Five years ago, we moved my son into a college dorm. I know that makes us sound like thousands of other families with 18-year-olds, but my son has Duchenne muscular dystrophy. He couldn’t walk, and because of his waning arm strength, we all thought his college independence would be short-lived.

Instead, we are now getting ready to move him into that same dorm again, where he still lives independently, so he can begin the second year of his master’s degree study. He has spent those five years on Capricor’s deramiocel, a drug that received a negative FDA advisory committee vote last week on its secondary outcome: the definitiveness of its ability to stabilize heart function in a population of boys that included ones whose hearts were still stable. It was much noisier data than the significance seen in the subpopulation that already had signs of heart dysfunction.

The decision demonstrates something important: Designing a rare disease study consists of hundreds of tiny decisions. Sponsors make them with the best knowledge they have at the time but with no way of predicting each decision’s impact with high degree of certainty because they are the ones blazing the trail. But lots of sick people’s lives depend on getting these little decisions right.

One general example: Narrowing inclusion criteria gives you a better chance at statistical significance, but it introduces risk on reaching enrollment, which can then jeopardize the whole program.

Another specific one from my son’s study: The primary outcome is the performance of upper limb, which is divided into measurements of shoulder, arm, and hand function. You can use the total measure or an individual domain.

The total measure gives you more tasks that could potentially capture meaningful changes in your weaker (hand tasks) or stronger (shoulder tasks) participants. But it can introduce noise because lots of participants have already lost their shoulders, and lots of them are going to keep their hands. You could focus only on the arm tasks to shoot for better statistical significance, but are you then going to risk leaving meaningful change for your weakest and strongest participants on the table?

We’re into a second decade of clinical trials in Duchenne; by now, those of us in the Duchenne community are well aware of these challenges. We just don’t have enough data yet to guide the little decisions with certainty as the field continues to evolve. Other rare diseases are in the same boat.

That brings us to the pivotal question on regulatory flexibility in rare disease: What is the appropriate balance between the strictest possible definition of pre-specification and the necessary ability to learn as we go?

Striking that balance is an art, and it’s a task for the Food and Drug Administration that I don’t envy. There are passionate advocates on both sides of the scale. They have valid and legitimate points about patients’ urgent, unmet need, about the need to uphold pre-specification as the primary tenet of scientific experimentation.

Last week, the FDA chose to allow Capricor to submit additional upper limb data and analyses to refocus on the primary outcome. The agency weighed the completion and submission of a new biologic licensing application against the lives and function that would be lost during the time required for completion and chose to side with the patients. My family and I are grateful.

It is the kind of regulatory flexibility that can be exercised in rare disease without sacrificing rigor. My hope, as regulators turn their attention to the art of analysis for the upper limb data and navigate the totality of all those little decisions, is that their guiding principle is trying to understand what’s really happening to young men on treatment like my son, and that they use appropriate and rigorous methods for doing so. They are determining his fate.

Mindy Leffler developed the Duchenne Video Assessment, used in Capricor’s Phase 3 study, and consults on its data.