Case Series, Descriptive, and Cross-Sectional Studies

Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

Population of patients with condition of interest Sample

Sample Case Series, Descriptive, Group 1 Group 2 and Cross-Sectional Studies Bias External Validity Internal Validity Pawin Numthavaj, M.D. Section for Clinical Epidemiology & Biostatistics Chance Faculty of Medicine Ramathibodi Hospital Mahidol University August 2014 Conclusion

Hierarchy of evidence Natural Clinical Research Assign Exposure Exposure

Observational Studies Experimental Studies No With comparator comparator Systematic Review & Meta-analysis Descriptive Analytic Randomized Controlled Trials Randomized Controlled Trials Case Study Cross-Sectional Nonrandomized Non-randomized Intervention Studies Controlled Trials Case Series Cohort Observational Studies

Cross-Sectional Case-Control Non-experimental Studies Hybrid Studies (Nested CC, Case-Cohort) Expert Opinion BMJ 2001;323:334.1

Descriptive Studies

• Concerned about disease burden • Attempt to answer question • Who? • What? • Where? • When? • “First ideas” about causality and generate Case report and Case series hypothesis for further studies

1 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

Case report and Case series Example

• Detailed description of one or more cases of a • Description of series of infants born with disease that are unusual for some reason congenital cataracts and cardiac abnormalities • Never seen before in Australia (Gregg 1941) • Occur in unexpected individuals • Severe epidemic of rubella 6-9 mo. before children • Occur in unexpected places born • Now: we know that rubella affect babies born from infected mother

Propranolol vs. Infantile Hemangioma

• Identify potential health problems in outbreaks: • Léauté-Labrèze 2008. Case report of successful SARS, bird flu, swine flu treatment of a child with infantile hemangioma with obstructive cardiomyopathy with propranolol

Léauté-Labrèze C et al. N Engl J Med 2008;358:2649-2651.

Propranolol vs. Infantile Hemangioma

• 2008: First case report • Almost everyone still use Steroids • Multiple case reports follow • 2011: First RCT • Positive result • Multiple RCTs follow • 2013: Meta-analysis Cross-sectional Studies • Nowadays • Almost everyone now try propranolol first

2 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

now time Principle of X-Sectional Studies Exposure Outcome

Exposure Outcome Cohort • Conducted at “single point” in time • (Or a relatively short period) • “Snapshot” of population Exposure Outcome • Exposure and Outcome measured at one point in time or over a period* Exposure Outcome Case Control • Often in the same time • Can be descriptive or analytic Outcome Outcome Outcome • Depend on design Cross • Prevalence study (descriptive) Sectional • Comparison of prevalence among exposed and non Exposure Exposure Exposure exposed (analytic)

Population Population

Sample Sample Describe

Male 53% Mean age 45.30 yr Analyze Smoke 30% Mean SBP 143 mmHg + Exposure , – Outcome One time SD SBP 12 mmHg One time + Exposure , + Outcome Measurement Mean DBP 84 mmHg Measurement – Exposure, – Outcome SD SBP 11 mmHg – Exposure, + Outcome

Example Snapshots of Disease

• Prevalence of disease • Prevalence of Hand-Foot-Mouth disease in Bangkok • Morbidity Survey • Prevalence of post anesthetic spinal headache • Distribution • Mean and SD of length of descending branch of lateral circumflex femoral artery in Thai people

Prevalence of malaria 1995 1996 1997

3 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

Descriptive Cross-Sectional Studies Prevalence vs. Incidence

• What they can do • Prevalence • Trend analysis (forecasting) • Fraction of a group of people possessing a clinical • Planning condition/outcome at given point in time • Clue about cause (generate hypothesis) • Incidence • What they CANNOT do • Fraction of group of people initially free of outcome • Conclusion about cause of disease but develops condition over a given period of time • Over- or misinterpretation of data

Problem about descriptive data

• Vitamin C reduce URI symptom 70% • Placebo reduce URI symptom 60%

• Which one should we use?

Descriptive vs. Analytic Analytic Cross-Sectional Studies

• Prevalence • Measurement of association Descriptive Analytic • Prevalence ratio • Prevalence odds ratio • Diagnostic studies • Sensitivity •Describe •Explain • Specificity • Predictive values • Accuracy

4 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

Analytic Cross-Sectional Studies 2x2 Table

Disease No Disease Population (Outcome+) (Outcome−) Exposure+ Outcome+ Sample Exposure+ Risk factor Outcome− (Exposure+)

Exposure− Outcome+ One time Measurement Exposure− No Risk factor Outcome− (Exposure−)

Disease (O+)No Disease(O−) O+ O− E+ A B A+B

Risk factor A+B (E+) A B E− C D C+D

A+C B+D A+B+C+D No Risk factor C+D A+C (E−) C D Prevalence (of disease) = A+B+C+D

A+C B+D A+B+C+D

Disease No Disease

Measurement of Association Risk factor A+B A B • Prevalence Ratio (PR) No • Prevalence Odds Ratio (POR) Risk factor C+D C D

A+C B+D A+B+C+D

Prevalence of disease A among exposured (E+) = A+B

Prevalence of disease C among unexposured (E-) = C+D

5 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

1. Prevalence Ratio 2. Prevalence Odds Ratio

Odds of Exposure among Cases 퐸푥푝표푠푒푑 푐푎푠푒푠 푈푛푒푥푝표푠푒푑 푐푎푠푒푠 푃푟푒푣푎푙푒푛푐푒 표푓 푑푖푠푒푎푠푒 푎푚표푛푔 푒푥푝표푠푢푟푒 = ÷ = 퐴푙푙 푐푎푠푒푠 퐴푙푙 푐푎푠푒푠 푃푟푒푣푎푙푒푛푐푒 표푓 푑푖푠푒푎푠푒 푎푚표푛푔 푛표푛 푒푥푝표푠푢푟푒 퐴 퐶 = ÷ 퐴 퐶 퐴 + 퐶 퐴 + 퐶 = ÷ = 퐴 ÷ 퐶 퐴 + 퐵 퐶 + 퐷 Odds of Exposure among Non-cases 퐸푥푝표푠푒푑 푛표푛푐푎푠푒푠 푈푛푒푥푝표푠푒푑 푛표푛푐푎푠푒푠 = ÷ 퐴푙푙 푛표푛푐푎푠푒푠 퐴푙푙 푛표푛푐푎푠푒푠 퐵 퐷 = ÷ 퐵 + 퐷 퐵 + 퐷 = 퐵 ÷ 퐷 푂푑푑푠 표푓 푒푥푝표푠푢푟푒 푎푚표푛푔 푐푎푠푒푠 Prevalence Odds Ratio = 푂푑푑푠 표푓 푒푥푝표푠푢푟푒 푎푚표푛푔 푛표푛−푐푎푠푒푠 퐴 퐵 퐴 × 퐷 = ÷ = 퐶 퐷 퐵 × 퐶

No OA Example: OA knee and Obesity OA Knee Knee Total Obesity 80 20 100 No 40 60 100 Obesity No OA 120 80 200 OA Knee Knee Total Prevalence of OA knee 120 / 200 = 0.6 Obesity 80 20 100 Prevalence of OA knee among obese 80 / 100 = 0.8 subjects No Prevalence of OA knee among non-obese 40 / 100 = 0.4 40 60 100 subjects Obesity Prevalence Ratio 0.8 / 0.4 = 2.0

Interpretration: The probability of OA is 2 times higher for obese subjects than 120 80 200 non-obese subjects. OR the probability of OA is 100% higher for obese subjects than non-obese subjects.

Prevalence odds ratio Odds ratio

• The odds is the ratio of the probability that the • Probability of winning = 60% event of interest occurs to the probability that it • Odds of winning = ? does not. • This is often estimated by the ratio of the number of times that the event of interest occurs to the number of times that it does not

6 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

Odds ratio

• Probability of winning = 60% • Odds of winning = 60% : 40%

= P : 1-P = 0.6 : 1- 0.6 = 0.6 : 0.4 = 1.5

No OA Odds? OA Knee Knee Total Obesity 80 20 100 No 40 60 100 Obesity 120 80 200

Prevalence of OA knee 120 / 200 = 0.6 Prevalence of OA knee among obese 80 / 100 = 0.8 subjects Prevalence of OA knee among non-obese 40 / 100 = 0.4 subjects Prevalence Ratio 0.8 / 0.4 = 2.0 Prevalence Odds Ratio 80:20 / 40:60 Probability of dying = 50% 80x60 / 20x40 = 6.0 Probability of living on = 50% Odds of dying = 50%:50% = 50/50

Prevalence Odds Ratio Usefulness of Cross-sectional study

Prevalence Odds Ratio • Community 80 × 60 • = = 6.0 Screening (normal population) 20 × 40 • Health status • Associations between variables Interpretation: The ratio of the odds of having OA • Surveillance: repeated cross-sectional studies in the obese group relative to the odds in favor of • Clinical practice having OA in non-obese group. • Diagnostic study (illness)

7 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

When we found association… Hill’s causal criteria

• Spuriousness or artifact

Facet Series

• Confounding Control

Case Cross Case Studies Studies Cohort Studies Randomized Controlled Trials • Chance Sectional • Causation Temporality X X √ √ √ Strength X Up to the result Dose-response X Up to the result Consistency X Up to the result Biologic Plausibility N/A Reversibility N/A Specificity N/A Analogy N/A Experimental evidence X X X X √

Advantages Disadvantages of cross-sectional studies of cross-sectional studies (1)

• Good for describing the magnitude and • Length biased sampling: diseases that have distribution of health problems. long duration will over-represent the magnitude • Generalizability. of illness while short duration will under- represent illness • Quick, conducted over short period of time, easy, inexpensive. • Can study multiple exposures and disease • Prevalent rather than incident cases of disease outcomes simultaneously. are identified – exposures may be associated with survival rather than risk of development of disease.

45 46

Disadvantages Disadvantages of cross-sectional studies (2) of cross-sectional studies (3)

• Difficult to separate cause from effect, because • Confounding factors may not be equally measurement of exposure and outcome are distributed between the groups being compared conducted at the same time (difficult to and this unequal distribution may lead to bias establish temporal relationship) and subsequent misinterpretation. • Can assess only association but not a “causal association”.

47 48

8 Case Series, Descriptive, and Cross-Sectional 6-Aug-14 Studies

Bias in Cross-Sectional Studies

1. Selection bias - Sampling bias - Response and non-response bias

2. Information bias Section for Clinical Epidemiology & Biostatistics 3. Confounding Faculty of Medicine Ramathibodi Hospital

facebook.com/ramaclinicalepi www.ceb-rama.org