidence rate WOULD HAVE BEEN if it had the same age distribution as the standard
what are the 2 standard population age distributions that we can use for direct age standardization? and their specific purposes
- 2011 census population in Canada –> to compare diff time points
- 1960 world population –> to compare diff countries
how do you calculate what cancer incidence rate WOULD HAVE BEEN if it had the same age distribution as the standard? ex. for 2011 census population (3 steps)
- calculate % of population in each age group in 2011
- calculate age-specific incidence rates of given year X
- multiply age-specific incidence rates by proportion of that age group in 2011 and sum over all age groups
after standardization:
all populations have the same standard age distribution so now we can compare populations with different age structures (time points or countries) to see differences in cancer risk
how do we see if the risk of cancer in Canada is increasing over time?
Canadian population has increased and gotten older so we expect more cancer cases but must adjust:
1) for changes in population size –> calculate crude incidence rate (# new cases divided by population size) –> this increases with time!
2) for changes in population age –> calculate age-standardized incidence rate (risk in 1992 if we were as old as we were in 2011 AND risk in 2019 if we were as young as we were in 2011)
this shows that rates are not changing much
what does ASIR stand for? how is it changing in Canadians?
ASIR = Age-standardized cancer incidence rates
number of cases increasing but incidence rate stays constant
what does ASMR stand for? how is it changing in Canadians?
ASMR = Age-standardized cancer mortality rates
number of mortalities increasing but mortality rate decreasing due to reduced lung/colorectal/breast cancer mortality
what do we have to consider to determine whether someone will survive cancer? why?
must adjust for non-cancer causes of death –> remove probability of dying from other causes
lets us compare diff cancer sites, etc in diff people
why are there different incidence rates of cancer in different countries?
unequal access to vaccines, treatments, screening
why is there a high liver cancer incidence rate in Mongolia and Egypt?
high Hep B prevalence in Mongolia due to unhygienic medical and dental practices
high Hep C prevalence in Egypt
with lung cancer and smoking, how does this demonstrate that cancer takes a long time to develop?
historically, cigarette consumption increased then 20 years later we saw a big increase in lung cancer mortality
how has stomach cancer mortality changed? why?
DECREASED!
- reduced salt consumption (bc ppl stopped using salting to preserve food when fridge was invented)
- reduced Helicobacter pylori prevalence
what are the 4 etiologic factors we assess when looking at cancer risk factors?
- causes of cancer
- why do only some ppl develop cancer
- what increases/decreases risk of cancer
- what epidemiological study designs allow us to discover the causes and risk factors of cancer
describe Ochsner’s assessment of lung cancer and smoking throughout history (6)
- lung cancer was very rare
- then WWI caused increased tobacco consumption
- thus Ochsner hypothesized that lung cancer was caused by smoking
- great depression caused reduction of cigarette consumption
- once lung cancer mortality rates were recorded, Ochsner found correlation btwn sale of cigarettes and incidence of lung cancer –> but this was not causation!
- then in 1950s, 2 major epidemiological studies were released to show correlation but smoking habits didn’t change
why is it not helpful to just look at whether lung cancer cases are in smokers?
this doesn’t tell you whether smokers are more likely to get lung cancer
what do analytical studies measure?
measure association btwn exposure and health outcome
what group do analytical studies require?
control group
what are the 2 types of analytical studies and examples
- EXPERIMENTAL –> clinical trial
- OBSERVATIONAL –> cohort, case-control, ecological (decreasing strength)
what is causal inference?
looking at COUNTERFACTUALS
what are counterfactuals?
what would happen if you go back in time and remove the risk factor?
looking at counterfactuals from a study of lung cancer and smoking, what would you see if smoking caused lung cancer?
the counterfactual, absence of smoking, would not cause lung cancer
with smoking, would cause lung cancer
looking at counterfactuals from a study of lung cancer and smoking, what would you see if smoking did not cause lung cancer?
the counterfactual, absence of smoking, would cause lung cancer
with smoking, would cause lung cancer
are counterfactuals possible? why do we use them?
not actually possible but helpful to figure out the type of study
can we observe causal effects? why?
NEVER –> can only infer
bc we cannot observe the counterfactual
how can we infer causal effects for lung cancer and smoking?
compare outcomes in smokers and non-smokers to estimate causal effect
how would you design a clinical trial to see if smoking leads to lung cancer?
randomization of 1 smoking group and 1 control group –> then control risk of lung cancer
BUT THIS IS NOT ETHICAL!
what is the benefit of clinical trials for cancer research?
for cancer treatments and screening
what are the 5 downsides of clinical trials for cancer research?
- unethical to randomize potentially harmful exposure
- can’t randomize exposure
- need large population
- long trial
- limited number of comparisons can be made
why can we not randomize exposure for clinical trials for cancer studies?
exposure could be the environment –> can’t control this
and even if it were ethical to make someone smoke, can’t control whether someone smokes or not over long time
why do we need a large population for clinical trials in cancer studies?
cancers have low incidence
what is a case control study?
compare ppl with disease (cases) and without disease (controls) and see if they were exposed in the past
what is the odds ratio/risk ratio?
measure of ASSOCIATION or correlation btwn exposure and outcome
how do you calculate odds ratio/risk ratio?
(odds of exposure in CASES)/(odds of exposure in CONTROLS)
what does it mean when RR = 1
exposed and unexposed ppl have same risk of disease
what does it mean when RR > 1
exposed ppl have higher risk
what does it mean when RR < 1
exposed ppl have lower risk
is RR a measure of associaton or causation?
ASSOCIATION
what is the example of Graham and Wynder’s case control study for smoking and lung cancer?
surveyed ppl with and without cancer about their smoking habits
What did Graham and Wynder find in their case-control study? What was the criticism?
people who smoked had higher odds of having lung cancer
criticism bc retrospective study –> could be something else contributing
are retrospective or prospective studies better?
prospective
what is a cohort study? retrospective or prospective?
PROSPECTIVE
subjects have exposure now, then follow to see if they develop disease
what do cohort studies use instead of randomization?
relies on natural variation in a population to see differences in exposure
describe Doll and Hill’s cohort study and results
sent questionnaire to doctors about their smoking habits
followed doctors for 50 years to check for lung cancer deaths –> higher lung cancer mortality if heavier smoker
what is the problem with observational data?
didn’t randomize subjects for confounding factors –> ex. environment may play a role
describe the cohort study that looked at whether coffee causes cancer and the problem with the study
followed coffee drinkers to see if they develop cancer –> found INCREASED risk with more coffee
but confounding factors! coffee drinking strongly correlated with smoking –> cannot determine if coffee or smoking is causing the cancer
how can we control for smoking being a confoundng effect? and the results
STRATIFICATION –> separate data into groups based on smoking habits to adjust for smoking
look only at ppl who don’t smoke –> no significant relative risk btwn coffee drinkers and non-coffee drinkers
look only at ppl who currently smoke –> no significant relative risk btwn coffee drinkers and non-coffee drinkers
therefore, coffee doesn’t not increase the risk of cancer mortality
what is information bias? when does this happen?
error due to incorrect measurement of exposure and/or health outcome
ex. when ppl make mistakes in questionnaires
5 general steps for determining that HPV causes cervical cancer?
- HSV was prime candidate
- found HPV DNA in cervical cancer tumours –> now need epidemiological evidence
- case-control study using NAH to detect HPV but method not sensitive enough!
- case-control study using PCR to detect HPV with very good sensitivity
- determined HPV does cause cervical cancer
what did the HPV and cervical cancer show us?
that you need strong epidemiological studies, not just lab studies!
describe cohort study to see if NSAIDS reduce the incidence rates of colorectal cancer and the method of measuring time
see if they develop colorectal cancer after using NSAIDS
used PERSON-YEARS –> denominator of how long you follow the women, total number of years across all women
why did we need to control for age in the NSAIDS and colorectal cancer study?
some women had been using aspirin for >20 years so they were older than the other women
what were the results of the NSAIDS and colorectal cancer study?
using aspirin for more years = lower risk of colorectal cancer
after the observational cohort study, what did researchers do? and the results
did clinical trials –> randomized patients with and without aspirin, then see colorectal cancer rates
saw reduced cancer incidence and mortality –> therefore the observational data supports this experimental data
what are 4 things we consider when assessing preventative measures?
- what exposures cause the most cancers?
- are risk exposures modifiable
- can we screen for cancer?
- does screening reduce cancer incidence/mortality?
what is the population attributable fraction (PAF)?
the proportion of cancer cases that are attributable to a specific exposure
what 2 things does PAF depend on?
- RR measured in cohort/case control study
- prevalence of exposure measured with surveys
what happens to PAF if prevalence of exposure and RR are higher?
PAF is higher –> therefore more cancer cases caused by this exposure
if RAF = 95% for lung cancer cases and smoking what does this mean?
95% of lung cancer cases are attributable to smoking
how do we determine PAF using a graph?
Y = PAF
X = prevalence of exposure
with pre-defined curves for different RR values
if we know RR and prevalence of exposure, match up the point on the line to find PAF
what is screening? who do we screen?
examining ASYMPTOMATIC people to identify disease/disease precursor before it becomes symptomatic
do we screen to diagnose ppl?
no!! diagnosis only when they have symptoms
what are 2 requirements for screening?
- must be preclinical –> undiagnosed, asymptomatic but detectable
- must be benefit of this early treatment over late treatment
describe the study set-up for colorectal cancer screening
randomized control trial of flexible sigmoidoscopy screening
why do we need a lot of ppl for screening studies?
all asymptomatic so very few will have cancer
in screening studies do we follow them for long or short time? why?
follow for long time –> takes long time for screening to be effective
describe the results of colorectal cancer screening study
at the beginning, more cases in intervention group bc we are actively looking for the cancer
then later, intervention goes below control bc they can remove the polyp and prevent cancer
describe the results of ovarian cancer screening study
incidence is higher for intervention group bc the screening detects the actual cancer, not pre-cancer so maybe over-diagnosis
may be slightly higher mortality rate
why is it more harmful to screen for ovarian cancer?
over-diagnosis –> maybe some ppl would have remained asymptomatic
but then they receive harmful cancer treatment
what does the canadian task force on preventive health care determine about screening?
evaluates systematic review evidence to develop consensus recommendations for preventive interventions
which 4 cancers is screening recommended?
- Colorectal cancer
- cervical cancer
- lung cancer (if heavy smoker)
- breast cancer
which 2 cancers is screening not recommended?
- pelvic exams for non-cervical cancers
- prostate cancer
