Study Designs in Patient-Oriented Research

Descriptive Studies 

Analytic Studies 

Analytic studies are a more commonly encountered category of studies, involving comparisons between 2 or more groups. They are based on a research question and are etiologic, diagnostic, prognostic, therapeutic, and so on. Based on research architecture, the studies can be observational or experimental.

Observational Studies 

In observational studies, exposure is not determined by the investigator.

Cross-sectional Studies 

•No longitudinal component (Fig 2)

View Large Image Download to PowerPoint

Fig 2. Cross-sectional study.

•Data on exposure and outcome assessed at same time

•Strength: no waiting (fast and inexpensive)

•Weakness: limited inference between exposure and outcome

•Caution: potential biases (eg, response/participation bias [sicker patients are more or less likely to participate])

Examples of Cross-sectional Studies 

•Kramer HJ, Nguyen QD, Curhan G, et al: Renal insufficiency in the absence of albuminuria and retinopathy among adults with type 2 diabetes mellitus. JAMA 289:3273-3277, 2003

•Nash D, Magder L, Lustberg M, et al: Blood lead, blood pressure, and hypertension in perimenopausal and postmenopausal women. JAMA 289:1523-1532, 2003

Cohort Studies 

•Follows up groups of participants (eg, exposed and nonexposed) over time (Fig 3)

View Large Image Download to PowerPoint

Fig 3. Cohort study.

•Describes natural history, establishes temporal sequence

•Estimates of incidence and relative risk can be obtained

•Two common variations, prospective and retrospective studies, depending on whether exposure and outcome occurred concurrent with research or prior to research, respectively:

Prospective Cohort Design 

•Useful strategy to assess cause-effect relationship

•Important design for predictors that can otherwise have recall bias (eg, diet, alcohol)

•Requires large sample size and long follow-up periods

•Allows calculation of relative risk for given exposure and outcome

•Limitations: inefficient method for studying rare diseases; confounding factors can threaten validity of findings

•Caution: preventing losses during follow-up is important during conduct of study for accurate results; other biases (surveillance bias [more procedures in group with exposure])

Retrospective Cohort Design 

•Much less time-consuming and costly compared with prospective cohort studies

•Limited control over sampling of cohort and quality of predictor variables (collected in past)

Propensity Score Methodology 

•Improves causal inference in observational studies when compared groups are different at baseline with respect to intervention (exposure)

•Propensity score is model-based predicted probability of receiving intervention (exposure)

•Controls for confounding and minimizes loss of degrees of freedom in statistical analyses

•Outcomes in intervention and control group are “weighted” across values of propensity scores

Examples of Cohort Studies 

•Go AS, Chertow GM, Fan D, et al: Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med 351:1296-1305, 2004

•Ojo AO, Held PJ, Port FK, et al: Chronic renal failure after transplantation of a nonrenal organ. N Engl J Med 349:931-940, 2003

•Abbott KC, Trespalacios FC, Agodoa LY, Taylor AJ, Bakris GL: Beta-blocker use in long-term dialysis patients: Association with hospitalized heart failure and mortality. Arch Intern Med 164:2465-2471, 2004

•Kayler LK, Rasmussen CS, Dykstra DM, et al: Gender imbalance and outcomes in living donor renal transplantation in the United States. Am J Transplant 3:452-458, 2003

Case-control Studies 

•See Fig 4

View Large Image Download to PowerPoint

Fig 4. Case-control study.

•High yield of information from relatively few subjects

•Efficient design for rare outcomes

•Can be used to generate or test hypothesis

•Estimates odds ratio; estimates strength of association between predictor and disease (outcome); odds ratio is good approximation of relative risk if outcome is rare

•Limitations:

■Cannot evaluate incidence/prevalence

■Only 1 outcome usually studied

■Limited control over quality of predictor variables

■Increased susceptibility to bias: sampling bias (selection of controls from unrelated population), recall bias (smoking or alcohol intake history)

•Nested case-control design:

■Case-control study embedded within cohort study

■Particularly useful design for predictors (eg, biomarkers) that are expensive to measure in whole cohort

■Preserves advantages of cohort studies (ie, cases and controls are from same population and predictors are measured before outcomes)

■For biomarker studies, samples need to be stored until outcomes occur for diagnostic studies

Examples of Case-control Studies 

•Keller G, Zimmer G, Mall G, et al: Nephron number in patients with primary hypertension. N Engl J Med 348:101-108, 2003

•Alfrey AC, LeGendre GR, Kaehny WD: The dialysis encephalopathy syndrome. Possible aluminum intoxication. N Engl J Med 294:184-188, 1976

•Parikh CR, McCall D, Engelman C, Schrier RW: Congenital renal agenesis: Case-control analysis of birth characteristics. Am J Kidney Dis 39:689-694, 2002

Experimental Studies (Randomized Controlled Trials) 

•A properly planned and executed randomized clinical trial is powerful experimental technique for assessing effectiveness of an intervention

•Randomized, double-blind, placebo-controlled trial (Fig 5):

■Strongest evidence for cause-effect association

■Best possible design for some questions

■Randomization balances predictor variables across treatment groups (protecting against confounders):

○Block randomization: ensures that participant numbers are equally distributed

○Stratified randomization: promotes balance of a particular predictor of outcome

○Other: randomization by site in multisite trial

■Blinding helps avoid bias during data collection and assessment:

○Unblinded: investigator and participant know about intervention type

○Single-blind studies: only investigator or patient is aware

○Double blind: investigator and subject both are unaware

■Components of good clinical trial study protocol:

○Objectives:

□Primary aim or question: including predictor and response variables

□Secondary aim or question: including predictor and response variables

□Subgroup hypotheses, if any

□Adverse effects

○Design of study:

□Study population: inclusion criteria, exclusion criteria

□Sample size assumptions and estimates (including statistical power)

□Enrollment of participants: informed consent, assessment of eligibility, baseline examination, allocation of intervention (method of randomization)

□Intervention: description and schedule, measures of adherence

□Follow-up visit description and schedule

□Ascertainment of response variables: training, data collection, quality control

□Data analysis: interim monitoring, final analysis

□Termination policy

○Organization:

□Participating investigators: statistical unit or data-coordinating center, laboratories and other special units, clinical centers

□Study administration: steering committees and subcommittees, data monitoring and safety committee, funding organization

■Limitations:

○Costly in time and money

○Some research questions not suitable for experimental design (eg, etiology, adverse effects)

○Research interventions may not be feasible in clinical practice

○Can involve restricted scope and narrow study question

View Large Image Download to PowerPoint

Fig 5. Randomized controlled trial.

•Modifications of randomized trials:

■Factorial design: 2 separate questions in single cohort of patients

■Randomization of matched pairs

■Randomization of groups or clusters

■Crossover design: intervention followed by placebo in 1 group and converse in another group:

○Limitation: carryover effect

○Adequate “washout” periods needed

■Trials for Food and Drug Administration approval of new therapies:

○Phase I: unblinded, uncontrolled studies in few volunteers to test safety

○Phase II: relatively small, randomized, blinded trial to test tolerability and different intensity or dose of intervention on surrogate outcomes

○Phase III: relatively large, randomized, controlled, blinded trials to test effect of therapy on clinical outcomes

○Phase IV: large trials or observational studies, conducted after intervention has Food and Drug Administration approval, to assess rate of serious side effects and evaluate additional therapeutic uses

Integrative Studies 

Designing Studies for Medical Tests 

Additional Reading 

1. Feinstein AR: Multivariable Analysis: An Introduction. New Haven, CT, Yale University Press, 1996

2. Flather M, Aston H, Stables R: Handbook of Clinical Trials. London, ReMedica, 2001

3. Friedland DJ: Evidence-Based Medicine: A Framework for Clinical Practice. Stamford, CT, Appleton & Lange, 1998

4. Friedman LM, Furberg C, DeMets DL: Fundamentals of Clinical Trials (ed 3). New York, Springer, 1998

5. Gordis L: Epidemiology (ed 2). Philadelphia, PA, Saunders, 2000

6. Hulley SB: Designing Clinical Research: An Epidemiologic Approach (ed 2). Philadelphia, PA, Lippincott Williams & Wilkins, 2001

7. Riegelman RK: Studying a Study and Testing a Test: How to Read the Medical Evidence (ed 4). Philadelphia, PA, Lippincott Williams & Wilkins, 2000

8. Rothman KJ, Greenland S: Modern Epidemiology (ed 2). Philadelphia, PA, Lippincott-Raven, 1998

9. Concato J, Shah N, Horwitz RI: Randomized, controlled trials, observational studies, and the hierarchy of research designs. N Engl J Med 342:1887-1892, 2000

10. Rubin DB: Estimating causal effects from large data sets using propensity scores. Ann Intern Med 127:757-763, 1997

11. Concato J: Overview of research design in epidemiology. J Law Policy XII:489-507, 2004