lecture 6 Research design II part two (script)

8/3/2019 lecture 6 Research design II part two (script)

1/12


2/12

Before we start this lecture there is announcement about attendance and absences thosewho is exceed the limit of absence (10%) will be deprived and the dr. allowed theabsence of two lectures and he will put lists of absence during this week.

.

Now the lectureSlide2We have different types of randomized control trial or randomized study ordesign, we discuss each of these:

1-simple randomized design2-cross over design

3-factorial study or design

Note: slides from (7-12) are not included and they are self readingWe will start from slide13 (type of observational studies--:(

Dr said it is very important to be able to distinguish between different types ofrandomized study and different types of observational study so in the exam dr.will bring questions and this questions are examples and you need to understandwhat this present and what the type of design this present thats why this coarseis very much based on understanding not on memorizing

*Types of observational studies-:1-Cross-sectional studies

-Prospective cohort studies23-Case-control studies

Nested case-control studies-4Ecologic studies -5

#last two types will not be discussed and will be left as self reading

Slide14(cross sectional studies(

What dose cross section mean???You examine the subject as one single time so when you in the lab you docross sectional of specimen so what do you do here you cross section

time) )

1.It is easy and fast

2.Information is collected from subjects at a single point of time or atsingle cross section time

2


3/12

So you examine the patient once and you dont modify the way they liveor you dont give them any thing like drug or medication so to studythem you just cross section their time and examine them this is calledcross section of study ( )

3-Used to answer descriptive questionsFor example: - you want to see what the prevalence of a disease is? Or

what is the frequency of certain disease?So if you want to study this you have to bring a group of people once atsingle period of time, for example you want to study the frequency ofTaurodontism it is when the pulp chamber is elongated so the tooth hasvery elongated pulp chamber so you want to study the prevalence ofTaurodontism among group of people so what you do? At single point of

time you take ex-ray for all these people, each one will have an ex-ray forhis teeth and then you study those teeth with Taurodontism so by this wecalled this is cross section of study


4/12

5-Useful in estimating sample size (will be discussed later(

6-Not good in answering analytical questionsSo the only answer this scripted question so it's difficult to makeanalytical questions

*An association found may go in either direction, for example sometimesthe association between two factors goes two ways e.g.: alcohol causesdepression but people who are depressed also they may consume alcohol

*Risk factor may cause the outcome or vice versaSo this type of study (cross section) is not useful to study the relationship

between two factors when the effect and cause relationship is not clear orthe direction of effect cause is not identified e.g.: alcohol and depressionBut sometimes in certain studies it is very impossible for association togo the other way e.g.: smoking & facial wrinkles

So it is clear that smoking cause facial wrinkles but facial wrinkles cannotcause people to smoke

Slide 16 (Prospective cohort studies(What does cohort mean? Is a group of people with certain characteristic for example you want to study dental students in the second yearthey have certain characteristics for example age 19 years and so

they're named ( cohort(What dose prospective mean? You study a group of people foralong period of time not for across section of time

1-Sample is assembled prior to development of the outcome and followedover timeSo all people who are going to be included in your study should be free ofcondition that you are studying( (

for example you study the incidence of breast cancer all of women shouldnot have breast cancer and you will study this group for 20 year and you

will record the number 50 women who developed breast cancer amongthese 50 women there were 5 women who developed breast cancer forexample we say 5:50 or 10%have breast cancer

%510 (20 50(

2-Subjects are evaluated to make sure that they do not already havethe outcome being studiedSo these people should be free of the disease or outcome because

sometimes we study something does not necessary a disease

4


5/12

For example I want to study about fraud there is group of people Iask them (are you a cheater) so I have 300 students and thus theprevalence of cheaters for example is 90% or 250:300 but theincidence is different I have 100 students and I study them for 5 yearand they shouldn't become cheaters and thus the outcome is notnecessary that the outcome is a disease sometimes it is something

else such as cheating, a social study

3-Provide much stronger evidence in support of a causal relationshipBecause you study people over long period of time you can notidentify the cause and the effect, for example in the Dr.s study thetiming of tooth eruption he couldn't study the effect of prematureextraction of deciduous teeth because his study was cross sectionwhich was difficult (will divorce affect the eruption or not? forexample(

So it is difficult to study cause and effect but if he study a group of

people over ten years to see the timing of tooth eruption it is veryeasy to know that this person has a number of teeth lost so he cannext time detect the permanent will that emerge faster or slower

)(

Thats why I can study the occlusal relationship in much stronger wayin prospective cohort study

4-Reduce the possibility of reverse causality because it is always thecause to go in one direction

(( 5-Minimizing recall bias

What dose recall bias mean? It is bias or error in research becauseof there is a recall to something that happened to the patient but atthe beginning the patient didnt remember thats happened to him orwhen a person remembers a pervious exposure to a disease( (

For example, I have a number of people who are in my study; I willask them (have you had colic pain like appendicitis?) if there a

person 60 yrs old he will not remember but if the injury is painfulduring this research this is called recall bias)

60(

*Information about risk factor is collected ahead of diseasedevelopment

*Recall bias is a problem with case control studies developingdisease make subject remember an exposureSo you make a study and there are people having a disease and

others dont have disease and I study them over a long period of timeand the people who didn't have the disease when they felt pain this

5


6/12

pain isnt the first time but at the beginning they didnt remember sothey dont give accurate information

Slide 17 (Prospective cohort studies(1-Can be used to calculate incidence rate not prevalence

2-Incidence rate is the number of new cases of a particular condition in anat-risk population per unit timeSo you have to identify the period of time not cross section of time andyou have to study these over a long period of timeLongitudinal study3-

)Means over a long period of time )

4-Length of follow up time is based on how long it takes to develop thediseaseSo you have to select the suitable time to allow disease to develop forexample you want to study the incidence of breast cancer it isnt good tostudy women over 6 months because in this short period of time, you maynot have results but you have to study women over 20 yrs by this you willensure that the women have come in the stage of development of breastcancer of course some of them

Remember: in prospective cohort the participants should be free ofthe outcome at the beginning of the study and all of themshould have similar chance or the same possibility of

developing the disease or outcome ( )Slide 18

*Disadvantages: (of Prospective cohort studies(1-Take a long time to perform esp., when the disease develops slowly

For example, breast cancer develops slowly thats why women should bestudied over a long period of time

) (cross section 2-Costly and inefficient for studying uncommon diseases (fewer personswill develop the disease(

It isnt good in uncommon or rare disease because may be the diseasedoesnt happen

3-Bias due to loss of subjects to follow upBecause this study take a long period of time, people will give up sowhen we study something longitudinally some of your subjects may drop

this study, thats why loss of subjects is a possibility and it is a draw back

6


7/12

( )

4-Period of temporal changes may influence resultsTemporal is time so temporal changes will impact on outcome) )

Introduction of new instruments**Change in clinical practice

For example, acrylic fillings and now we have amalgam and compositefillings but in the 50th century they used acrylic fillings if someone madea study on acrylic over a long period of time this isnt good because theresult is no longer important and this filling will be old fashionOther example about the implant it is important now rather than dentureand if someone made a study on dentures so there sometimes a change inclinical practice

And introduction of new instrument( (

5-Answer of research question may become less relevant when the studyis completeSo the studies are always renewed such as study of dentistryFor example, when the dr. was a student they told them that the theory oftooth eruption is the attachment and reattachment of periodontal ligamentnow last year dr. told us that the most acceptable theory is contractility ofthe hyproplast and hydroststic pressure

(20 ( Slide 19 (case-control studies(

*Subjects are assembled based on whether they have experienced theoutcome (cases) or not (controls(

So you group people in two groups based on whether they have theoutcome or not

So one of them have outcome (case) and other doesnt have outcome(control(

Not like prospective cohort study that should be free of conditionsFor example, I choose people with gingivitis (cases) and I choose peoplewithout gingivitis (control(

And then I study them

*Frequencies of risk factors are compared between cases and controls

Slide 20 *advantage: (ofCase-control studies(7


8/12

*Efficient especially for studying uncommon diseasesSo if you have uncommon disease the best way is to do the case here ifyou're in case-control studies >> I choose 5 people with disease (cases)and 5 people without disease (control(

So they should belong to same population or the same range or same race*Disadvantages:

Cannot be used to determine prevalence and incidence1- Because the cases have disease already so what I shall study!?!2-Selection bias: loss of cases/controls prior to their selection (a case died

prior to assembly of cases, then the sample wouldnt be representative(So sometimes when you study there are loss of cases or control andsometimes this loss can be prior to their selection such as cases died priorto assembling sample so sample wouldnt be representative) (

3-Recall bias*Cases are more likely to remember exposures than controls

( )For example, we have cases with cancer may report previous exposures

because they have been more aware about their health and subjected tomany previous tests

)5 5 , ( Slide 21Case-control studies >>> *Can be matched and unmatchedMatching*

1-Individual matching*Each case is individually matched with one or more controls

So the person with a disease should be matched with a normal personboth of them should be the same age and same population, for example:45yrs old man as case matched with 45 yrs old man as a control (

)

Frequency matching2-*Controls are matched to cases as a group

*Similar distribution of cases and controls on each matched variableE.g., males with range of 20-40 yrs account for 30% in both groups

)20-40 30( %

Slide 22

*Advantage of matching:-Eliminate confounding

8


9/12

The cases and control should be the same population and same age andsame exposure to risk factor so this is eliminating confounding

)prostate cancer )

Disadvantage of matching*

1-Increasing difficulty and cost of identifying controls esp., with limitednumber of potential controlsSo sometimes it is very difficult to find a match of certain peopleE.g., 45 yrs old male with prostate cancer

(45(

2-Matching for a variable will not enable to study its impact on theoutcome

E.g., matching for smoking to study the effect of diesel fumes on lungcancer )

disadvantage(

*Best to avoid matching except in small studies where it is difficult toadjust statistically for all possible confounders unless if matching is

used

So I always avoid the matching unless there was statistically I willuse the matching because it is necessary

Slide 23 /case-control cont\"*How many controls per case to enroll

The best is for one case to select one control

*Greatest efficiency with equal numberBut sometimes the disease is very rare and only I have 3 cases ofthis disease so may be it isnt wise to select 3 controls so I choosefor each case a number of controls

*enough cases cannot be obtained such as in rare conditions increase the power of the studySo I choose for each case 3 controls, and thus I increase thecapacity of study

*there are more than one variable/confounder to match on

9


10/12

So if you have more than one variable, you should choose 3 casesand 3 controls and each one have to study a particular variable

Maximum: 4 controls per case*

#Slide 24 & 25 is self reading and not included

Slide 26 (Specifying a hypothesis(

*What are you hoping to prove before data collection(what are ugoing to do(?

So before you start, you have to put hypothesis for your researchdesignFor example, the study is prevalence of caries among dental

students so the hypothesis will be (what is the prevalence ofcaries in dental students(?This is the research question is called hypothesis))

*it has two forms:1-null form

There is no difference or no association or no correlation, sowhen you study the effect of smoking on lung cancer you say(smoking is independent of lung cancer or doesnt cause the lungcancer?) this is called hypothesis

2-Alternative formThere is a difference

There is effect smoking on lung cancer

#you always put hypothesis in two forms (null & alternative) allthe time

*Study hypothesis is stated in both the null and alternative forms

-Statistical analysis is based on inferential reasoningSo for example, you want to study effect of smoking on lungcancer you put hypothesis in null form (smoking is independentof lung(

But in the end of the study you find that smoking cause lungcancer by 90%In this case you reject the null hypothesis and accept alternativeform

Slide 27 (Inferential reasoning(

10


11/12

*Assessing the probability that an association found in a samplecould have occurred by chance if there were no true associationin the population

*If the probability that the association could have occurred bychance falls below (p


12/12

Waad,Ruba,Sara,Dalia,Alaa,Amna:Each of us has his way in life, but wherever we go every otherpart of the bears

12

Documents

lecture 6 Research design II part two (script)