Half a seizure cannot occur. Yet in a landmark cannabidiol trial for Dravet syndrome, the reported seizure frequency fell from 12.4 to 5.9 per month in one treatment group. Those decimals came from careful bookkeeping. A family counted whole events, investigators converted observations into comparable monthly rates, and the published paper summarized a group of children and young adults whose starting frequencies varied enormously. Reading that number backward reveals what the trial actually tested.
A diary before the first dose
Seizures in Dravet syndrome typically begin in infancy and can take several forms. In the 2017 randomized trial, 120 participants continued their existing antiseizure treatment. Researchers added either a purified cannabidiol oral solution or a matching placebo. Before the treatment comparison began, caregivers recorded seizures over a four-week baseline. The study then followed participants during a 14-week treatment period.
Why ask families who already knew the condition so well to begin with another four weeks of counting? A remembered “usual month” gives particular weight to frightening days and unusually quiet stretches. A dated diary gives the starting and treatment periods a more comparable basis. Both records can still contain missed or misclassified events. Writing down in advance which events count gives families and researchers a common rule to work with.
The primary outcome concerned convulsive seizures. A fall in that count can be important while other types of seizures continue. The trial registry entry preserves the definitions and planned outcomes, useful context when a shorter account refers only to “seizures.” The shorter word can accidentally enlarge the scope of the result.
What the middle of the group tells us
In the cannabidiol group, median monthly convulsive seizure frequency went from 12.4 at baseline to 5.9 during treatment. The corresponding placebo figures were 14.9 and 14.1. These are group medians. The person in the middle of the baseline ranking need not have been the person in the middle of the treatment ranking. Dividing 5.9 by 12.4 therefore does not recover the trial’s estimate of the typical individual change.
Take two imaginary diary entries. A child who goes from twenty events to ten has ten fewer seizures. A child who goes from four to two has two fewer. Both record a fifty percent reduction, although daily life may have changed in quite different ways. The investigators used each participant’s own baseline to calculate a percentage change, then compared the treatment groups. The adjusted difference favored cannabidiol by 22.8 percentage points.
Who reached zero?
Seizure freedom asks a much stricter question. A participant whose count falls from twenty to one has had a large reduction and still misses that endpoint. In the trial, five percent of participants assigned cannabidiol became seizure-free during treatment, compared with none assigned placebo. The authors did not find that particular comparison statistically significant. The evidence for a reduction in convulsive seizure frequency was clearer than the evidence about reaching zero.
Different families, different counts
When investigators studied cannabidiol in Lennox–Gastaut syndrome, they selected drop seizures as the primary outcome. Those attacks carry a particular risk of falls. A chart for that trial might look almost identical to the Dravet chart, yet the tick marks stand for different events in a different population. Combining the two headlines without looking at the diaries would erase a clinically meaningful distinction.
Even a meticulous count captures only part of a family’s experience. It does not show how long an event lasted, how long recovery took or whether a treatment caused troublesome adverse effects. The Dravet investigators reported more adverse events with cannabidiol than with placebo. Set beside that information, the decimal in the headline becomes more useful. It describes a measured change in recurring events during a defined period, and leaves room for the rest of a child’s day.