Last week's blog looked at a paper which analysed a couple of cohort studies to show the benefit of improving your diet - primary prevention. This week it's the turn of treatment, with a clever study design. The researchers compiled data from trials published over a 20 year period to see how the risk of sudden death from heart failure has changed. Their hypothesis was that as better treatments have become available, fewer people should die suddenly.
The authors conclude that this is indeed what happened. It's a nice study but I'm not completely convinced: figure 2 has a line drawn through it to indicate linear decline, but to me the data don't look linear: seems more like everything flat but 3 studies in the early years with high rates.
They must have thought of that (the methods mention a sensitivity analysis which omitted each single study in turn to test for outliers) and in the NEJM the statistical refereeing will have been heavyweight, but it's always important to inspect the raw data before processing through regression models etc.