To the Editor: Stone and colleagues must be complimented for their study of gestational diabetes mellitus (GDM) published recently in the Journal.1 Although the study was comprehensive, in that it linked two pertinent records for 99.3% of the women delivering in Victoria in 1996, their analysis was necessarily limited to the parameters recorded in those two data sets.

Stone et al do not refer to relevant recent work from Toronto, Ontario,2 and Sunshine, Victoria,3 which has already determined the maternal risk factors for GDM: increasing age, racial origin, family history of diabetes mellitus, and pre-pregnancy body mass index. Data on the latter two of these four factors were specifically noted as not available to Stone's group, but could have been mentioned as established risk factors to consider in patient management.

The findings of Stone et al regarding age are similar to those previously reported. However, the relative risk for GDM in the Sunshine cohort increased from 25 years of age (odds ratio, 1.9; 95% CI, 1.3–2.7).3 Recasting the all-Victorian data1 with "< 25 years" as the reference datum would most likely produce a similarly significant result to that found in the Sunshine study.

Again, the Victorian data on racial origin are but the Sunshine data writ large. Stone and colleagues had to choose a reference datum for this parameter, but have not avoided a problem that bedevils all such work in the "New World": there is as yet no definitive Australian reference datum available for racial origin. For example, the designation "Australian" may be used, but parents who are themselves second or third generation Australian-born may have, say, pure Maltese ancestry, giving them a high risk of GDM and thus distorting the reference datum. I have argued this case in more detail elsewhere,4 and took measures to circumvent the problem in the Sunshine study.3

Among their findings, Stone et al confirm that there is a significantly increased incidence of macrosomia in GDM-affected infants. I agree. Their work is unquestionably the definitive statement of Victoria's GDM-related macrosomia status (as at 1996), but it lacks one vital ingredient: it defines macrosomia, but does not cite the source of the data used as the study's benchmark. To enable comparisons with their work in future years, could we please have that benchmark referenced so that others may use it too?

In reply: We thank Davey for his comments, which provide us with the opportunity to highlight the benefits and limitations of reports using population-based data. The value of population-based data is that the reported incidence, risk factors and outcomes reflect current practice in the whole of Victoria and are not subject to bias introduced by local referral patterns or clinical practice. Our study1 shows that, in addition to established risk factors for gestational diabetes mellitus (GDM), the reported incidence varies according to hospital size and geographic location, demonstrating the type of bias that can occur. In addition, the large number of subjects in our study (over 60 000) enables more accurate analysis of subgroups.

A limitation, already highlighted in our discussion, is that we are restricted to the parameters available within the data sources used. Davey and Hamblin's article2 demonstrates the difficulty of obtaining individual patient data on body mass index, racial grouping, and family history of diabetes. Even working at the hospital level, they had to extrapolate from population-level data to derive an estimate of these risk factors among the control subjects.2 Given that their study population is a subgroup of ours,1 it is no surprise that the two studies showed similar results.

An important implication for providers of health services is that, with increases in the age at which mothers give birth and in the number of births to Asian-born mothers,3 we predict that the prevalence of GDM in Victoria will rise.

Davey correctly points out that our article does not refer to a relevant 1997 study by the Toronto group.4 However, we do actually refer to a later publication by the same group.5

The problem of ethnicity and migration arises in studies of conditions that are not only polygenic but also a result of complex interactions between a person's genes and his or her environment.

Lastly, the source of information on macrosomia was 1996 population data. We have since produced a percentile chart of weight (g) for gestational age (weeks) based on 15 years of Victorian data.6