LIFESCI 2A03 Final Content

Exposure

The word used in observational studies for the independent variable (ex. age, education status, smoking)

Intervention

The word used in interventional studies for the independent variable (ex. administration of a drug, use of a therapeutic procedure)

Risk ratio

The probability of an exposure causing an outcome. Reported in cohort studies, cross-sectional studies, and clinical trials

P(event) in exposed/P(event) in non exposed

Odds ratio

Probability of an event occurring versus not occurring. Reported in case-control studies and cohort studies.

Which study provides the strongest evidence of a causal relationship?

Cohort studies because:

• Grouped based on the independent variable (exposure)

• Longitudinal

When are the odds ratio and risk ratio the same?

When the event is rare

Interpreting RR and OR when P and CI are statistically significant

>1 = increase of R/O in exposed group

<1 = decrease of R/O in exposed group

Main cause of cancer

Uncontrolled and increased cell proliferation

Tumour heterogeneity

Many types of cells make up the tumour

Hierarchical model of tumour growth

1. Cancer stem cells (CSCs): high self-renewal capacity, drives cancer growth and relapse

2. TA cells: limited proliferation

3. Non-proliferative cancer cells

Driving force of carcinogenesis

Accumulation of DNA alterations and mutations in dividing stem cells

Evidence of the SC division theory of cancer

• Most cancers arise in tissues with high turnover and are uncommon in tissues with low turnover

• Cancer incidence increases with age, especially in issues with high turnover

• Organs/tissues with very low regenerative capacity are less likely to develop primary tumours

Prognosis

Outcome of patients with the affliction

Neural stem cell niches

1. Subventricular zone (SVZ)

2. Subgranular zone (SGZ)

NSC niches and GBM relapse

NSC niches may be sources of CSCs that drive GBM relapse

Incidental research

Researchers are examining an incidental outcome that was a side effect of something else and therefore don’t decide the protocol

Hazard ratio

Same thing as risk ratio but in the context of survival, used in survivorship studies

Prognostic factors

Predictor of the prognosis of a patient (could be confounders in a study)

K-M analysis + purpose

• Statistical analysis used to measure the probability of patient survival time after an intervention

• Only analyzes 1 prognostic factor

• Can compare two groups via a log-rank test

• Studies the efficacy and safety of a treatment

Progression-free survival (PFS) + what is it an alternative for?

Main event of interest is progression, measuring the time between treatment and observed progression. Alternative to overall survival (OS) since PFS appears much sooner (making studies shorter) and is not always proportional to survival

Multivariate Cox regression analysis + purpose

• Calculates hazard ratios per risk factor

• Can analyze multiple prognostic factors simultaneously

• Assess the survival of a treatment while controlling for confounders and covariates

Confounders vs covariates

Confounder: extraneous, not of interest, covariate variable linked to the predictor and the outcome variable

Covariate: 3rd variable that is linked to the outcome variable but not the predictor

Risk vs prognostic factors

Risk factors: increase the probability of developing a condition

Prognostic factors: associated with/influence the outcome in people who have the condition

Historical controls + advantages

Groups and data from past, external studies being used as control groups. Often used in clinical studies and help reduce patient recruitment + costs

4 types of comparison groups

1. Standard treatment (active)

2. Placebo

3. Dose-response (several groups with different dosages of the treatment)

4. Historical

Parallel CT design

Subjects stay in their assigned groups. Cheaper and faster

Crossover CT design + advantages

Subjects switch to the other group after a washing period. Minimized variation between groups, no additional participant recruitment

Simple randomization + ideal sample size

Subjects are assigned using pure chance. Ideal for large sample sizes

Block randomization + ideal sample size

Randomizes participants within fixed-size blocks. Ideal for sample sizes less than 100

Stratified randomization + ideal sample size

Balances groups across important covariates (ex. sex, age). Ideal for sample sizes less than 100

Why is blinding/masking important?

Increases internal validity and reduces performance, assessment, and attrition bias in experiments

Early phase 1 (0) of CT

• Exploratory

• See whether and how a treatment affects the body (NOT therapeutic or diagnostic)

• Micro doses

• Very small number of participants

Phase 1 of CT

• First application in humans

• Test safety + tolerability with low doses

• n=20-100

• Healthy volunteers

Phase 2 of CT

• Therapeutic trial phase

• Finding the optimal dose + safety

• n=100-500

• First application in patients with disease

Phase 3 of CT

• Therapeutic confirmation

• Confirming the safety + effectiveness and comparing it to other drugs

• n=1000-5000

Phase 4 of CT

• Long term observation

• Therapy optimization + safety monitoring to see interaction with other drugs

• Very large number of patients

Arms in CTs

Another word for groups. Comparative arm = control group

Primary outcome

The most important/of interest outcome for evaluating an intervention. 1 or more can exist in a study

Secondary outcome

Not as important as the primary outcome, but still of interest. May become primaru in future studies. 1 or more can exist in a study

Objective/hard outcomes

Unambiguous and consistently measured, ideal to minimize bias and improve validity

Subjective/soft outcomes

Based on interpretation of the assessor. If used, attempts should be made to reduce bias (ex. calibrated tools, blinding, randomization, standardizing data collection)

Summary statistics + chart types + issues with it

Summarizes data, therefore hiding raw data. Includes line charts, bar charts, and boxplots. Important information, such as distribution, outliers, and variability, is hidden, which may lead to an incorrect conclusion.

What chart is suitable for small data sets?

Scatterplots with bars representing the mean

What chart is suitable for medium data sets?

Bee swarm plot with error boxes representing variance

What chart is suitable for large data sets?

Dot plots with boxplots as visual aids

Pie charts + issues

Used to display categorical proportions using size. Represents 1-dimensional data + human perception struggles to compare areas

Boxplots versus beeswarm plots

Boxplots: summary statistics for large data sets
Beeswarm: show individual data points, provides raw data, for medium data sets

What is the recommended colour palette + aspect ratio for graphs?

Monochromatic/colour blind safe palettes, landscape format that is 50% wider than tall

What effect does sample size have on variation?

Reduces variation, which increases statistical power

What graph should be used to represent continuous data + what should it not be used for?

Line graphs. Do not use for discrete data, creates misleading continuity

R²

Correlation coefficient, describes the strength of the correlation between two variables

Placebo effect

Patients believe that they are experiencing a positive, real effect from something that is not medicinally active

Psychological mechanism behind the placebo effect

Classical conditional, which creates an association between two stimuli (some factor of the treatment + the experienced effect)

Physiological mechanism behind the placebo effect

The brain can release chemicals so that a real change is experienced

Nocebo effect

Negative expectations about a treatment cause an adverse reaction

How is the placebo effect controlled for?

Including placebo groups as negative controls, double-blinding, randomization, and minimizing variation between groups

When do you fail to reject Ho?

If the p-value > alpha and the confidence interval includes the null value (0 for means, 1 for RR/OR)

Which methods of controlling for confounders/covariates do you NOT need to know the specific confounders/covariates for?

Randomization of a large sample size and block randomization

Restriction

Controls for known confounders by restricting the study to 1 confounder (ex. a study consisting only of females)

Matching

Controls for known confounders by balancing variables across groups to reduce variation

Is randomization feasible for observational studies?

Usually no

Immunohisto/cytochemistry

Determines the expression of proteins in cells, fluorescently labels them. Allows us to find the expression of a certain gene and idenitify specific cells

RNAi

Knocking down genes to study cause-effect relationships and gene function in specific processes

RT-PCR

Detects gene expression on the transcriptional level by measuring mRNA

Western blot

Detects gene expression on the translational level by measuring protein

Clinical vs pre-clinical experiments

Clinical: provides relevant info for expression and prognosis in humans

Pre-clinical: opportunity to study cause-effect relationships, easier to manipulate (less ethics), but not feasible to conduct CTs

Describe the 3 trends of clinical trial phases

• Participant number increases

• Number of unhealthy subjects/subjects with the target affliction increases

• Length of the phase increases

Emergency use authorization (EUA)

A vaccine hasn’t completed clinical trial but is approved for use. Compromises ongoing clinical trials because it is unethical to withhold an effective treatment from the placebo group

Ethical crossover

Giving the placebo group the intervention after the trial is complete (you can;t withhold it from them ethically)