Randomized Controlled Trials Questions and answers

systemic reviews and meta-analyses randomized controlled double blind studies cohort studies case control studies The top 4 types of evidence on the hierarchy in order in order from best to least best in vitro (test tube) research animal research ideas, editorials, opinions case reports case series The worst 5 types of evidence on the hierarchy in order from worst to least worst randomized clinical trials These are experiments which involve patients and are designed to evaluate average causal effect of treatment. They're always comparative, as a new tx is compared to a control in the same tx. controls In RCTs, these can be standard tx (standard of care) or placebo if we don't have standard of care average causal treatment effect (ACTE) This estimates the difference in mean outcomes between units assigned to the tx and units assigned to the control. Using RCTs, this is just a comparison in mean outcomes for tx and control arms of the study. mean differences Testing for this is often done with extremely basic stats (e.g. T-tests) randomization This principle is a result of us wanting there to be no difference, on average, between tx and control arms. We do this via random assignment. Patients are generally entered sequentially as you treat patients when they appear with condition and not some time later, therefore, we need to ensure this is designed so that it's preserved over time and treating clinical is unable to influence allocation blinding This principle comes from the need of patients and clinicians. Otherwise, knowledge of tx allocation could influence loss to follow up, patient or physician reported outcomes, adjunct tx. double blind This refers to blinding both patients and clinicians. It's not always possible (open heart surgery) but for drug studies should be insisted on statistical significance Since the difference between two randomized drugs is all random, we can test as if the only differences are due to random error. Convention is to call two groups statistically significant if they have a p-value for a test of less than 0.05 p-value This is the probability of obtaining test results at least as extreme as the results from the randomized trial but where the null hypothesis is still true adherence It's crucial that all RCTs report this or else estimates can't be interpreted by reader. This common issue in medication trials may be influenced by perceptions of drug side effects and/or lack of efficacy. This means we estimate ACTE for participants at the level of this seen n the trial. Most concerning when this differs between tx and control arm since we worry that thus blinding failed intention to treat (ITT) Principle whereby data are analyzed according to group assignments, regardless of how subjects actually completed the study. Gives an estimate of the effectiveness of the medication at adherence levels seen in RCT but underestimates efficacy of the drug if adherence were 100%. Generally seen as biased but this is small and in a conservative direction (issue for safety trials) confounder This is a common cause of both the exposure and outcome. May give misleading estimate of the risk. This always leaves some uncertainty about whether we've found all the confounders. ethical A study can be considered this if participants have potential to benefit, answer is in doubt (equipoise), participants are informed and give consent, and there is no coercion, even indirectly. This means many things (e.g. smoking) can't even be studied. paralle group In this trial structure, participants are individually randomized to tx and control arms crossover trial In this trial structure, participants are randomized to sequences of tx and control, such that participants typically receive multiple allocations over time. Includes washout period. cluster randomized In this trial structure, participants are randomized by centre to tx and control arms, and the analysis is done comparing centres phase 1 This phase of trials has the goal of testing drug safety in humans. It's essentially never randomized. Objective is to puzzle out what dose to use for tx and this is often obtained from dose-escalation experiments. Also involves studies of drug metabolism and bioavailability. These are very small trials a long way away from market. phase 2 This phase of trials estimates effectiveness and safety of the medication. It's often randomized and includes a small group of very closely followed participants. Can be used to pick between competing molecules with similar potential in early pre-clinical studies. phase 3 This phase of trials is the most important one for many drugs. Not all drugs get post-marketing RCTs. This is almost always randomized. The tx is given to large groups of participants to estimate effectiveness and side effects. It's usually what gets a drug to market. phase 4 This phase of trials includes post-marketing investigation of safety or effectiveness (if not randomized, then cohort study), which often evaluates long term side effects or new indications. Can also be used to look at meds in special populations not considered in previous phases like complex patients (DM, CVD), ethnicities not in original trial, genders/sexes underrepresented in original trials baseline Randomization only works on this. Data collected after this can potentially be confounded. This includes measures of trials success like adherence. Anything after this is observational. systematic reviews and meta-analyses Generally, these approaches will summarize the body of evidence using the highest quality data available. Main challenge is often deciding which RCTs are comparable to each other. They're higher levels of evidence because they summarize across the field but often exclude complex patients observational meta-analysis This reviews and summarizes lower quality evidence like observational studies which are biased loss to follow-up This should always be reported by both tx and control arms. sometimes a high loss is unavoidable but generally ideal to have this less than 5%. If it's quite different between tx and control arms, that's very concerning. RASS In this study of drug effectiveness, patients were randomized to losartan, enalapril, or placebo and followed for 5 years. Primary endpoint was change in glomerular mesangial fractional volume in kidney biopsies (made follow up harder), retinopathy endpoint was a 2 step or greater progression in retinopathy.