Section 1.1: The Philosophy Behind Randomised Trials

Imagine you're developing a website or a digital product and have two design options—Design A and Design B. You believe Design A might perform better but wish to avoid bias in your judgment. Thus, you decide to conduct a randomized statistical trial to objectively assess which design is superior.

You divide your audience randomly, presenting half with Design A and the other half with Design B. Using metrics such as 'clicks' and 'time spent,' you analyze which design prevails. After a set period, the results indicate Design B outperformed Design A, prompting you to proceed with Design B for your final product.

This process is commonly referred to as an "A/B" test. While my example is simplistic, real-world applications often involve multiple options and more intricate algorithms, yet the core principle remains: the best solution emerges victorious.

This methodology is so widespread that it's fair to say a significant portion of today's online revenue stems from such A/B tests. However, my focus here is on a more pressing issue—transitioning from digital design to the realm of clinical trials.

Section 1.2: From Web Design to Clinical Applications

Suppose you are exploring treatment options for a disease. Following the standard randomized trial protocol, you would evenly distribute your test group and administer each treatment to an equal number of patients. This resembles an A/B test.

While you can determine which treatment is more effective, a critical concern arises: unlike web design, clinical trials involve human subjects. This means that during the trial, many participants may receive suboptimal treatments. This raises ethical questions.

Some might argue, “It’s a trial, and participants are aware of their involvement!” While I recognize that some degree of sacrifice may be necessary for the greater good in clinical research, what happens when neither the participants nor the researchers fully understand the risks involved? To illustrate this, let’s examine a historical case.

Section 1.3: The ECMO Case Study

In the 1970s, Robert Bartlett from the University of Michigan pioneered a revolutionary technique known as extracorporeal membrane oxygenation (ECMO) to address respiratory failure in infants. This method extracts blood destined for the lungs, oxygenates it externally, and then returns it to the heart.

Despite its effectiveness, ECMO carries risks like blood clots. One successful case does not warrant its widespread application without clinical trials.

Section 1.4: Zelen's Algorithm: A New Approach

To address the ethical challenges posed by randomized trials, biostatistician Marvin Zelen proposed "adaptive" trials in 1969, which focus on a "play-the-winner" strategy. Imagine placing colored balls representing treatment options in a hat. A ball is drawn at random, and if the treatment succeeds, another ball of the same color is added, increasing its chances of being selected in subsequent draws.

Section 1.5: The ECMO Clinical Trial Controversy

Between 1982 and 1984, Zelen's algorithm was employed to evaluate ECMO's effectiveness compared to standard treatment. The results were shocking—while several infants died from conventional treatments, those receiving ECMO showed remarkable survival rates.

Despite the compelling evidence, the scientific community was hesitant to accept such a small sample size as statistically significant. Subsequently, a study in the 1990s in the UK involving 200 infants used traditional randomized trials, resulting in similar conclusions but at the cost of more infant lives.