A couple of weeks back there were numerous reports such as this one saying that research has shown that the use of medication such as ibuprofen can increase the risk of miscarriage by 2.4 times. The figures that then followed did not seem to support this assertion and as usual there was no link to the research. However, a bit of googling turned up an email address for the lead researcher, Dr. Anick Berard of the University of Montreal. So I thought I’d ask her directly:
Dear Dr Berard
Your research into links between NSAIDS and miscarriage has been widely reported in the British press. I am considering blogging about it myself so if you have the time, I would be grateful if you could clarify something:-
One newspaper states that your team “examined 4,705 miscarriage cases and found that 352 (7.5%) had taken a non-aspirin NSAID. [These] were compared with 47,050 women of a similar age who had not lost a baby during pregnancy. A total of 1,213 of them (2.6%) had used an NSAID. You are then quoted as saying that this is associated with a 2.4 fold increase.
I was wondering how this figure was arrived at. Dividing the percentages produces a ratio of 2.9 not 2.4. In any event, would that be the appropriate calculation to make? I can see that would be the case in a cohort study where the comparison was between miscarriagerates among those who used NSAIDs as opposed to those who did not but that does not appear to be the case here.
I may be misunderstanding but I would have thought that the difference in percentages was the important figure in a case control study, in this case 4.9%.
I received a prompt reply:
Dear Mr Hawcock,
The percentages you are referring to are for the crude estimate of risk. However, once we took away the effect of the underlying diseases, the differential effect of health care utilization as well as use of other medications and sociodemographic characteristics between cases and controls that could also be increasing the risk of miscarriage, our estimate decreased to 2.4 (we are interested in the pharmacological effect only). Hence our adjusted estimate tells us that women taking NSAIDs during early pregnancy are 2.4 times more likely to have a miscarriage than those who don’t and that was seen for all types of NSAIDs and dosages.
Please let me know if you need more information.
Envoyé de mon iPhone
Le 2011-09-07 à 18:40,
This did not resolve my confusion, so I replied:
Dear Dr Berard
Thank you for your reply to my email yesterday.
I am still a little confused and will use some simplified figures to illustrate my confusion:
Suppose we have Group A consisting of 4,000 women who have suffered miscarriage. Of these 200 (5%) had used NSAIDs during the early stages of pregnancy. In the control Group B there are 40,000 women who did not suffer miscarriage. Of these, 800 (2%) used NSAIDS during the early stages of pregnancy.
NSAID use is thus 2.5 times commoner in Group A than in Group B and thus clearly is a risk factor. Is it correct though, to say that NSAID users are two and a half times more likely to suffer miscarriage? Even if the miscarriages of the NSAID users in Group A had all been caused by NSAID use, this still leaves 3800 which were due to other causes. Thus I would have thought that NSAID use made the chances of miscarriage 4000/3800 or a little under 1.053 times more likely.
This time there was no reply. So am I right or am I Dunning-Kruger made manifest?