LifeSci 2A03 Test 1

Case-control studies

Observational studies aim to study ....

observational (in particular epidemiological) studies aim at studying the patterns (distribution) and causes of diseases in people to control them.

during Bias

performance:

(Placibo effect? We want the outcome to come out in a certain way) so differences in care provided to study groups (e.g. from knowing the application of treatments)

attrition ;recall/ report bias

Incomplete or loss of data,inaccurate data due to errors in bias recalling or recording (self reporting)

assessment/observer/ measurements/ detection bias:

errors in assessing/recording outcome (e.g. wishful thinking; fabricating/manipulating data)

How to avoid allocation/sampling/recruitment/ selection bias :

Come up with the plan first

random selection and allocation of subjects + allocation concealment (blinding, people don't know which group they are going); rigorous criteria; appropriate sample size (bigger, better); prospective trials (collecting data in future, you have more freedom than retrospectctive bc it is already collected i cant change how it was collected)

What is science?

The pursuit and application of knowledge and understanding of the natural and social world following a systematic methodology based on evidence.

What are the 6 characteristics of science?

1. Based on empirical knowledge

2. Provides rational/natural explanations

3. Repeatable and reproducible

4. Testable

5. Involves experimentation

6. Generality of principles

What does it mean to say that science is based on empirical knowledge?

- Knowledge derived from observation or experimentation.

- Data collection (qualitative or quantitative).

- Observations - often the starting point of a scientific investigation.

What does it mean to say that science provides rational explanations?

- Seeking explanations in terms of rational/natural causes.

- Very few questions are off-limits in science, but the answers science can provide are limited to the natural world.

- Supernatural causes are not used in science.

- Regeneration is not miraculous - there is a biological explanation and can be studied

What does it mean to say that science is repeatable and reproducible?

- Scientific results are always subject to confirmation by the same (repeatable) or other investigators (reproducible).

- If experiments can be repeated/reproduced via a reliable method/procedure > confirmation > reliable findings becomes more powerful

- Sometimes due to impracticality, temporal difficulties, or expense, it's not possible to repeat/reproduce experiments > observation lack reliable confirmation

What does it mean to say that science is testable?

- A neat theory is not sufficient. A scientific explanation (hypothesis) To a question or problem must be testable through observation or experimentation. For an idea to be testable, it must generate specific expectation. (Prediction)

- A set of observations and results are expected if the idea is true or false.

- Sometimes to test ideas, the development of new tools may be required

What does it mean to say that science involves experimentation?

- Experiment = a planned, controlled setting where there is intervention/manipulation of a natural process to observe the effects.

ex: cut a limb of a salamander and observe growth of a new limb daily

- An organized and detailed series of steps to validate or reject a hypothesis.

What does it mean to say that science seeks generality of principles?

Generality is the establishment of universal principles that express how the natural world works under certain conditions.

How general depend on the condition

common patterns:

law and theories

What is research?

The systematic, planned, and multiple-step process that uses discoveries to advance knowledge.

If purpose is contribution towards science> scientific research

What are the steps of the scientific method?

1. Define/identify a problem/question (usually drives from observation)

2. Formulate a hypothesis (potential explanation)

3. Test hypothesis.

4. Analyze data/results.

5. Does the data support the hypothesis? If no, formulate a new hypothesis and perform new experiments.

6. Draw conclusions.

7. Communicate results.

Why scientific research is important?

- To provide solid knowledge by identifying and characterizing new phenomena, materials or organisms

- To solve a scientific problem for which solution is needed

- To develop scientific inventions (e.g. cloning, gene therapy)

- To provide basis for many government policies and industry productivity

What are the types of research studies?

Primary, secondary, fundamental and applied

What is primary research?

A study that involves the collection and analysis of original or raw data. There are many types; for example: observational, experimental, etc.

What is secondary research?

A study that involves further analysis of studies already performed, such as narrative/traditional literature reviews, systematic reviews, or meta-analyses.

What is fundamental research?

The investigation of basic principles and reasons for occurrence of a particular event, process or phenomenon.

• AKAS 'theoretical' or 'pure' research=to expand knowledge

• Curiosity-driven.

• The outcome of basic research forms the basis for many applied research studies. non-clinical (subjects other than humans)

What is applied research? (+types and explanation)

concerned with solving practical problems

of interest to humans such as understanding human disease via the use of model organisms or systems - pre-clinical (non-human) : Pre-clinical research is also referred to as translational research, as it refines and advances the application of a discovery or knowledge often learned via basic research studies, bridging thus the gap between basic and clinical research.

vs. clinical (humans as subject): Unlike non-clinical research, clinical research is the study of health and illness in people. It is designed to learn how to prevent, diagnose, and treat illness

Classification of a study with respect to the nature of the study (type of question addressed)

Descriptive (who, when, where) study does not try to quantify the relationship between variables but instead tries to give us a picture of what is happening in a population, i.e. observing the basic features of its distribution in terms of time, place, and person (e.g. patterns of disease occurrence within a population).

Analytical (why,how) relationship b/w variables (e.g. the effect of an intervention or exposure on a given outcome --- ie. why something is happening)

Classification of a study with respect to the method used for data collection

Observation ( no manipulation of system) or experimental,interventional (manipulation of system) system=organism+enviro that organism live

Classification of a study with respect to time

longitudinal (Prospective, retrospective) = over time (takes time), prospective: collecting data till future (data doesn't exist), retrospective:looking back at history of data (data already exist, maybe it hasn't been use before or published) cross-sectional: snap shot (short time), no separation in time

Classification of a study with respect to data analysis

Qualitative, quantitative, or mixed

Why is scientific research IMPORTANT?

- Provides solid knowledge by identifying and characterizing new phenomena, materials or organisms.

- Solves a scientific problem for which a solution is needed.

- Develop scientific inventions.

- Provides the basis for many government policies and industry productivity.

Human samples initially used for chromosome analysis were derived from

testicular tissue in which haploid meiotic cells were often present

Inductive Reasoning

1. Observation

2. Pattern

3. Tentative Hypothesis

4. Theory

(specific observation- generalization):

- Summarizes a set of observations and serve to generate hypotheses and develop theories.

- "This ice is cold, therefore all ice is cold".

Deductive Reasoning

1. Theory/RQ

2. Hypothesis + Prediction

3. Observation/Experimentation

4. Confirmation

(generalization- specific observation)

- Used to test whether hypotheses and theories developed by induction are 'correct'.

- "If all green apples are sour, then this green apple will be sour".vv They both interplay

What is a hypothesis?

A statement or idea proposed as a tentative explanation to a research question based on previous observations. It should be simple, specific and testable. Implies a relationship between 2 variables (independent and dependent) and leads to predictions. (expectation based on the hypothesis). question -> hypothesis -> prediction -> experiment

Testing a Hypothesis

1. Generate predictions by using the rules of deductive logic. Prediction more detail what we expect/ help us to conduct the experiment and data how to perform study

2. Collect (via observation or experimentation) and analyze data in order to Reject (or fail to reject) a hypothesis (CONCLUSION)

Deductive syllogism

If Hypothesis then Prediction

Evidence that is consistent with your prediction will

strengthen a hypothesis but will never let us know for certain that the hypothesis is true.

If you have observations that are inconsistent with your prediction

then you reject your hypothesis.

Null Hypothesis h0

Usually a statement of no effect or no difference. A hypothesis that there is no significant difference between two sets of results.

Alternative/Alternate Hypothesis ha

A hypothesis that is contrary to the null hypothesis. Usually the research hypothesis.

Let's assume that a scientist fails to reject the null hypothesis, what do you ask?

1. What other confounding factors could have affected their data collection and the fact that they did not observe differences between groups?

2. What could they do next?

Positive controls

Groups where a phenomenon is expected. They ensure that there is an effect when there should be an effect, by using a treatment that is already known to produce that effect. (show that the system works!)

Give you reference point by giving you view on extremes/ findings of another study when i repeat it can become my control if I do the exact same presudure (reference point)

Negative controls

Groups where no phenomenon is expected. They ensure that there is no effect when there should be no effect.

Give you reference point by giving you view on extremes/ findings of another study when i repeat it can become my control if I do the exact same presudure

Why include pos/neg controls?

- They serve as reference points.

- They allow us to control for confounding factors (other variables that may affect the results).

Hypothesis-driven studies use

the scientific method and deductive reasoning to test hypotheses.

Experimental Design

The design of an experiment refers to the structure of the experiment with particular reference to:

- The set of treatments included in the study.

- The set of experimental units included in the study.

- The rules and procedures by which the treatments are assigned to the experimental units or vice versa.

- The measurements that are made on the experimental units after the treatments have been applied.

Somatic

Give rise to body cells during organogenesis, tissue maintenance and tissue repair.

Correlation

indicates the existence of a trend (e.g. increasing,

decreasing) between two or more variables

- explained by some equation e.g. linear regression - strength/degree of relationship measured by corr. coefficient

Causation

indicates that one event (dependent/response

variable) is the result of the occurrence of another event (independent/predictor variable)testing

Case Study: Brain Freeze

Case Study: Brain Freeze

• Ha: the rate at which you eat ice cream affects whether you will experience an ice cream headache

• Ho: there is no effect of the rate of eating ice cream on the incidence of ice cream headaches

ie. there is no statistical difference in either observed effects (they both either cause headaches or they don't)

• The experiment: 145 middle school students were randomly placed into 2 groups:

• Eat slowly group: eat ice cream in over 30 seconds (100mL)

• Eat fast group: eat it in 5 seconds (100mL)

• Results: found that in slow group, 9% had headaches whereas fast group 27% has headaches

a chi-square test can be done to see it these values are statistically signiticant

• Drawing conclusions.

if researchers rejected the mo: data supports Ma or is consistent with the ha

if failed to rejected the Ho: they didn't detect a difference between groups and there is no

evidence to support Ha this basically means they accepted the null hypothesis

Type 1 (alpha) error

False positive)- rejecting the null hypothesis when it's true/which the null hypothesis is rejected, but in reality we should have failed to reject he is pregnent

type 2 (beta) error

(False negative) - non-rejection of a false null hypothesis/

a type Il error is one in which we fail to reject the null

when we should have rejected it she is pregnent

Null hypothesis significance testing (NHST)

1. Set α = level of significance = type I error (e.g. α = 0.05 → arbitrary!)

2. Choose statistical test (e.g. X2 test) + calculate test statistic and P value

If P < 0.05 (OR CI does NOT include the 'null' hypothesized value) → Significant differences between groups → Reject Ho → fail to reject Ha (STATISTICAL conclusion) → Evidence supports association/effect (BIOLOGICAL conclusion)

• If P ≥ 0.05 (OR CI DOES include the 'null' hypothesized value) → No significant differences between groups → Fail to reject Ho → reject Ha (STATISTICAL conclusion) → No evidence of association/no effect (BIOLOGICAL conclusion)

Confidence interval (CI)

1. Set α = level of significance = type I error

2. Calculate CI with a Confidence Level % = 1 − α They also estimate magnitude and precision of the estimated effect/association → more common in clinical studies

DIRECTIONAL HYPOTHESIS

• RH may state 'direction' of effect/association (positive or negative)

- e.g. eating ice cream FAST increases the incidence of headaches • When? - evidence suggesting so - the opposite is not of interest or possible • H0 u≤0; Ha u>0 OR H0 u≥0; Ha u

Plays a role

Cause and effect

We can have more than 1 or more Hs under the same RQ and 1 or more P under the same H

In vitro (preclinical testing)

Use of cell/isolated cell

Ex vivo (preclinical testing)

Isolated tissue/organ

Its somewhere in between

In vivo

Model organism/ use of animal-treatment apply to full organism

clinical

Human

hypothesis generation

• Some research studies are appropriate for

hypothesis generation - e.g. descriptive studies

test hypotheses (hypothesis-driven)

Some research studies are conducted to test

hypotheses (hypothesis-driven) - e.g. analytical studies

litreture review (secondary)

Not new things just summarizing or reviewing finding. we are not analyzing data we can use it to generate hypothesis because they put studies together and say blah blah blah bring up this research question

Experimental

Why/how question- always h-driven studies prospective (bc in experiment you are gonna see effects in future) +analytical by nature (by default)

Observational

Why/how question

Analytical

H-driven studies

descriptive studies (Question)

What/who/when/where question

Cross-sectional study

It's a one time point study which is mostly analytical but it depends on purpose of a study and it can sometimes be a descriptive study too

If there is separation of time between my independent and dependent valuableI

longitudinal

IMPERFECTIONS OF SCIENCE

Science = a human enterprise = NOT PERFECT

1) Limitations of the study design (human design them )

2) Technical limitations (equipment like microscope is not perfect

3) Limited abilities of scientists to discern differences (we have limit in our eyes

4) Wishful thinking (bias, we want to conclude in a certain way therefore bias and wrong conclusion)

important to be aware of these, include them and reflect on them

Systematic error

can lead to under or over-estimation of association between variables (find a relationship thats not actually there)

can occur at any stage

It can occur:

- before (e.g. planning/design stage)

- during (e.g. data collection, data

analysis)

- after (e.g. publication)

Intentional or un-intentiona

before Bias (planning)

sampling/recruitment/ selection bias :

Unequal (not representative, ex i use online survey to grab data from participant, bias bc some people might not have access to computer and wifi to answer) representation of subjects during: sampling stage

allocation bias : Unequal (not representative) representation of subjects during: allocation stage (can also happen during)>after sampling when I want to assign them into groups

How to avoid bias avoid during?

performance:

Blinding

Attrition; recall/report bias:

prospective studies; rigorous criteria (specific guideline plan)

assessment/observer/measurement/detection bias:

a priori hypotheses; avoid 'torturing' the data; proper statistical tests; blinding; calibrated tools (you should have plan on what statical test you want to use before collecting data, don't try each test and then use the best one)

After bias (publish)

reporting:

reporting selective outcome

citation:

citing mainly studies that support own findings

publication:

preference of journals for studies with significant results or those that preferentially cite the journal

How to avoid after bias

reporting:

transparency in communication of procedures and results (e.g. raw data

citation:

equally cite both types of studies

publication:

equally accept both types of studies; - journal of negative results (bcc both provide knowledge don't cherry pick)

Case Study: Human Chromosome Count

• in 1923, Theophilus Painter reported that there were 48 chromosomes in a human cell (limitation in microscope)

Methodology:

1. had serial sections of testicular cells in metaphase and chromosomes were clumped together and curled in

2. Camera Lucida drawing of chromosomes (bad pic)

CHROMOSOME COUNT - LIMITATIONS

1. Source of tissue (testicular)

- low # of spermatogonia (low sample size)

2. Tissue not fixed - damaged chromosomes

3. Chromosomes clumped together or curl up (not spread out)

4. Chromosomes split among sections

- hard to re-construct

TECHNICAL IMPROVEMENTS

1. Fixation of tissue (preservation) - prevented fragmentation

2. Hypotonic solution + squashing - spreading of chromosomes

3. Use of colchicine

- mitotic arrest

- further DNA condensation

after using improved techniques:

- inaccurate counts due to sample

source and size

- biased by previous results - skeptical to challenge previous count

• this number was believed and accepted for 33 years - scientists used the same technique he used to

experiment

In 1956 new techniques (improved) were used

Used human embryonic fibroblasts (better source of sample more chromosomes)

- Increased sample size! to count chromosomes (now we can repeat counting)

This can change result

o 1956 - Tijo and Levan reported 46 chromosomes

• so what was wrong with Painter's approach?

Sampling error? (chromosome aberrations)

Some of the sample came from patient who had mental disorder and some of these patient might had abnormal number of chromosomes

Technical error?

1. clumping; sectioning + reconstruction (one section of chromosome could be in more than one tissue section so you count two but there is only one)

2. aberrant staining (with dye)

(Maybe it wasn't staining close to centromere of chromosomes but it was staining the arms so you basically gonna count two instead of one)

• people often see what they expect/want to see - this is bias

All of the above biases could happen

WHY AND HOW TO MINIMIZE BIAS

Researcher:

1. Be aware of all potential sources of bias (reflect) (e.g. confounders, etc.)

2. Include control groups, matching, randomization, blinding when

feasible

3. Declare limitations of the study

Editors and reviewers:

1. Create and ensure

implementation of guidelines

and policies for publication

2. Detect bias during review

process

3. Avoid publication bias

Always gonna be bias tho

IMPERFECTIONS OF SCIENCE - SUMMARY

Study design + technical limitations + judgement calls (report them)

• Unwillingness to part from previous findings OR influenced by others

• Pressure for publishing 'flashy' studies (study that show significant result)

• Evidence AND conclusions may change over time due to

changes in study design:

e.g. technological advances, size of the study

Important in methodology (chromosome case study)

...BUT: Science CAN self-correct over time = essential strength!

temporary result; it might change and its okay

REPLICATION and PEER-REVIEW is CRUCIAL!

We think a study is valid and true till you prove its not!

Study design/Research plan for H-driven studies (analytical) test hypothesis

Define variables of interest (indep + dep) + confounders

method for sampling + assignment of subjects to groups

sample size - how you gonna obtain your subject from large target population are you gonna include randomization? Etc

how data is collected + analyzed

-its important to plan carefully bc you cant change anything during study if you don't, you gonna waste time and money even more

Types of errors

Errors: Differences between observed and true values

Reflect on errors that might happen during planing stage so you can minimize it

Types (based on sources):

Random due to chance or natural variation in population - minimized by replication ( always gonna be there) even with a given group/ similar individuals - even have same diet you sugar blood gonna be different (genetically, and you cant do anything about it)

Systematic (bias) - minimized by randomization and/or blinding

Statistical (previous lecture)

CONFOUNDERS

Third valuable that is different in subjects so it can impact on relation between variables and if its not control they may effect on our conclution. So its important to reflect on them

Variable that is NOT of interest to the experimenter

It can distort estimation of association between variables of interest if not controlled- somehow related to dependent and independed values

Confounding = a type of bias

e.g. association between drinking coffee and heart attack (smoking = confounding factor, smoke> drink coffee, smoke>heart attack)

if identify confounding valuable:

gather data about it in case

do best to control its impact also include it in your research study in analysis section

- age is usually common confounding factor

How to control confounding factors?

We might not know that they even exist-know that

1. Inclusion of control group/s + matching for known

confounding factors - should be in random way (e.&, known risk or prognostic factors demographics/ prognostic factors: factors that we know it somehow related to the outcome variable that we measuring ex: smoking thing as confounder, if I know there is a relation between smoking and heart attack smoking is a potential prognostic factor bc it has been shown ex: if age can influence try to keep average age same in two groups so the difference is not bc of age (matching)

2. Randomization (during sampling stage) of subjects + large/adequate sample size for known + unknown confounding factors / control treatment

3. Inclusion of confounder in statistical analysis for known confounding factors for which there is available data

CONTROL (COMPARATIVE) GROUPS (specially in H- driven studies)

Reference point for comparison with treatment

(exposed) group/s

A good control is: (can have more than one)

similar in potential confounders (control and treatment) ('matching') > only differ in independent value

Differ in predictor variable

- They strengthen the research study!

Types of negative control

Vehicle /placebo - administration of 'vehicle' used to administer substance of interest or of inactive substance (alaki- should not see an effect. Matching on mythology ex give two groups pill, one with effect one with not - bc if we just give them one group a pill and see impact in result , the difference maybe just due to the pill and any pill can has the effect and its not bc of the gradient of the pill

Sham procedure - administration of surgical procedure only (without the actual treatment) - actual in treatment group

Non-treatment - no intervention of any kind (give them nothing)

RANDOMIZATION

It can be applied at different stages

sampling

allocation to groups

• Minimizes bias (e.g. selection bias; effects of unknown confounders)

• Different types (e.g. simple, stratified, block)

May not be always feasible (e.g. observational studies)

Sample size

'LARGE' enough

representative

affects effectiveness of randomization

costs

• Determined by # of biological replicates

• Related to power (1-B) of the study

BLINDING

• Minimizes bias (e.g. performance, detection/assessment bias)

• Different types/levels

(who) single vs. double vs. triple

(when) full vs. Partial

STUDYING CAUSATION

Not easy - good design is key!!

1. Controlling unknown + known confounding variables

e.g. randomization, matching

2. Best: interventional studies

longitudinal (prospective) = temporal separation of variables

manipulation of predictor (indep) variable

3. Inclusion of dose-response (D-R) analysis in design

CAUSALITY? (Example)

Temporality is not sufficient!

Predictor = blood type

(A, AB, B, o)

Outcome =

coronavirus infection (COVID+ or healthy)

H: coronavirus infection is

associated with blood

type

Internal validity

extent to which the observed results represent the truth IN the population studied

External Validity

Extend to which the study results apply to subjects OUTSIDE the population studied

VALIDITY OF A STUDY

Assessed based on whether: findings can be replicated

Characteristics of sample population (representative?)

Truth in study>> Generalizability >> Truth in real life

Assessed based on research design

DRAWBACKS OF EXPERIMENTAL STUDIES

1. Applicability/Generalizability of results

2. Some variables cannot be manipulated e.g. marital status

3. Ethical considerations

True or False?

True/ Scientific questions are best answered when study is planned and performed properly

Best practices for good internal and external validity:

Diminish impact of confounders and other biases

Controls

Randomization

Blinding

Adequate sample size

Decision on study design depends on:

purpose = RQ

resources, time

feasibility

RESEARCH ARTICLES (1)

Original research after

• Author(s) conducted an original

experiment or study.

• Typically contain the following sections: Methods, Results, & Discussion.

• Specific focus

REVIEW ARTICLES (1)

Summarise existing research

• Author(s) did NOT conduct original research.

• Do not contain sections such as Methods & Results

• General focus

SYSTEMATIC REVIEWS (1)

Original research (arguably)

the researchers use an organized method of locating, assembling, and evaluating a body of literature on a particular topic using a set of specific criteria, in which case the study is called systematic review

Systematic reviews typically include a description of the findings of the collection of research studies investigated

Author(s) have set out to answer a defined research question by collecting and summarising all empirical

evidence that fits eligibility criteria.

A systematic search that attempts to identify all

studies

Assessment of the validity of the findings of the

included studies (e.g. risk of bias)

Clearly stated objectives with pre-defined eligibility

criteria for studies

Explicit, reproducible methodology

Meta-analysis (1)

If the systematic review also includes a quantitative pooling and statistical analysis of the data, the study is referred to as a meta-analysis

Peer REVIEW (1)

Quality assurance process in academia

GENERAL. MULTI- DISCIPLINARY DATABASES

Google Scholar, Web of Science

DISCIPLINE-SPECIFIC DATABASES

Inspec, BIOSIS, Political Science Abstracts

Final stage (1)

Complicating the peer review process

Retracted Papers