Study Design Categories, Definitions, and Levels of Evidence
- Prospective, randomized, controlled trials and meta-analyses of prospective, randomized, controlled trials.
The randomized, double-blinded, controlled trial is the gold standard of study design. To achieve this ranking, the study allocation must be blinded to the investigator both before and after the randomization and the assignment to intervention group. This design provides protection from allocation bias by the investigator and from bias in the assessment of outcomes by both the investigator and the patient.
Unfortunately, double-blinding after intervention assignment is not always possible because procedures or side effects often vary substantially among study interventions in ways that are obvious to both the investigator and the patient. In some cases, however, it should be possible to blind the investigator and the patient until the randomization has been made, even if intervention delivery cannot be blinded. Furthermore, randomized trials are still subject to imbalances if sample size is inadequate to the question of interest, and may be less than optimal owing to other design elements such as choice of sample, measurement approach, appropriateness of the comparison condition, or loss to follow-up.
Meta-analysis of randomized studies offers a quantitative synthesis of previously conducted studies. The strength of the evidence from a meta-analysis is based on the quality of the individual studies included. Moreover, meta-analyses can magnify small systematic errors in individual studies.
Because the results of meta-analyses and the randomized trials upon which they are based do not always agree and because meta-analyses performed by different investigators to address the same issue can reach contradictory conclusions, meta-analyses of randomized studies are placed in the same category of strength of evidence as are randomized studies, not at a higher level.
- Prospective, nonrandomized, controlled trials; prospective cohort studies; prospective case series; and cross-sectional studies.
In nonrandomized controlled trials, allocation to intervention group is made on the basis of:
- Birth date.
- Chart number.
- Bed availability.
- Day of clinic appointment.
- Any other strategy known to the investigator before informed consent is obtained from the patient.
Imbalances can occur in intervention allocations under such circumstances but can also be taken into account statistically.
A prospective cohort study is a longitudinal study in which a defined group of individuals who share a certain condition and/or who are exposed to a particular intervention are followed over time and compared with another group of individuals who are not exposed to the intervention under study. The general population from which the cohort is drawn or another cohort can be used as the comparison group. Alternatively, subgroups within a cohort can be compared with each other.
A prospective case series is an uncontrolled study involving either consecutive cases or nonconsecutive cases. Although the evidence from this type of study is weaker than that from a controlled study, this design may be the only one practical in certain settings. Evidence from a consecutive case series is considered stronger than that from a nonconsecutive case series, and population-based series drawn from a well-defined population are stronger than non–population-based series.
Cross-sectional/correlational studies are widely used to investigate the relationship among variables in one or more groups of individuals. Depending on the research question, they may be the highest level of design possible and may even be the most appropriate to examine the relationships among multiple variables. Analytic strategies, including multivariate approaches, need to be matched to the question at issue. A combination of cross-sectional and longitudinal data collection can be explored with very sophisticated path analytic approaches, with the imputation of causality. Quality can nevertheless vary considerably, depending on:
- Choice of measurement.
- Adequacy of sample size.
- Representativeness of sample.
- Quality of statistical analysis.
- Retrospective studies.
Retrospective studies include:
- Case-control studies.
- Retrospective case studies.
- Individual case reports.
In case-control studies, two groups of individuals—one with a disease or condition (cases) and a similar group without the disease or condition (controls)—are compared. The cases should be representative of all cases in a population but do not have to include all cases in a population. The controls should come from the same population as the cases.
Retrospective studies are prone to numerous biases, including selection bias and lack of information about confounding variables from medical records or memories of the study participants (recall bias). Incident cases are preferable to prevalent cases to help reduce bias.
Evidence obtained from retrospective studies, especially case reports, is among the weakest forms of evidence. However, systematic qualitative research, while often retrospective in nature, improves substantially on anecdotal case reports.
- Opinions of respected authorities based on clinical experience, consensus statements from expert committees, or authoritative reviews.
The evidence supporting conclusions or recommendations in opinions, consensus statements, or authoritative reviews can be highly variable and subject to important biases. This category of evidence is usually considered the lowest in many levels-of-evidence rankings.