Objectives To learn the common study designs. To learn the hierarchy of study designs. To learn the advantages and disadvantages of study designs. To be able to select a study designs.
Introduction Interested in Exposure, and/or Eg. smoking, medicine, risk factors, etc. Outcome Eg. cancer, cure, disease, etc. Exposure Outcome
Introduction Observational studies Interventional studies
Observational Studies 1. 2. 3. 4. 5. 6. Ecological studies Case report Case series Crossed sectional studies Case control Cohort study
1. Ecological Study Focuses on the comparison of groups or aggregate data, rather than individuals Quick, inexpensive Help in formulating hypothesis Disadvantages Inability to link exposure with disease (ecological fallacy) Inability to control for confounding
Ecological Study Eg. Per capita smoking in different countries and cancer rate per 100, 000 adults
2. Case Report Documenting rare / novel observation. May indicate a new link between exposure and outcome. Disadvantages It may occur by chance Single case
Case Report BMJ Case Reports 2014; doi: 10. 1136/bcr-2014 -204986 Green urine in a postoperative patient An 85 - year-old man underwent reversal of loop ileostomy. The integrity of the anastomosis was tested by an injection of methylene blue into the bowel lumen. Postoperatively the patient voided ‘green urine’ and alarmed the nursing staff and junior duty doctor. Routine microscopic examination and culture of urine revealed no abnormality. Liver function test and full blood count was also normal. This discolouration of urine resolved spontaneously in the next few days.
3. Case Series Similar with case report but observe more cases. To document potential link between an exposure and outcome. Disadvantages It may occur by chance Few cases Absence of control
Case Series According to previous example, how would you imagine a case series to be? Collection of case reports, reporting methylene blue causing green urine. Eg. 10 case reports, and compile them into a case series.
4. Cross-sectional Study Data collection on exposure and outcome are conducted at the same time. Snap shot of the studied population. Descriptive: Prevalence of cervical cancer Analytical: Association of risk factors (ie. age, SES) and presence of cervical cancer Cheap and quick. Able to study multiple risk factors and outcomes at the same time. No loss to follow-up
Cross-sectional Study Disadvantage Unable to determine causality (cause and effect) Not suitable for rare exposure Unable to establish temporal relationship (ie. alcohol consumption and depression) Exposure Outcome
5. Case Control Study Relatively cheap and quick Can investigate many exposures Suitable for rare diseases No loss to follow-up Exposure Study start at outcome Outcome
Study start at outcome Exposed Disease Non exposed Exposed Non Disease Non Exposed
Case Control Study Disadvantage Selection bias Information bias Not suitable for rare exposures
6. Cohort Study Useful for rare exposure Study multiple outcomes with single exposure Temporal relationship can be established Risk of measurement bias can be minimized Study start at exposure Exposure Outcome
Disease Exposed Non disease Follow up Study start at exposure Disease Non exposed Non disease
Cohort Study Disadvantages: Not suitable for rare diseases Expensive Time consuming Loss to follow-up bias problem
Observational Studies 1. 2. 3. 4. 5. 6. Case report Case series Ecological studies Crossed sectional studies Case control Cohort study
Qualitative Studies Exploratory, Flexible, Complex Authors and world perspective Common categories: Phenomenology Grounded theory Ethnography Case study
Qualitative Studies Common methodology: Interview Individual Focus group Semi-structured, structured Observation Individual Group Location Document analysis
Qualitative Studies How has the meaning and practice of informed consent in research changed over the last 35 years? Miller T and Boulton M (2007) ‘Changing constructions of informed consent: Qualitative research and complex social worlds’ Social Science and Medicine 65 (11) 2199 -2211. ‘Factors that help injecting drug users to access and benefit from services: A qualitative study’ Substance Abuse Treatment, Prevention and Policy 2 (31)
Introduction Observational studies Interventional studies
Interventional Studies Randomized controlled trials True experiment Randomization Confounder distribution Minimize bias Blinding Controlled
Interventional Studies Randomization (equal chance) Toss of coin Computer generated random number Ms Excel STATA SPSS IVRS
Interventional Studies Blinding Single / Double Participant / Investigator Concealment of randomization Double dummy Assessor and applicator
Cured New treatment Population Not cured Randomization Cured Standard treatment Not cured Study start at exposure outcome
Interventional Studies Disadvantages Ethical issues Expensive Time consuming Need strong justification
Hierarchy of Study Designs Systematic review and meta analysis of RCTs Cohort studies Case control studies Cross sectional studies Case studies Ideas, expert opinions, editorials Anecdotal Level of evidence
Features ECOLOGICAL EXPERIMENTAL COHORT CASE CONTROL CROSS SECTIONAL Study Type Descriptive Experimental Observational Features Aggregate Controlled E/not E known for Case chosen – Outcome and measures exposure by all participants – incident or study factor experimenter Followed up over a prevalent and then studied long time period assessed for study simultaneously factor Controls often matched Sampling None – measures Often convenient Diverse, large Prevalent or Ideally a random entire populations samples (eg. incident cases sample of the eg. countries hospital settings) Control – target population controls preferred Directionality Non-directional Always forward Backward Non or backward (incident) Non (prevalent) Used when Establishing For therapeutic or Exposure rare, Frequent Outcome and association in preventive Outcome frequent exposure, Rare exposure frequent, summary statistics assessment Retrospective for outcomes Study factor not rare disease, long changing over induction time, Incidence can be established elsewhere.
Features ECOLOGICAL Determining Not possible Causality Strength of Low as not at study individual level Problems Ecological fallacy Cost Low, if data available Information bias Major Potential Bias Controlling for confounding EXPERIMENTAL COHORT +++ ++ MOST POWERFUL Most powerful observational study CASE CONTROL + CROSS SECTIONAL Not possible Next most powerful Weakest observational study but useful for prevalence. Volunteerism, Loss of Defined by Can’t use for rare compliance participants outcome, unable to outcomes, look at multiple diseases with short effects of study durations. factor Expensive High and time Less in time and Low consuming money cf cohort Selection bias, Selection bias – Selection bias Information bias dropouts matching controls Information – Information bias – Recall bias Recall if Recall bias, poor retrospective records Randomisation – Collect data on all Matching to control Collect data on all Check it worked potential for strong potential Collect data on all confounders, confounders. potential collect data on all confounders potential confounders
How do we select? Research questions Ethical considerations Resources
Case Scenario 1 Wilm’s tumor is a kidney cancer that affects children very early in life. It is more common in West Africa than most parts of the world. The reasons for this are obscure. The fact that most affected children have the disease at birth, or soon after, suggests that the cause of the cancer may act in utero. You wish to design a study to generate hypotheses about the possible exposures in utero that could lead to Wilm’s tumor.
Case Scenario 2 Vitamin E is known as an intracellular antioxidant that might protect against cancer, but studies on animals and humans have yielded somewhat conflicting results. Select a study design to examine the effect of vitamin E on the risk of cancer.
Case Scenario 3 An epidemiology graduate student finds evidence in the literature that childhood sunlight exposure may affect adult breast cancer risk. Select a study design to explore this hypothesis
Case Study An exploratory study of the effect of mahjong on the cognitive functioning of persons with dementia Sheung-Tak Cheng, Alfred C. M. Chan and Edwin C. S. Yu INTERNATIONAL JOURNAL OF GERIATRIC PSYCHIATRY Int J Geriatr Psychiatry 2006; 21: 611– 617.
References Bonita R, Beaglehole R, Kjellström T. Basic Epidemiology 2 ed. Geneva: World Health Organization; 2006. Rothman KJ, Greenland S. Causation and causal inference in epidemiology. American journal of public health 2005; 95 Suppl 1: S 144. Gordis L. , Epidemiology 4 th ed. Philadelphia: Saunders Elsevier; 2009. Mason J (2002) Qualitative Researching (2 nd edn) London: Sage Publications.
Acknowledgment to Chew CK for preparation of the core contents of this presentation.
Dear CK, I’m just summarising some suggestions given by the other speakers regarding your topic on study design. Also, as the actual GPCR course has increased to 2. 5 days, the slot for Study design has been increased to 1. 5 hours. So kindly amend the contents for the 1. 5 hours of lecture ya. The following are the suggestions of information to be added into your powerpoint: 1. To state on randomisation method: eg: toss of coin if only 2 study groups, using sealed envelope, generate of random tables. Etc 2. To state on how to do blinding 3. To include study design for Qualitative studies, methods and pros and cons of different methods: ie: face-to-face interview, focus group etc 4. To provide a case study using a published article and further discuss or critique on the study design. The dateline for submission of new slides is by 10/11/14(Monday). Also, please add in an acknowledgement slide for yourself in the powerpoint. Do send it to lai at laiwh@crc. gov. my