{"id":5968,"date":"2025-05-25T17:10:59","date_gmt":"2025-05-25T17:10:59","guid":{"rendered":"https:\/\/journalofmedicaloptometry.com\/vol4issue2\/?p=5968"},"modified":"2025-05-29T13:56:54","modified_gmt":"2025-05-29T13:56:54","slug":"the-most-important-research-trial-metric-for-clinical-practice","status":"publish","type":"post","link":"https:\/\/journalofmedicaloptometry.com\/vol4issue2\/volume-3-issue-2\/the-most-important-research-trial-metric-for-clinical-practice\/","title":{"rendered":"The Most Important Research Trial Metric for Clinical Practice"},"content":{"rendered":"

doi:10.62055\/70688991Nq<\/a><\/p>\n

I love key performance indicators (KPIs). Whether it\u2019s in my practice, my finances, or my running, I\u2019ve come to appreciate the clarity they offer. They help me track progress, adjust course, and most importantly, make decisions rooted in data rather than guesswork.<\/p>\n

In my business, I use revenue per OD hour to analyze efficiency. In my life, I measure weekly mileage, books read per month, and my monthly saving and giving percentages. Why? Because KPIs simplify complex systems and anchor decisions in something more reliable than a hunch.<\/p>\n

But what about in the clinic?<\/p>\n

In medicine, we often substitute precision with abstraction. We lean on statistical terms like p-values<\/em>, confidence intervals<\/em>, and relative risk<\/em>\u2014concepts that rarely translate to the patient sitting in our exam chair. Ask any clinician to explain p-values in plain English, and you\u2019ll likely hear some variation of \u201cit tells us whether the result is significant.\u201d But that\u2019s not entirely accurate. Worse, it\u2019s practically useless when trying to communicate what a treatment might actually do<\/em> for a patient.1<\/sup><\/p>\n

That\u2019s why I believe the Number Needed to Treat (NNT)<\/strong> is the single most important trial metric in clinical practice. It\u2019s not perfect, but when it comes to decision making at the point of care, NNT gets us as close as possible to answering the question every patient is silently asking: is this treatment worth it for me?<\/em><\/p>\n

What is NNT?<\/strong><\/h2>\n

The Number Needed to Treat is exactly what it sounds like: how many patients you need to treat to prevent one bad outcome or to achieve one desired effect. Mathematically, it\u2019s the inverse of the absolute risk reduction (ARR)<\/strong>. That\u2019s the difference between the event rate in the control group and the event rate in the treatment group.<\/p>\n

NNT = 1 \/ ARR<\/strong><\/p>\n

Let\u2019s say a new medication reduces the chance of progression from ocular hypertension to glaucoma from 10 percent to 5 percent. That\u2019s an ARR of 5 percent, or 0.05. Divide 1 by 0.05, and you get 20. So, you\u2019d need to treat 20 people to prevent one case of glaucoma. That\u2019s your NNT.2<\/sup><\/p>\n

It\u2019s a simple, intuitive concept. More importantly, it\u2019s one that patients can understand. You can say, \u201cIf we treat 20 people like you, one will benefit from the medication.\u201d That doesn\u2019t require a background in biostatistics to grasp.<\/p>\n

Why It Matters<\/strong><\/h2>\n

Unlike relative risk reductions (which can sound more impressive than they are), NNT forces us to look at the actual benefit. And it naturally invites a conversation about trade-offs: cost, side effects, adherence, and patient values.<\/p>\n

Want to prescribe statins for primary prevention? The NNT to prevent one cardiovascular event over five years in low-risk patients can be as high as 1003<\/sup>. That means 99 patients take the drug without benefit, exposing themselves to potential side effects, cost, and hassle.<\/p>\n

By contrast, treating bacterial keratitis early with fortified antibiotics? The NNT to prevent vision loss is often low, meaning it\u2019s highly effective and worth doing. NNT anchors our treatment plans in real world outcomes, not just academic abstractions.<\/p>\n

Common NNTs in Medicine and Eyecare<\/strong><\/h2>\n

Some well-known examples:<\/p>\n