14.Randomization08.Parallel - sporedata/researchdesigneR GitHub Wiki

1. Use cases: in which situations should I use this method?

  • Parallel randomization is used when evaluating the efficacy (i.e., a placebo is involved) or effectiveness (i.e., comparing two or more treatments) of different therapeutic, diagnostic, or prophylatic interventions, as well as screening mechanisms.

2. Input: what kind of data does the method require?

3. Algorithm: how does the method work?

Model mechanics

  • Randomized controlled trials are still considered the gold standard when it comes to the evaluation of efficacy or effectiveness of clinical interventions. This position is related to their ability to address confounding through measured and unmeasured confounders.

  • External validity tends to be problematic for trials since recruitment practices and strict protocols lead to the recruitment of patients who are fairly different from real-world patients. Specifically, patients in trials tend to be richer, more educated, male, and not from minority groups. This problem can be partially addressed through pragmatic trials.

  • Balance of baseline variables is not a requirement for avoiding biases, as some remaining imbalances in measured variables might be balanced by unmeasured variables [1].

  • Guidelines for the Content of Statistical Analysis Plans in Clinical Trials [2].

  • In a trial where the randomization was flawless, but a large percentage of patients were excluded after randomization, the exclusion of the patients could be influenced only by events that happened after the randomization. If the events that motivated the exclusion are associated with treatment outcomes, for example, this may introduce bias in the study, since the sample would be "manipulated" based on the response the participants had to the treatment. For example, let's say that the treatment does not work very well for very sick patients, and that participants who are sicker have a higher chance of dropping out of the study. If all participants who dropped out of the study are removed, the sample will be unbalanced in favor of the healthier patients.

  • Attrition related to treatment outcomes (especially adverse events) could lead to bias. Another one could be that the patient didn't like to get into a certain arm -- this could happen if the trial is unblinded, or if they have a feeling that they are in an intervention arm that they do not want to be (if you are randomizing beta-blockers vs placebo, for example, people in the beta-blocker group will have their heart rate go down and they could break the randomization). In these situations, bias will be present as the attrition is related to the outcome as well as to the assignment. This is probably why in practice randomized trials are not as pristine as most people would like to believe and require a certain level of caution.

Reporting guidelines

Data science packages

Suggested companion methods

Learning materials

  1. Books *
  2. Articles

4. Output: how do I interpret this method's results?

Mock conclusions or most frequent format for conclusions reached at the end of a typical analysis.

Tables, plots, and their interpretation

5. SporeData-specific

Templates

Data science functions

References

[1] Senn S. Seven myths of randomisation in clinical trials. Statistics in medicine. 2013 Apr 30;32(9):1439-50.

[2] Gamble C, Krishan A, Stocken D, Lewis S, Juszczak E, Doré C, Williamson PR, Altman DG, Montgomery A, Lim P, Berlin J. Guidelines for the Content of Statistical Analysis Plans in Clinical Trials. Jama. 2017 Dec 19;318(23):2337-43.

[3] Senn SJ, Lewis RJ. Treatment effects in multicenter randomized clinical trials. Jama. 2019 Mar 26;321(12):1211-2.

⚠️ **GitHub.com Fallback** ⚠️