There’s something to be said for papers that state the obvious. That’s generally that it’s not actually obvious until it’s been shown to be the case, ’cause it always might have been different. (For those of you who struggle with the idea about studying the obvious, have listen to this superb paper by Ruth Gilbert about sleeping position and SIDS.)
Take NSAIDs and ulcers in children. In adults, it’s pretty well established that the habitual user of large doses of aspirin has a greater risk of popping an ulcer than those who don’t indulge. In children, this has been more debatable, with large RCTs comparing paracetamol and ibuprofen showing no difference in adverse effect rates. A very large cohort of patients in Italy has been analysed to demonstrate that, as with adults, those kids who’ve been given NSAIDs compared to no drugs have a higher rate (OR ~3 ) of admission with proper, proven, upper GI complications. The same study shows that the best guess for paracetamol, when compared to no drugs, is OR ~2.
Case closed? NSAIDs evil? Worse than paracetamol? P’raphs not, m’lud. If we examine the confidence intervals (a rough & ready way of seeing if two things are equivalent, inferior/superior, or ignorant) then we see:
NSAID 2.9 (2.1 to 4.0)
Paracetamol 2.0 (1.5 to 2.6)
Both 3.1 (1.5 to 6.2)
There’s no classical statistically significant evidence here that paracetamol is safer (the data would support it being worse – could be 2.6 vs 2.1 NSAID, and both may be safest of all OR 1.5) but the balance of probability rather than proof would sit with the paracetamol being kinder.
The next step in undertaking the balancing should be to look at the absolute rates involved, to calculate the meaning of this to an individual family. The paper estimates an absolute rate of problems of 2.4 / 10,000 ED visits. If even 1 in 10 children taking analgesics attended the ED, this would be 2.4 / 100,000 children. Again, taking an extreme case of this being the non-drug rate, and the point estimates of the OR for common analgesics leads an NNH of 20800 children for NSAIDs vs. nothing or 41600 children for NSAIDs vs paracetamol.
Does this alter your view of the initial findings?