Intro and Evaluating Evidence Flashcards
What is the importance of chronic disease prevention?
What is chronic disease?
- Diseases of long duration and slow progression
- May be non-communicable (non-infectious)
- Examples: cancer, heart disease, diabetes
- Accounts for >60% of all deaths globally
Today in Canada, 84% of males and 90% of females reach at least age 65.
Why this change? [5]
- Sanitation
- Improved medical care
- Vaccinations
- Improved social safety net
- Nutrition (decreased prevalence of deficiency diseases; however, current diets are increasing risk of chronic diseases)
Will today’s children die earlier than their parents?
Maybe not, but their quality of life in later years is decreasing due to chronic disease.
What is a risk factor?
- Any substance or condition that increases one’s risk (likelihood) for developing the condition.
- Modifiable = can be altered (e.g., smoking, diet, physical activity, sleep patterns)
- Non-modifiable = cannot be altered (e.g., age, genetics)
What are some dietary risk factors? [10]
- High sodium
- Low nuts/seeds
- High processed meats
- Low seafood omega-3 fats
- Low vegetables
- Low fruits
- High sugar-sweetened beverages
- Low whole grains
- Low PUFAs replacing CHO or saturated fats
- High red meats, unprocessed
What are the Bradford Hill Causal Criteria? [5] + [4 bonus]
What are concerns of the first 5?
-
Consistency of association: relationship observed repeatedly
- concerns: consistent errors across studies; inability to find consistent relationships due to methodological differences across studies (different tools used to collect information, different populations studied)
-
Strength of association: effect size
- concerns: often only weak associations in nutrition studies
-
Dose response: statistically significant linear trend
- concerns: threshold effects → nutrient/outcome relationships are not always linear; misclassification → food records vs. food frequency questionnaire
-
Biological plausibility: theoretical explanation/mechanism
- concerns: unknown mechanism for effects of nutrients on diseases; foods vs. nutrients (complex systems); what part of the food is responsible for the effect?
-
Temporality: exposure precedes the outcome
- concerns: did diet cause disease? Or did disease cause a change in the diet?
Bonus:
- Specificity: exposure causes a specific outcome
- Analogy: comparison to a similar biological system can be made
- Coherence: fits in with existing knowledge
- Experiment: true experimental study finds the association
What are 5 challenges is measuring change?
- The level of original measurement matters.
- There may be a minimal detectable difference (i.e., can only detect differences greater than the measurement error of the instrument used)
- Starting points matter → floor/ceiling (i.e., already as high/low as it can get, outcome won’t differ) effects; regression to the mean)
- Variables change naturally over time (e.g., blood pressure)
- Tools of measurement must be both reliable and valid.
What are some advantages of survey research? [3]
- Cheap and convenient
- Can survey a broad geographic area and large number of subjects
- Relatively inexpensive compared to other research methods
- May serve as a basis for hypothesis development and future research ideas.
What are limitations of survey research? [4]
- Non-response bias
- Interviewer bias
- Misreporting
- Poorly designed questions (may be misunderstood / might be asking the wrong thing)
Compare cross-sectional, cohort, and case-control studies regarding the timing of outcome and exposure.
Cross-sectional: O + E
Cohort: E (follow over time) O
Case-control: O (look back for) E
What are advantages of case-control studies? [4]
Useful for studying:
- Rare conditions
- Conditions with a very long lag between exposure and outcome
- Usually requires fewer subjects than cohort or cross-sectional studies
- Generally quicker and cheaper than cohort studies
What are limitations of case-control studies? [6]
1. Difficult to establish temporality
Did exposure precede disease OR did diet/lifestyle change because of disease?
2. How far back?
Must determine appropriate time period in the past in terms of biological plausibility
E.g. For folic acid and neural tube defect – appropriate to ask about intake from 1 year ago BUT for ω-3 fatty acids and heart disease, how far back to ask about? – 1 year, 5 years, 10?
3. Reliance on Memory
Difficult to recall past practices accurately
4. RECALL BIAS
Cases more motivated than controls to search their memory to make sense of why they have the condition
Thus, cases may be more likely to report past exposure
5. Misclassification
Misclassification of cases
Misclassification of exposure status
6. Cannot calculate disease incidence or prevalence
Proportions of cases in the study sample ≠ proportions with the outcome in the population
What is the difference between an odds ratio and risk?
Risk (Probability) = likelihood of event of event A /total number of events
Odds = likelihood of event A /likelihood of alternate event
Example: Risk vs. Odds of rolling a 6 on a die
Risk = 1/6
Odds = 1/5
How is an odds ratio interpreted?
OR > 1 → exposure more likely in cases than controls (There is a positive association between the exposure and outcome)
OR < 1 → exposure is less likely in cases than controls (There is a negative association between the exposure and outcome OR there is a protective effect of the exposure)
OR = 1 → no difference in odds of exposure between cases and controls
When is an odds ratio considered statistically significant?
An OR is considered statistically significant when the confidence interval (CI) excludes 1.0
The Confidence Interval is the range of values within which the true population value likely exists
Eg. 95% CI
95% chance that the true value for the population lies within the CI
Eg. 99% CI
99% chance that the true value for the population lies within the CI
For α = 0.05 (Type I error), a 95% (1 – α) CI is used
Eg. OR 3.01 (95% CI 1.86-4.85)
Sample odds ratio = 3.01
95% chance that the true population value lies somewhere between 1.86 and 4.85
Generally, the larger the sample size, the smaller (and more precise) the CI
What three questions should you ask yourself when you interpret an odds ratio?
Is the OR >1 or <1?
Does the CI include 1?
Is the CI very wide or very small?
What is considered a ‘strong’ odds ratio?
>4 or <0.25 = strong
2-4 or 0.25-0.5 = moderate
1.5-2 or 0.5-0.67 = possible but weak
<1.50 or >0.7 = possible but very weak
Which of the Bradford Hill Criteria for Causation cannot be shown with case-control or cross-sectional studies?
Temporality
Describe the advantages of prospective cohort studies. [5]
- Can establish timing and directionality of events
- Can calculate incidence (new cases)
- Avoid potential for recall bias
- May be easier (or at least more feasible) than a randomized controlled trial
- If well designed, can assess many outcomes
Describe limitations of prospective cohort studies. [4]
- Confounding: Exposure may be related to another (unknown) factor that is associated with the outcome
- Level of exposure may change over time
- Loss to follow up: Particularly problematic if people who withdraw from the study have different characteristics than those who remain involved
- Large sample size is needed if outcome of interest is rare
How is relative risk interpreted?
RR < 1 → outcome less likely to occur in exposed group than non-exposed group
(Negative association between exposure and outcome)
RR >1 → outcome more likely to occur in exposed group than non-exposed group
(Positive association between exposure and outcome)
RR = 1 → no relationship between exposure and outcome
With RR can say:
“Exposed are x times as likely to have the outcome”
Eg. RR = 0.70, Subjects who took Vitamin C supplements were 0.70 times (or 70%) as likely to get heart disease
Eg. RR = 2.2 Subjects who drink milk were 2.2 times as likely to have strong bones compared to those who did not drink milk
When is relative risk statistically significant?
RR is statistically significant when the confidence interval does NOT include 1
What type of studies can odds ratios be calculated for?
What about relative risks?
Odds Ratios may be calculated for cohort studies, case-control studies or experimental studies
Relative Risks are used in cohort studies and experimental studies (where risk is compared between treatment group/control group)
RR should NOT be calculated for case-control studies
Describe advantages and limitations of cross-over design.
Advantage: Subjects serve as their own control, therefore, no differences between treatment and control group; Can use smaller sample size
Limitation: Can not use crossover design if treatment leads to a permanent change (Eg. education program)
What are advantages of randomized controlled trials? [5]
- Controls for possible confounders → aims to have comparable groups by randomization
- Level of exposure controlled/manipulated by researcher
- Control group for comparison (ensure effects are due to treatment)
- If blinded, reduces bias
- Provides evidence about temporality of associations
What are limits to randomized controlled trials? [5]
- Cost
- Feasibility → ethical implications
- Large sample size needed for statistical power
- Poor compliance and drop outs
- External validity/generalizability
What are 5 nutrition RCT challenges?
-
Duration
- Longer interventions → difficult to have compliance with dietary interventions over long term
- Shorter interventions → long enough to have impact?
-
Controlling exposure
- Difficult to completely ‘control’ intake of a nutrient
- Difficult to have good compliance with nutrition interventions
-
Level of exposure
- Cannot have zero intake group of essential nutrient
- Typically comparing ‘some’ intake with ‘more’ → how to define ‘some’ and ‘more’; ‘some’ may already be enough for outcome of interest → threshold effects
-
Nutrients vs. food
- Will supplement of omega-3 fatty acid produce the same effect as obtaining omega-3 fatty acid from eating fish?
-
Difficult to accurately measure nutrient intake
- Rely on memory and subject’s ability to accurately recall foods consumed & portion sizes
- Foods consumed vary meal-to-meal and day-to-day
- Similar foods may vary in nutrient content.
When is OR, RR, or mean difference significant?
OR, RR → significant if CI does not cross/include 1
Mean difference → significant if CI does not cross/include 0
What are advantages of systematic reviews? [3]
- Summarizes current research
- Systematic methods for data retrieval limits potential for bias
- Considered to be evidence of the highest level in the hierarchy of research designs.
What are limitations of systematic reviews? [4]
- Limited by publication bias
- Differences in methods for selecting and assessing quality of papers that may affect results
- Averages may miss effects in certain groups
- Systematic reviews may not always be updated → to ensure information is up to date:
- Check publication date
- Look for new papers published after that date
→ Results from systematic reviews and meta-analyses still need to be interpreted critically.
One of the issues when trying to find a dose-response relationship in nutrition is that nutrient-outcome relationships […].
Generally do not have linear relationships
What are some limitations of cross-sectional studies? [4]
Limitations: Non-response bias; misreporting, cannot show disease incidence or temporality
A confounding factor influences only the outcome.
True or False?
False.
Confounding factors influence both the exposure and the outcome.
A confounding factor influences both the outcome and the exposure.
True or False?
True.
What is an ecological study?
In epidemiology, ecological studies are used to understand the relationship between outcome and exposure at a population level, where ‘population’ represents a group of individuals with a shared characteristic such as geography, ethnicity, socio-economic status of employment.
Describe the hierarchy of evidence for clinical decisions.
Describe sigmoid relationships and threshold effects in nutrition science.
- A standard dose-response does not generally appear in nutrition as it does in pharmaceutical science.
How are forest plots interpreted?
- Each ‘black box’ shows the OR (RR or mean diff) for an individual study → the lines extending from the box shows the 95% CI
- The size of the black box reflects the weight of the study in the meta-analysis (and usually, the sample size)
- The clear weighted diamond is the weighted mean of all the studies
Of all the 9 studies in this forest plot, how many showed a statistically significant effect of the treatment?
3
What is a forest plot?
A forest plot, also known as a blobbogram (lol!), is a graphical display of estimated results from a number of scientific studies addressing the same question, along with the overall results.
In this example, the forest-plot shows that overall, eating processed meat is associated with increased risk of colorectal cancer, and this was statistically significant.
What is the difference between an odds ratio and risk?
Risk (Probability) = likelihood of event of event A total number of events
Odds = likelihood of event A likelihood of alternate event
Example: Risk vs. Odds of rolling a 6 on a die
Risk = 1/6
Odds = 1/5
What is an odds ratio?
= Odds of exposure among cases Odds of exposure among controls
Exposure refers to independent variable, could be nutrient intake, supplement intake etc.
What is relative risk?
Likelihood that outcome occurs in exposed group compared to unexposed group
Ratio of the risk among exposed to the risk among unexposed
Null hypothesis → relative risk = 1 (no relationship)
E.g., Subjects who took Vitamin C supplements were [RR] as likely to die from heart disease as subjects who did not take Vit C supplements
When is relative risk statistically significant?
RR is statistically significant when the confidence interval does NOT include 1
What is risk difference?
Risk Difference = (Risk of outcome in exposed group) – (Risk of outcome in unexposed group)
Null hypothesis: RD = 0
Which of the Hill Criteria for Causation are met by this evidence?
ALL, but strength of the relationship because RR =/= 2.
However, in nutrition studies, a strength of relationship RR of 2 is not necessary to decide causality.
