Critical Appraisal of Randomized Controlled Trials

Intention-to-Treat Analysis

• An ITT involves analyzing patients who drop out or cross over into the alternative arm in the arm they were randomized to.
• In most trials where the outcome is something that improves over time with treatment, the last observation or measurement is used as the final outcome, referred to as the 'last observation carried forward' (LOCF).
• In trials where the outcome is something to be avoided, those who drop out can be assumed to have had the outcome.
• ITT analysis cannot compensate for large losses to follow-up, and if more than 10% are lost, even the ITT results should be viewed with caution.

Types of Outcomes

• Surrogate outcomes: biomedical markers that patients cannot feel or report.
• Patient-relevant outcomes: things that ultimately change for a person.
• Clinical outcomes: outcomes that are directly related to the patient's health.

Continuous vs. Dichotomous Outcomes

• Continuous outcomes: compared using mean.
• Dichotomous outcomes: analyzed using risk difference (RD), absolute risk difference (ARD), number needed to treat (NNT), relative risk (RR), or odds ratio (OR).

Relative Risk and Odds Ratio

• RR of 1.21 means 1.21 times the chance of healing on drugs compared to placebo at 5 days.
• RR can also be expressed as a 21% relative increase in the chance of being healed at 5 days with the drug.
• Odds ratio (OR) of 1.49 means the odds of the outcome are 1.49 times higher in the treatment group.

Critical Appraisal of RCTs

• Quality is determined by how well the trial ensures equal distribution of known and unknown risk factors between treatment and control groups.
• Conduct of the trial should minimize systematic bias and random error to provide accurate results.

Sources of Bias in RCTs

• Selection bias: using methods other than randomization to allocate treatment.
• Performance bias: systematic differences in care provided apart from the intervention being evaluated.
• Attrition bias: systematic differences between groups in withdrawals from the study.
• Detection bias: systematic differences in outcome assessment.

Assessing Internal Validity of RCTs

• Was the subject's treatment allocation randomized?
• Was the study double-blinded?
• Was intention-to-treat analysis performed?
• Was the subject's treatment allocation randomized?

Randomization

• Randomization ensures equal chance of allocation to treatment or control groups.
• Randomization improves the likelihood of similar baseline characteristics between groups.
• Randomization eliminates bias introduced by investigators assigning patients to treatment groups.
• There are two types of randomization: simple and restricted.

Outcome Measures

• Investigators need to clearly define all outcome measures before initiating the study to avoid fishing for significant results.
• The primary outcome measure is the most important efficacy parameter predetermined by the authors.
• The primary outcome measure is used to determine sample size.
• Secondary outcomes are used to examine other outcomes of interest.
• Validated tools or assessment scales should be used for outcome assessment to allow for comparisons between similar trials.