A very brief description of RCTs and LSTs pertaining to pharmacoepidemiological research

1. Randomized Controlled
Trials (RCTs) & Large
Simple Trials (LSTs) in a
nut shell

2. Randomized Controlled Trials (RCTs)
• Provide unbiased estimates of drug’s beneficial effect(s),
risk of adverse outcomes.
• Are considered the ‘gold standard’ for demonstrating the
effectiveness of a new medication
• Generally used to evaluate drug’s beneficial effects
• Most commonly used during the premarketing phases of drug
development to demonstrate a drug’s efficacy (and to gather
general information concerning drug safety profile).
• By randomization, researcher hopes to ensure equal
distributions of confounding factors in the population.

3. *** General features of RCTs
• Larger sample sizes: equal distribution of confounders
• Blinding: prevents bias
• Choice of Control group: active controls are better than placebo
• Exhaustive data collection and analyses
• Results can be generalized
*** During pre-marketing phases of drug development, RCTs
– involve highly selective (specific to study) subjects.
– are designed to be sufficiently large to provide evidence of
beneficial clinical effect of a drug .
– are rarely large enough to reliably estimate the risk of rare
events.
– generally evaluate beneficial clinical effects.

4. LIMITATIONS OF RCTs
• It may be unethical to conduct a study in which patients are
randomly assigned a ‘potentially’ harmful treatment.
E.g., a RCT to test the hypothesis that second-hand cigarette
smoking increases the risk of heart disease.
• Complexity and cost of premarket RCTs, with their detailed
observations and resource-intensive follow-up, make very large
studies of this type generally infeasible.

5. LARGE SIMPLE TRIALS(LSTs)
• Is the best solution when it is not possible to completely control
confounding.
• The US Salk vaccine trial (1950s) is an early example of LST.
• More recently, large randomized trials have been used to test
the efficacy of therapeutic interventions, especially in
cardiology, or to evaluate dietary supplements for primary
prevention of cardiovascular disease and cancer.
• Used successfully to evaluate the risk of adverse drug effects;
• LSTs are very large RCTs made simple by reducing data
collection to the minimum needed to test only a single
hypothesis (or a few hypotheses).
• Randomization of treatment assignment is an important feature
of LSTs.

6. Yusuf and Hasford’s statements about LSTs
• Yusuf et al. (1984) – in very large randomized studies of
treatment-related mortality, the researcher must collect only the
participants’ vital status at the conclusion of the study (minimal
but relevant data is sufficient).
• Hasford (1994) – “large trials with lean protocols” include only
relevant baseline, follow-up, and outcome data.
• With simplified protocols it is relatively easier to test hypotheses
related to pharmacoepidemiology.

7. When are LSTs appropriate?
1) Very important research question (priority and cost involved)
E.g., (i) A minor medication side effect (headache or nausea)
may not be a simple issue for the individual patient but, is not a
top priority in research by LSTs.
(ii) If the question involves the risk of premature death,
permanent disability, hospitalization, or other serious events,
the cost may well be justified.
2) ‘Uncertainty’ must exist (blinding)
• Both patient and doctor should be substantially uncertain about
the appropriateness, of each of the trial treatments (for that
particular patient).
• If the patient and doctor are reasonably certain (aware) that one
or other treatment is inappropriate then it would be unethical for
the patient’s treatment to be chosen at random.

8. • Apply this principle to evaluate if it is appropriate to conduct an
LST to test a hypothesis related to the risk of an adverse
clinical event.
3) ‘Study Power’ is substantial or significant
• Study power: no. of events observed during the study period
(determined by incidence rate, sample size, duration of
observation and follow-up)
4) Confounding
• LSTs will only be needed if…
(i) absolute risk of the study outcome is small; confounding by
indication is probable;
(ii) relative risk is small; residual confounding is probable;

9. • LSTs would not be necessary if…
(i) absolute risk or relative risk is large
(ii) confounding (by any source) is not an issue;
5) Association between exposure (treatment) and outcome is not
variable across the sub-groups (i.e., no effect modification)

10. When are LSTs feasible?
1) Simple hypothesis (not very complex)
2) Simple treatment (a single drug at a fixed dose for a short
duration)
3) Simple, easily-defined and measurable outcomes
4) Appropriate follow-up methods are available
5) Population (cooperative and motivated to remain in the study)

11. The End