(5) designing a study ii (cont.)

Upload: jehad-ali-al-ammari

Post on 04-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 (5) Designing a Study II (Cont.)

    1/8

    Types of observational studies

    1. Cross-sectional studies

    2. Prospective cohort studies

    3. Case-control studies

    4. Nested case-control studies

    5. Ecologic studies

    Last time we discussed the first two studies, we said the cross sectional

    study is done at a single point of time, it takes a short period of time, and

    its less expensive.

    The other type is called prospective cohort study, prospective means that it

    has something to do with the future,so you select your sample and youfollow up the sample for a long period of time.

    Slide (19-23): case control study

    Subjects are assembled based on whether they have experienced the

    o utcome (cases) or not (controls).

    In case control studies you take for example patients who have cancer we call them

    case, and also you select other normal people we call them controls, then you observe

    certain things among these cases and also among these controls.

    Frequencies of risk factors are compared between cases and controls .

    You select 50 patients who have lung cancer and then you select 50 normal people, and

    then you will ask these patients questions about risk factors like smoking, if they

    smoke or not, and then you ask the controls normal people if they smoke or not, so you

    will study the frequency of the risk factors or the effects of the risk factors which is

    smoking on people with lung cancer and on people without lung cancer. So this called

    case control study, why Case control? Because one patient has the case and theother one is normal control.

    1

  • 7/30/2019 (5) Designing a Study II (Cont.)

    2/8

    Advantages :

    Is Efficient, especially for studying uncommon diseases.

    Cases and controls must originate from the same population .

    You cant take the people with cancer among the Jordanian and then the

    controls from the Chinese people that work in madeenet 2l 7asan!!!

    Because they are from different populations.

    Disadvantages:

    Cannot be used to determine prevalence and incidence.

    Because they already have the disease, if you want to see theincidence of cancer, you have to select healthy people and study

    them over a long period of time and record new cases of cancer.

    Selection bias: loss of cases/controls prior to their selection (a

    case died prior to assembly of cases, then the sample wouldnt

    be representative)

    You want to study people with cancer, and you prepared a research for fifty

    cases with cancer, but at the selecting stage, you may discover that some

    patients died, so by this you lose some of the cases, and this lead to selection

    bias, you need now to select new people, which takes time.

    Recall bias

    Cases are more likely to remember exposures than

    controls,

    Psychologically, people with cancer (or any other disease) are different

    from normal people, all the time they remember their exposure to cancer,

    and that may affect their participation.

    E.g., cases with cancer may report previous exposures

    because they have been more aware about their health

    and subjected to many previous tests.

    2

  • 7/30/2019 (5) Designing a Study II (Cont.)

    3/8

    People with cancer are very cautious about their health, and thats why

    they are very aware, thats why they will provide more details about any

    previous problem related to cancer that happened with them, but normal

    people because they are healthy, they are not caring about the details of

    things that happened to them, thats why these two groups are different,

    and thats called recall bias, Because people with cancer will remember

    more details to provide you in the research more than the normal people

    who havent been experienced with cancer.

    Can be matched and unmatched

    Matching (two types):

    1. Individual matching

    Each case is individually matched with one or more controls.

    For example a 45 yrs old man as case is matched with a 45 yrs old

    man as a control.

    2. Frequency matching

    i. Controls are matched to cases as a group.

    We are not matching at individual levels; here we matchone group with the other group

    ii. Similar distribution of cases and controls on each

    matched variable is established.

    iii. E.g., males with range of 20-40 yrs account for 30% in

    both groups

    If you want in the case group to select ages from twenty

    to forty, and this age range account for thirty percent of

    the number of cases, you have also to select the same

    percentage of people from twenty to forty years in the

    control group.

    Advantage of matching

    You eliminate confounding; we already discussed what confounding

    means.

    Disadvantage of matching

    Increasing difficulty and cost of identifying control, specially with

    limited number of potential controls.

    3

  • 7/30/2019 (5) Designing a Study II (Cont.)

    4/8

    its not easy to find two persons with the same circumstances in the

    same population, its difficult and can be costly, and in control group,

    people are not interested to test a drug on them, we can select fifty

    people with cancer and apply a drug, or a medication on them, butthe control group is not easy to select because they will not accept to

    try medications on them thats why they have to provide some

    source of motivation to them, like paying him with high cost, thats

    why it is difficult and expensive.

    Matching for a variable will not enable to study its impact on the

    outcome

    We want to study the effect of diesel fumes in the development of lung

    cancer, so what we do? we have to match for smoking, because smoking

    also leads to lung cancer, thats why when you select the cases, you haveeither to select cases who are all smokers, or cases who are all non

    smokers, you cannot haphazardly go and select people without knowing

    that these people are smokers or not.

    Best to avoid matching except in small studies where it is difficult to

    adjust statistically for all possible confounders unless if matching is

    used.

    in general although matching is good and proficient, it is difficult and

    expensive, so its better to avoid matching, but we do matching

    when we have a very small study, which is very easy to handle, like if

    you want to study only five cases of cancer.

    How many controls per case to enroll?

    Greatest efficiency with equal number, like fifty cases and fifty

    controls.

    Adding additional controls, we can make the control group bigger

    than the case group when :

    4

  • 7/30/2019 (5) Designing a Study II (Cont.)

    5/8

    Enough cases cannot be obtained such as in rare conditions to

    increase the power of the study.

    If we want to study a very rare condition, like a very rare cancer, and

    we only have five cases; its not wise to bring only five controls, so

    we have to increase the control group.

    There is more than one variable or confounder to match on.

    If we have more than one variable that can be acting as confounders,

    like the effect of smoking on people who breathe diesel fume in the

    lung cancer. Smoking is a variable you have to be aware of, thats

    why in these cases we have more than one variable. Sometimes the

    age is an important variable you have to select people with the sameage in this case you can select more people in the control than in the

    case group.

    Maximum: 4 controls per case, if you have a case group with twenty

    people dont make a control group with more than eighty people.

    Slide (24, 25)

    Self reading: not included

    Slide (26): Specifying a hypothesis

    What are you hoping to prove before data collection? Whats your aim in

    the research? This is the hypothesis.

    Hypothesis, can take two forms:

    Null form

    There is no difference; or there is no relationship, for example I want

    to study the effect of smoking on cancer, I put the null form of my

    hypothesis: there is no relationship of smoking on lung cancer, andduring the study I try to reject the null hypothesis, and prove that

    smoking causes lung cancer, which is the alternative form.

    5

  • 7/30/2019 (5) Designing a Study II (Cont.)

    6/8

    Alternative form

    There is a difference; or there is a relationship

    Study hypothesis is stated in both the null and alternative forms, when

    we put the hypothesis we put it in two forms, for example I want to

    study the effect of smoking on lung cancer, I will have two hypothesis,

    the first hypothesis; smoking is not related to cancer this is the null

    hypothesis, then smoking is related to lung cancer this Is the

    alternative hypothesis, and in my study I want either to reject the null

    hypothesis or to confirm the alternative hypothesis (the same meaning),

    sometimes the results will be deviated to the null form, and the null

    hypothesis become confirmed rather than rejected, and the result will

    be there is no relation between smoking and lung cancer in this casethe null hypothesis is not rejected, and you get results that you are not

    wishing to, or you are not hoping to.

    Statistical analysis is based oninferential reasoning

    Slide (27): Inferential reasoning

    Assessing the probability that an association found in a sample could haveoccurred bychance if there were no true association in the population.

    If the amount of chance in your sample is large and not reduced, some of

    your outcomes or findings may be based on chance. In this case - when an

    association found by a sample could have occurred by chance- there will be

    no true association in the population and your sample will not represent

    the population.

    Thats why your role is to reduce the amount of chance, so that your

    sample represent your population, you dont want the effect of chance tolet you confirm your results when your results cannot be applied on the

    population.

    If the probability that the association could have occurred by chance falls

    below five percent, null hypothesis is rejected and alternative hypothesis is

    accepted.

    We have a small range of probability of chance, this chance shouldnt be

    over five percent, in the study you cannot eliminate chance, because the

    sample is not hundred percent representative to the population, but youhave to reduce the chance, I want the chance only to have less than five

    percent effect, so thats why I put probability less than five percent, this is

    6

  • 7/30/2019 (5) Designing a Study II (Cont.)

    7/8

    the maximum amount of chance that I can allow in my study otherwise if it

    exceeds this percent this means that the outcome is not valid.

    We all know that smoking causes lung cancer, but possibly -by chance- I examined a

    small group of people in my study and I proved that smoking does not cause lung

    cancer, but this is wrong, because it happened by chance and does not represent the

    population.

    Slide (28): A study with more than one question

    Collecting data on more than one outcome

    Collecting data on additional risk factors for the same outcome

    does not answer multiple questions

    If you want to answer more than one question you should have multipleoutcomes, the effect of diesel fume on lung cancer has just one outcome

    but smoking may affect the lung cancer, gingival health, and heart disease.

    Multiple outcomes

    Different stages of the same disease process

    Smoking on angina, MI and death

    Different disease processes influenced by the same risk factor

    Smoking on gingival health, lung cancer and heart disease

    Outcomes unrelated to one another

    The effect of Smoking on health and cost.

    Slide (29): Sample size

    Depends on the statistical test used

    Discussed later

    Slide (30): Ethical approval

    Its related or applied to the Human research committees; I cannot do any

    research that I want just like that!!!

    7

  • 7/30/2019 (5) Designing a Study II (Cont.)

    8/8

    3. Protect the rights of subjects.

    4. Insures that the subjects fully informed.

    5. Insures that subjects have consented to participate, like signing a

    contract (mwafaqa 5a6eyye)

    6. Insures that the risks are reasonable and much less than the new

    knowledge or benefit that the study will provide. (fawa2ed lazem

    tkuun aktar men 2l 2draar)

    7. Insures that confidentiality is maintained, you shouldnt relies the

    data you collected, like pictures and radiographs you used in the

    research.

    One of our colleges in just wanted to make a research on the beauty, and howattractive the face is, he wanted to bring nice females and to have their faces

    photographed and then studying it, in our conservative society this is unethical, and

    it was rejected.

    Thank you LAZGA for helping me SO MUCH in doing this lecture :S!!!(4 minutes!!!)

    As we all know from dr.ashraf, most of the question will be from the slides, but to

    understand the slides we have to refer to the lectures, the worst thing in studying the

    lectures is that there are no pictures AT ALL! :S

    Why do we always read the regards before we start the lecture!!! O.o ???

    Good luck in the coming exams

    DONE BY:

    AMMAR ANAGREH

    8

    THE END