Sunday, August 31, 2008

Gobbledegook - update

Last month I wrote about a 'pregnancy increases your risk of heart attack three to four fold' story in The Sun, The Star and The Daily Telegraph. Rereading the 2006 study from which the papers took their soundbite the significance of this stood out where it hadn't before:
We found a higher incidence (6.2 versus 2.8 per 100 000 deliveries) and a lower case fatality rate (5.1% versus 7.3%), however, than a recent analysis that used data from a California database.10 Because the California estimate of incidence was based on data from 1991 to 2000, our estimate may reflect improved identification of cases or may reflect a true increase in the number of cases

(sorry for the bolderising, but this is important).
So, according to the study authors, since 2000 it's plausible that identification and reporting of heart attacks during delivery has improved, leading to a higher figure for its incidence. They go into a bit of detail:
Improved identification of cases may have resulted from the advent of the widespread use of troponins, which has likely resulted in the detection of small events that were previously undiagnosed. Increased detection would explain an increase in the incidence of acute myocardial infarction and a decrease in the case fatality rate. Another possible explanation for an increase in the incidence over time is the increased number of births to older women, who may have more cardiac risk factors. Between 2002 and 2003, the birth rate rose 6% for women aged 35 to 39 years and 5% for women aged 40 to 44 years.8 Since 1981, the birth rate for women aged 40 to 44 years has more than doubled.8 Another possible explanation for the higher incidence is better ascertainment of cases in the NIS database.

all of which have me nodding my head in agreement. Two paragraphs later (and they're not very long paragraphs either, so the two statements come pretty close together):
We do not have the incidence of acute myocardial infarction for women who were not pregnant. Petitti et al,14 however, published the incidence of myocardial infarction among reproductive-aged women in a large health maintenance organization. Using the age-specific rates of myocardial infarction per 100 000 women-years derived from that study and applying them to the age distribution of the women in the present study, we would have expected to find 250 myocardial infarctions as opposed to the 859 found. Therefore, the risk of acute myocardial infarction appears to be approximately 3 to 4 times higher in pregnancy.

(the line that had the PR person so excited)(and the newspapers). Clicking on the hyperlink that takes you to the reference section, and the full title of the Petitti et al study mentioned above gives it as Petitti DB, Sidney S, Quesenberry CP Jr, Bernstein A. Incidence of stroke and myocardial infarction in women of reproductive age. Stroke. 1997; 28: 280–283. A visit to the archives of Stroke and a look at line two of the Subjects and Methods section tells us that the study was conducted
May 1, 1991, through August 31, 1994, in northern California and July 15, 1991, through August 31, 1994, in southern California.

Right, so: detection and reporting between 1991 and 2000 may have led to an underestimation of incidence in one paragraph, but two paragraphs later, the same time period is being used to estimate a three to four fold increase in heart attack risk for pregnant women. Dye not reckon that maybe the 859 cases of heart attack in pregnant women found in this study were down to improved detection and reporting, and that the 250 cases in the general female population of 1991 may have been an underestimation? Just saying.
Science and health reporting isn't a job for jobbing reporters, it's a job for investigative journalists. And while I'm at it, it's high time health correspondents abandoned their slavish respect for medical authority and started reading the bleeding studies they trumpet so very loudly.

No comments: