Conducting Research IV

Study Design

Hypothesis formation
Hypothesis testing
1. Study Design Descriptive studies Analytical studies Case reportCase serialreportsCross-sectionalstudiesEcological studiesObservational studiesExperimentalstudiesCase-controlstudiesCohort studiesProspectiveRetrospective(historical)RandomizedControlled ClinicaltrialsRandomizedControlled fieldtrialsNon-randomizedexperiments
"There are only a handful of ways to do a study properly but a thousand ways to do it wrong"
II- Analytic studies
1- Observational studies. The investigator simply observe the nature al course of the events, noting who is exposed & non exposed and who has & has not developed the outcome of interest (passive investigator). We have:
a- Case-control study. b- Cohort study.
2- Interventional studies (clinical trial). The investigator himself will allocate the exposure (active investigator).

*Cohort study.

The group(s) is defined on the basis of presence or absence of the exposure to a suspected risk factor for a disease\outcome. All subjects then followed over a period of time to assess the occurrence of that outcome.

Design of cohort studies

→
((Starting point))
Out-come
Total
What will happen??


+ve
-ve

Exposure

+ve
a
b
Exposed

-ve

c
d
Not-exposed

Total

Present
Free
N

#Strength;

Establish the temporal relationship between exposure and outcome.
Allow direct measurement of incidence.
Examine multiple effects of single exposure.
Suitable for rare exposure.
Provide information on confounders.

*Limitations;
1- Expensive and time consuming.
2- Validity of the results can be seriously affected by losses to follow up (bias).
3- Insufficient for the evaluation of rare disease.


Selection of exposed population
For Rare exposure (e.g. radiation), we need specially exposed people (as individual in certain occupation, or exposed to special event or treatment).
For common exposure (e.g. smoking), we need specially chosen people (can obtain complete information e.g. members in certain profession as doctor, teacher…etc., or student in particular college or school).

The source of exposure data can be obtained from:

* Pre-existing data. * From the study subject. * Direct physical examination.
* Direct investigation. * Direct measurement of environment.

Selection of comparison population

Internal comparison group, used when a single cohort entered the study.
External comparison group, when specially exposed cohort enters the study.
Multiple comparison groups. Very valid.

Estimation of risk ((is there an association?))

In cohort study, we can measure the incidence (absolute risk) directly. The incidence among exposed (Ie) and incidence among non-exposed (I0). Relative risk (RR) is the measurement of association in cohort study and can be calculated as follow: (RR) = (Ie)/ (I0).
If RR=1 →No association, If RR>1 → Positive association
If RR<1 →Negative (inverse association, possibly protective).
Attributable risk (AR) provides information about absolute effect of the exposure: (AR) = (Ie) - (I0). AR indicate the number of cases among the exposed that can be attributed to the exposure it self.
Attributable risk percent (AR%): (AR)/ (Ie) x 100% , estimate the proportion of the disease among the exposed that can be attributed to the exposure it self.

Ex: To determine the relationship between cigarette smoking and CHD, 8000 healthy individual with age > 45 years were enrolled in a study. 3000 of them were smoker. Within 10 years, 84 of the smoker and 87 0f non-smokers develop CHD.
1- What is the design of the study? (Cohort study).
2- Draw 2x2 table.


Out-come (CHD)

Total

+ve

-ve

Exposure

(smoking)
+ve
84
2916
3000

-ve

87
4913
5000
Total
171
7829
8000


3- Is there any relation between smoking and CHD?
We measure the association by estimating the RR = (Ie)/ (I0). Ie(CHD in smokers) =(84/3000)x 0.1 = 0.0028 per year. I0(CHD in non- smokers)=(87/5000)x 0.1 = 0.0017 per year. RR=0.0028/0.0017 = 1.6 (what does this mean?)

*Case control study.

Subject are selected on the basis of whether they do (cases) or don't (control) have a particular disease / outcome under study, the groups are then compared with respect to the proportion having a history of an exposure or characteristic of interest.
Design of case control studies.
← ((Starting point))

Out-come

Total
What happened??

+ve

-ve

Exposure

+ve
a
b
Exposed


-ve
c
d
Not-exposed

Total

Present
Free
N

#Strength;

Quick & inexpensive regarding other analytic studies.
Depend on already available data.
Suitable for rare diseases &diseases of long latency period.
Can examine multiple etiological factors for a single disease e.g. CHD with smoking, diet, physical exercise…..etc.

*Limitations;

1- Not establish the temporal relationship between exposure and outcome (both exposure and outcome have already occurred.
2- Can't compute the incidence.
3- More prone for bias compared with other analytic studies.
4- In efficient for the evaluation of rare exposure.


Definition of cases: Homogenous disease entity (clear definition of the disease).Depend on strict diagnostic criteria (e.g. MI diagnosed by chest pain more than 30 min. ECG changes and enzyme changes).

Sources of selection of cases

Hospital-based case control study.
Advantages: common, easy and inexpensive.
Limitations: only sever cases of the disease enter the hospital (selection bias).
Population based case control.
Advantages: avoid selection bias and describe the picture of the disease in the population.
Limitations: difficult, costly and not routinely done.

Selection of control

It is important to select appropriate control group. The control group should be selected to be comparable to the cases and there is no control group fit for all situations. We have the following types of control:
1- Hospital control;
Advantages: common, easy, more willing to cooperate and inexpensive.
Limitations: They are diseased.
2- General population control.
Advantages: represent healthy population.
Limitations: difficult, costly and not routinely done as it is difficult to contact with healthy individuals.

Estimation of risk ((is there an association?)).

In case control study we can't calculate the incidence (absolute risk) because we start already disease population (cases) and non disease (control) people. Hence, we use the" Odd's Ratio" (OR) which is measure using the following formula: OR= ad/bc


Ex: To determine the relationship between smoking and CHD, patients with CHD had been compared to patients from orthopedic department; the two groups were matched for age and sex, and were asked about smoking history, the following had been found:

Out-come (CHD)

Total

+ve

-ve

Exposure

(smoking)
+ve
112
176
288

-ve

88
224
312
Total
200
400
600


1- What is the design of the study? (Case control).
2- Is there any relation between smoking and CHD?
We measure the association by estimating the OR= ad/bc
OR= (112 x224) /(176 x 88) = 1.62 (What does this mean?).

II- Interventional studies (Clinical trail).

An epidemiological design that can provide data of such high quality that it closely resemble the controlled experiments done by basics science researches. Similar to cohort study, individual are included on the basis of their exposure status, but the difference is that the investigator himself will allocate the exposure (active investigator) while in cohort (passive investigator).

Design of randomize trail:

Defined population (Acceptable and eligible group)
↓
Randomization (Each individual has the same chance to be included in group 1 or 2).
Group 1: Exposure (New medical or surgical treatment, or preventive measures).
Group 2: No exposure (No or traditional treatment, or placebo)
↓
Follow up for a period → Improvement or not → "Compare"

#Strength:

The temporal relationship between exposure and outcome can be demonstrated with confidence.
Many factors and confounders under study can be controlled.
Small- moderate differences (10%-20%) which can't be demonstrated by observational studies can be demonstrated in interventional studies.
It’s the strongest and the most direct epidemiological evidence to judge a causal association.


*Limitations;
1- The major disadvantage of this study is that this study doesn’t represent the real life situation because in real life we can't allow one factor to operate while keeping other factors constant.
2- Expensive and time consuming.
3- Ethical issue: for certain factors, there is some doubt about the benefit or harm to the exposed individuals.

Types of interventional studies

1- Preventive (Primary preventive) trail: Evaluate whether the agent or procedure reduce the risk of development of the disease among those who are free of that disease
E.g.1 Polio vaccine trial ((give the vaccine for the first group and placebo for the other, then compare the rate of polio development in both groups).
E.g. 2 fluorination of drinking water in first group and no Flore in drinking water for the other, then compare the rate of dental carries in both groups.
2- Therapeutic (secondary prevention) trail: Conduct among patients with certain disease to determine the ability of an agent or procedure to diminish symptoms, prevent recurrence or decrease risk of death from that disease.
E.g.1: Coronary artery surgery and medical treatment study for patient with MI to decrease mortality rate.
E.g.2: Radical mastectomy and simple mastectomy surgery for female patient with CA breast.

Selection of study group

The experimental group should be:
Sufficient sample size.
Sufficient number of outcome or end point (use high risk group e.g. study CHD, we chose male above 40 years old).
Need to have accurate and complete follow-up information.


The experimental group should inform about the aim of the study, about the possible side effects or benefits of the agent or may have only placebo→ Acceptable group→ screening for eligibility (exclusion of unfit individuals), Causes of exclusion:
Definite history of end point under study e.g. MI.
Definite need for study treatment e.g. Aspirin for patient with Rheumatoid arthritis.
Contra indication for the study treatment e.g. Aspirin in DU.


→ So we have Acceptable and eligible group → We do "Randomization" to control known and unknown confounders.

Maintenance of Compliance.
Non-compliance is the major problem in the interventional studies which is related to the complexity and length of the follow up. This could be due to:
Development of side effect(s). Forgetting to take the treatment.
Withdrawal from the trail. The intervention become contra indicated.

Enhancement of Compliance

Inclusion of interested and high reliable group (high risk group).
Frequent contact with participants.
Provision of incentive.
Implementation of "run in" or "wash out" period before actual randomization.

Ascertainment of out come

It's important to have complete, accurate and similar method of collection of information from both study groups and also a complete follow up of study participants over the duration of the trail. Some end points require short period follow up (e.g. mortality rate after acute MI) → Better contact with all participants during the entire study period. While other require long period of follow up (risk of dying from chronic disease) and may need many years of follow up →Difficult to maintain complete ascertainment.
The ascertainment of outcome can be affected by knowing the treatment status of participants especially if the outcome is subjective (headache, nausea…etc.), this will lead to "observational bias". This observational bias can be decreased by doing blinding of the exposure status, we have:
Single blind design. The investigator alone is aware about the allocation of the intervention.
Double blind design. Neither the participants nor the investigator (responsible for the ascertainment of outcome) know to which intervention is allocated.
But blinded trails are usually complex and difficult to conduct and some time even impossible (evaluation of programs involving change in life style as exercise, cigarette smoking…etc.), so we do un blind trail.
In order to ensure that all the aspects of the program are identical except for actual treatment (experiment), the comparison group is assigned to receive "Placebo". Placebo is inert substance in distinguish from active substance.
The use of placebo will minimize the bias in the ascertainment of both subjective and side effect of treatment. If a study doesn't use placebo control, it is impossible to tell whether subjective outcome are due to actual, to extra attention participants receive or merely to their believe that the treatment will help e.g. gastric freezing study in treatment of DU.


Measurement of association
Relative risk (RR) "as in cohort" is the measurement of association between exposure and outcome of interest in treated and comparison groups.

Statistical power of clinical trail

The statistical power of clinical trial to detect a difference between treatments groups depend on: 1- Sample size; should be sufficiently large.
2- Total number of end points, and to achieve sufficient number of end points we do:
3- Selection of high risk population. E.g. study mortality rate from CHD, we select old age group.
4- Length of follow up period.
5- The difference in compliance between the groups.

Ex: An investigator wished to investigate whether giving aspirin in low doses reduces the risk of myocardial infarction (MI). The participants in this study were over 22000 healthy males' physicians who were randomly and equally assigned to receive aspirin or placebo. The study samples were followed over an average period of 60 months. The investigator found that 8 physicians in the aspirin group experienced MI attack during the course of the study comparing with 50 physicians in the placebo group. Regarding the above information answer the following questions:
Determine the study design used.
Construct 2x2 table for the results showing the exposure and the outcome.
Calculate the appropriate measures for risk (per year), and for association.

Ex: To test the hypothesis that BCG protects against leprosy, 200 cases of leprosy and 200 individuals with no leprosy (similar age, sex distribution) were examined for the BCG scar. The results were as follows: Scar +ve among cases: 50, Scar +ve among controls: 150 , The remaining in each group was scar negative.
I. What is the study question?
2. What kind of a study design is this?
3. What is the exposure?
4. What is the outcome?
5. From this data, can you conclude that BCG protects against leprosy?