The dept librarian is back and I finally was able to get the whole article. Recall I had noted my surprise, and they did write a good article applying those very sophisticated tools of both statistics and metanalysis jargon. Also their discussion was appropriately circumspect (in contradistinction to their abstract), noting possible problems with the data, and their conclusion is the need for another large trial (Oh my God another expensive multicenter...!).
However the major problems here are as follows: The total participants in each subgroup ranged from 364-764 (thus picture the Ns of adverse reactions). None of the large trials were in this analysis, and despite the sanctity of mathematical formulas as well as the sophisticated science some worship called metaanalysis, we have seen several instances now where very promising metaanlyses proved erroneous when a very large , well designed, and well monitored multicenter trial was performed (after which some added
them to metaanalyses and continued to rave!).
There are other problems here, one of the trials was listed as ASA vs no treatment instead of placebo (minor). We know that the clinical diagnosis of preeclampsia is often wrong, especially in multipara, but here one supposes that might mask effects. However aspirin is known to decrease GFR in some patients with underlying renal disease, and I always worry that a little less GFR means a little less protein filtered, and ? Less in the urine. Finally, given the views in those articles I sent that critiqued subgroup analyses, the number of endpoints dictates that a p of 0.005 would be needed before one thinks of significance, and in fairness to the authors several are quoted as <0.001. However, as the article continues even here there would be false positives.
In the last analysis, there has always been a small effect in the aspirin trials, but is it much larger if started prior 16 wks? I do not think the data for a conclusion is there.
Again, it been a good exchange, and now to other things.