Retraction of paper on maternal diabetes

Editor—In a paper in the BMJ three of us (GH, LMI, and RTL) claimed that the excess of birth defects and perinatal deaths among births in women with pregestational diabetes was significantly lower in Norway compared with the northeast of England.¹ The analysis of the data from the Medical Birth Registry of Norway was, however, wrong and led to a wrong conclusion.

There are codes in the registry for different types of diabetes. By mistake Norwegian data included codes for gestational diabetes in the diabetes group, although it was intended to exclude those. We failed to realise the mistake even when the possibility was raised in correspondence in the journal.²

After the error was discovered and discussed at a joint meeting in Newcastle,³ one of us (NM), a clinician, reviewed the clinical records of 696 mothers with a code for diabetes in the Norwegian registry. When the definition in the registry was restricted to one code of diabetes noted before pregnancy, 80% (206 of 257) were confirmed to have had pregestational diabetes. The rate of false negative diagnoses in the registry by this definition is not known.

Using this narrow definition, and including births with an estimated gestational age above 24 weeks, we performed a reanalysis of our data. Risks among children of women with pregestational diabetes in northeast England and Norway are shown in the table.

Table 1.

Perinatal mortality and birth defects in babies born to mothers with and without pregestational diabetes, 1 July 1994 to 30 June 1997 in northeast England and Norway

	No affected	No not affected	Rate per 1000	Odds ratio (95% CI)*
Perinatal mortality†
Northeast England:
Maternal diabetes (n=304)	13	291	42.8	4.4 (2.3 to 7.6)
No maternal diabetes (n=181 132)	1014	100 502	10.0
Norway:
Maternal diabetes (n=704)	11	693	15.6	2.2 (1.1 to 3.9)
No maternal diabetes (n=181 132)	1312	179 820	7.2
Birth defects†
Northeast England:
Maternal diabetes (n=309)	17	292	55.0	2.3 (1.3 to 3.8)
No maternal diabetes (n=101 755)	2472	99 285	24.3
Norway:
Maternal diabetes (n=704)	19	685	27.0	1.3 (0.8 to 2.1)
No maternal diabetes (n=181 132)	3633	177 499	20.0

^*Adjusted for maternal age with exact 95% confidence intervals.

^†Difference between odds ratios adjusted for maternal age in northeast England and Norway was not significant (P=0.10).

Although there is a tendency for higher risks in England, the differences in odds ratios are not significant. Adjustment for a rate of 20% of false positive diagnoses of diabetes increases the tabulated odds ratios in Norway from 2.2 to 2.5 for perinatal death and from 1.3 to 1.4 for birth defects, which further reduces the estimated differences in data between the two countries. We believe that closer involvement of clinicians in both countries may improve the validity and usefulness of similar comparisons in the future.