Clinical Trials and Hand Surgical Literature

Brent Graham

Overview

The main concern with any type of scientific report is the presence of bias in the study.
Bias may enter into a study at many different levels: Sample, the intervention or observation of interest, the methods of measurement, the statistical analysis, the interpretation of the findings.
Selection of the study sample is often the most obvious and important source of bias. Samples of convenience drawn from a particular setting, such as an academic institution or subspecialty clinic, are unlikely to be representative of the general population of interest, especially if the sample is small.
Various study designs attempt to address the issue of bias in the study sample by assigning a control group.
In the uncontrolled case series, which remains the most common sample reported in the hand surgery literature, there is an implied historical control although making comparisons to other series in the literature is often inappropriate.
A study can be blinded to the provider of care and or the patient. A “blinded” study is a study wherein the evaluator does not know which group the patient or variable being examined is in. A “double blinded” study is one wherein neither the patient nor the evaluator knows which treatment was given, or whether one was given at all (dependant on study design).

I. Hierarchy of Clinical Research Designs

Randomized control trial
- Advantages
  - Controls for both known and unknown biases by randomly assigning eligible patients to treatment groups
  - Comparison of results is usually very intuitive and clear
  - Gold standard for defining evidence
  - Can establish cause and effect relationships
- Disadvantages
  - Cost; primarily related to needs for data collection/security and the time required to recruit enough participants into the study
  - Generalizability may be poor because study sample does not necessarily resemble the patient population of interest
  - Answers only the question posed by the study
  Randomized controlled trials are most effective when the time required to ascertain the result is relatively short. Interventions that may require a lengthy period before the outcome is known are difficult to study with a randomized trial because costs may become substantial.
Cohort

Cohort designs are used to establish associations between exposures and outcomes by identifying samples with and without exposure to factors of interest and then following through time to observe an outcome.
- Advantages
  - Feasibility, especially for the study of rare exposures unless the outcome requires a long time to be identified
  - Useful for studying occupational exposures
- Disadvantages
  - Confounding; the samples are not randomly chosen so spurious associations may be seen if the sample is somehow nonrepresentative
  - Can only identify associations, not cause and effect
  Beware the retrospective cohort represented as a prospective cohort study! The measurements made on the sample may not be of adequate quality if the cohort was assembled for another purpose and then secondarily studied for the current question.
Cross-sectional study

All measurements are made at once in a sample drawn from the population of interest; distribution of variables within the sample may allow conclusions to be made about various associations.
- Advantages
  - Feasible; since all measurements are made at one time, costs are relatively low.
- Disadvantages
  - Can only identify associations, not cause and effect
Case control

In contrast to the cohort study, the case control starts with the outcome and then works backward to identify exposures of interest in the sample and in a control group matched to the study sample for important confounding variables.
- Advantages
  - Feasible; because the samples sizes are relatively small and the outcome has already taken place, costs related to acquiring the data are usually small.
  - Effective for studying rare occurrences.
  - Results are usually expressed as an odds ratio, which is easily understood by users of the study.
- Disadvantages
  - Predictor variables are often measured retrospectively leading to bias.
  - Can only identify associations, not cause and effect.
  - Because the process is controlled by the investigator, the selection of controls may be biased; should have at least two (or more) controls for each case although this is rarely done in the hand surgery literature.
Decision analysis

This approach models a clinical condition and systematically evaluates the expected utility of various decisions about testing, therapy, etc. Informs clinical
policy about a particular condition but does not necessarily provide any insight into the treatment of an individual.
- Advantages
  - Feasible; does not always require the collection of data from patients but may use data from the literature.
  - Transparent; users may be able to fully replicate the model because investigators are obligated to fully define all the variables.
  - Flexible; sensitivity analysis reveals the variables that are most influential over the decision regarding the most desirable strategy. This may provide insights into the problem that would not be evident in an experimental design like a randomined trial.
- Disadvantages
  - Measures the utility, or preference for a given health state or outcome, which may not be fully intuitive to clinician
Health services research

Sometimes called “outcomes research,” this type of approach involves the analysis of data obtained for another purpose, frequently billing information for large health care aggregates like Medicare.
- Advantages
  - Large volumes of data available for analysis.
  - Usually the studies are inexpensive because the data have already been collected.
  - Can provide interesting insights into the process of care.
- Disadvantages
  - Can only identify associations, not cause and effect
  - Usually more effective as a method for the generation of new hypotheses rather than the answering of specific research questions
  Although only a prospective study design can demonstrate cause/effect relationship these are often prohibitively expensive. A substantial body of lower-level evidence that identifies a strong and consistent association may be almost as convincing. It is now common and required by many of the journals to state the level of evidence of the publication along with the abstract (Table 37.1).

II. Framing a Research Question

Whether planning a clinical research study or evaluating clinical research reported in the literature, it is useful to consider the characteristics of an appropriate research question. Think of the research question as FINER if it meets these criteria:

Feasible

Interesting

Novel

Ethical

Relevant

These are the questions reviewers and editors should be asking themselves when evaluating manuscripts submitted for publication. Stated another way: Is it new? Is it true? Does it matter?