Who was enrolled, and against what
The enrolled group is narrow, and narrow in the right way: adults whose immune thrombocytopenia had failed corticosteroids and at least one recommended subsequent treatment, randomised 108 and 108 through a central web-based system across several centres. The primary endpoint was prespecified as durable response at six months rather than a platelet count on one good day. That distinction accounts for most of the difference between a number that means something to a patient and one that means something to a graph.[1]
The comparator is danazol alone. That makes the question the trial answers a specific one: in this group, does adding baricitinib to danazol help? It does not answer whether the combination beats the treatments a haematologist would otherwise reach for next. The absolute difference is 25 percentage points, so roughly four patients must be treated with the combination for one additional durable response, which is a good number in a group defined by having already run out of options.[1]
Patients and caregivers knew their assignment; the investigators judging efficacy did not. For an endpoint anchored in laboratory platelet counts, that ordering protects the measurement reasonably well. It does not protect everything unblinding can touch, and the honest way to hold the result is that the effect is large enough that expectation alone is an unlikely explanation, not that expectation was excluded.[1]
What 108 patients can say about safety
The safety set is 108 in the combination arm and 105 in the monotherapy arm, three fewer than were randomised, because it counts only patients who received at least one dose. At least one adverse event was reported by 48.1 per cent and 44.7 per cent, a gap of 3.4 points. Each arm had two discontinuations for adverse events and one event of grade 3 or worse.[1]
Read carefully, those numbers say two things: the added drug did not visibly raise toxicity at this dose over this duration, and the trial was never sized to find a rare harm. With about a hundred patients per arm and six months of follow-up, an adverse event occurring in fewer than one patient in a hundred can pass unseen, and a harm that takes a year to appear cannot be seen at all. Neither of these is a criticism of the trial; both are limits of its denominator.[1]
The version currently available is the unedited accepted manuscript, which the journal flags as subject to further editing before final publication. For a phase 2 result that is a reason to hold the decimals loosely rather than the direction. The direction is that a combination assembled from two already available drugs earned a larger trial in a group of patients who are hard to help, and durability past six months is a question that trial will answer.[1]