Clinical

Research 1/5

Introduction to Clinical Research

David Narita, MD Angkor Hospital for Children March 29th-30th , 2011

Steps in starting clinical research Form your question Do research about your question (get all the information) Ask for advice/input Refine your question (hypothesis, aims, ideas how to conduct, shell of results table) Check excitement levels Choose a study Observational Studies Descriptive Studies Simple description of a community Based on: - Routinely available data e.g. death statistics - Data obtained in special surveys - Does not show a relationship between exposure and effect Simply describe what is observed at a given time Types of Studies Observational Descriptive Analytical - Ecological - Cross-sectional - Case-control - Cohort Experimental Analytical Studies Randomized controlled trials Analyzes the relationship between health status Field trials and other variables Community trials Ecological/Correlational Studies Used to study groups or populations (rather than individuals) Comparing populations in different countries Comparing the same population at different times Easy to carry out Limitations: - Difficult to interpret - Cannot show the link between individual exposure and development of disease - Cannot consider other factors which can contribute to development of disease - Ecological fallacy (incorrect belief or idea) - Cannot show link between individual exposure and development of disease Cross-sectional/Prevalence Studies Measure the prevalence of disease Exposure and disease are measured at the same time Cannot determine if exposure preceded or followed development of disease Cheap and easy Regular cross-sectional surveys can be used to assess the health care needs of a population Cross-Sectional Study Survey Survey should have a clear purpose and needs to: - Be well-designed - Have an appropriate sample of sufficient size - Have a good response rate

Clinical Research 2/5

Case-Control Studies Compares people with a disease and people without a disease to identify the exposure Start by selecting (choosing) cases - the sample should represent all the cases from a specified population Select controls from the same population Cases and controls should be selected independent of exposure status Case-Control Study Case-control studies are: - Cheap and easy - Used to investigate causes of disease - Best for rare diseases - Longitudinal (Data are collected from more than one point in time) - Usually retrospective (Disease status is determined first, then exposure) Determining exposure status - Ask the patient or caregiver (retrospective data) - Biochemical measurements - Established recording system (e.g. patient chart or employment data) The odds ratio (OR) Measures the association between an exposure Disease Punnet Square and a disease Yes No Total Ratio of the odds of exposure among cases compared to controls Yes a b a+b If OR = 1, then the exposure is not associated Exposure No c d c+d with the disease If OR > 1, the exposure is associated with Total a+c b+d a+b+c+d developing the disease - x times more likely to develop the disease than those not exposed If OR < 1, the exposure is associated with NOT developing the disease (may be protective) Cohort studies Also called follow-up studies Longitudinal Can be retrospective or prospective Gives the best information about disease causation Can take a long time Sometimes need large study population Expensive

Cohort Study

Clinical Research 3/5

Experimental Studies Test the effectiveness of an intervention - e.g. new drug, medical treatment, behavioral intervention, etc. Randomized Controlled Trials (RCT) Randomized Frequently used to test the Control Trial effectiveness of new drugs or medical treatments Participants are randomly assigned to either an intervention (treatment) or a control (no treatment or placebo) group - Blind: participants do not know which group they are in but researchers do - Double blind: both participants and researchers do not know which group they are in Field Trials Involve people who are disease-free and non-hospitalized but thought to be at risk Sometimes involves a lot of participants Useful for evaluating interventions that reduce exposure (eg: Sabin oral polio vaccine) Community Trials Treatment groups are communities, not individuals (eg: villages randomly assigned to intervention or control group) Used for diseases which are influenced by social conditions Can be difficult to control confounding factors Most commonly used for health promotion interventions to change behavior eg: Condom promotion to reduce HIV transmission, education to encourage healthy eating Choosing a Study

Errors in Research Random error The sample is different from the population because of chance 3 main sources of error - Individual biological variation: Random variation that naturally occurs

Clinical Research 4/5

- Sampling error: Random errors in choosing participants (usually because we dont have a large enough sample) - Measurement error: Random errors in measuring the exposure or outcome need clear protocols Sample size The easiest way to reduce random error is to increase the sample size (the number of people in the study) Epidemiologists use formula to decide on sample size based on: - The necessary level of statistical significance - How common the disease is in the community - The size of the groups to be compared - The chance of missing a real event Systematic Error Results do not show the true situation because of an error (bias) in data collection 2 main types: - Selection bias - Measurement bias Selection Bias Occurs because there is a difference between the people who agree to participate in a study and those who do not, and between those who remain in a study and those who leave early Examples: People who are sick are more likely to volunteer to participate Heavy smokers are less likely to respond to a survey on the health effects of smoking People who are very sick may not need to withdraw from the study Measurement Bias Happens when they do not measure correctly what they are supposed to measure ie individual measurements or classifications of disease or exposure are wrong If it occurs equally in both groups being compared, it results in an underestimate of the true strength of the relationship Recall Bias People who have a disease are more likely to remember the exposure because they already know a relationship exists Bias can exaggerate (make bigger) the effect of the exposure e.g. heart patients more likely to admit history of little exercise Or bias can underestimate the effect of the exposure e.g. mothers less likely to admit to smoking or drinking during pregnancy Observer Bias Investigators ask questions or interpret the information for cases and controls differently Confounding Occurs when there is a second exposure in the study population that is associated with both the exposure and the outcome Important because: - It can change the direction of an association e.g. something that appears to protect a person from disease may actually cause the disease - It can make people think something causes a disease when it does not Is not a problem if the confounder occurs with the same frequency in both groups The most common confounders are sex and age

Clinical Research 5/5

Controlling confounding Design stage Randomization: Randomly assigning participants to a group should distribute confounders equally (if the sample size is large enough) Restriction: Limit the study to include people who have certain characteristics (eg: non- smokers) Matching: Participants in the 2 groups are matched for confounding variables (eg: age and sex) Analysis stage Stratification: Groups are analysed in categories (strata) of the confounding variable (eg: measure an association within age groups or by sex) Statistical modeling: Statistics can be used to measure the associations between many variables and identify variables that are confounders Validity & Reliability Validity: whether the results of a study show the true situation; whether the test measures what it is supposed to measure - Internal validity: How correct the results are for the population under study - External validity (generalizability): How much the results of a study apply to the general population - A study can have high internal validity but low external validity eg: if the sample population does not represent the general population Reliability: whether the results are repeatable Ethics in Epidemiology Informed Consent - Patients have the right to withdraw from the trial at any time - Researchers have an obligation to the study participants and others whose health may be affected by the results - They must tell communities what they are doing and why - They must inform communities of their results and their significance (what they mean) - All proposals for studies must be submitted to an institutional ethics committee before work begins Confidentiality - Researchers must respect personal privacy and confidentiality at all times - Researchers need to obtain permission before accessing patient records and other information Respect for Human Rights - Tension can arise between the interests of a group and the interests of the individual - Need to assure people that their interests will be protected Scientific Integrity - All scientists have the potential to behave in an unethical manner, perhaps in part because of the pressure to succeed.