Our main result is that nearly one in five of women with various risk factor for DM and normal OGTT in pregnancy were diagnosed with manifest diabetes. This is more than three times the expected rate of 6 % registered in the Danish population at 50-60 years of age (12). As GDM was excluded during pregnancy the high incidence may come as a surprise for these women. A similar but shorter study found almost 6 % had progressed to type 2 diabetes with a follow up period of 10 years in women where GDM was excluded (13). When extrapolated further up to nearly 30 years in our follow-up, this result matches the incidence we found. Another long-term study with 23 years of follow-up, however, did only diagnose type 2 diabetes in 5.5 % of women after normal pregnancy (14). An Iranian study had 15 years of follow up study and found a yearly progression rate of 0.4 % per year with a mean age at follow-up of 36. This translates into 12 % after 30 years when women were at their mid- to end-of-forties years of age (15). They could report that family history and BMI were significant risk factors when repeatedly checking up on the women. The only significant hint of what may lie behind is that fasting glucose stayed with a strong association to later development of DM when adjusted for covariates, similar to our findings. An important caveat in comparison, is of course the different ethnicities and DM prevalence in the discussed studies (13-15). The point is that even in a Scandinavian population with low incidence of type 2 DM, the risk of manifest DM is also a potential outcome when classical risk factors are present.
We find that our high incidence rate of DM is an overlooked issue; nevertheless, this is validated in the registries ascertaining the diagnosis. Our long-term follow-up is resonated in recent systematic reviews and meta-analysis that pointed at the length studies; even shorter studies (<5 years) of DM after GDM showed higher rates just after pregnancy than comparing to those with longer follow-up. It is also known that a history of GDM leads to faster development to DM compared to those with a family history of DM and obesity alone. The challenge of the late debut is to look for DM after normo-glycemic pregnancies, which was the aim of this study (16).
The number of pathological OGTT threshold values also has an impact on developing type 1 or type 2 diabetes. The probability of remaining diabetes-free decreases linearly even with one pathological value when followed for more than 15 years (14). In our study one single abnormal value (i.e. borderline OGTT) increased the risk of DM profoundly together with known risk factors when looking at DM after pregnancy with risk factors for DM and differentiating on borderline OGTT. Herath et al showed that maternal age ≥ 30 years and neonatal weight above 3.5 kg increased the risk for later DM but not so family history, previous GDM or parity (13). Abnormal fasting glucose, which is suggestive of a beta cell dysfunction, had high risk of later being diagnosed with type 2 diabetes, similar to our findings. Inactive lifestyle with smoking and weight gains adds more to peripheral glucose intolerance and is rightfully blamed for diabetes epidemic worldwide. However, early testing by fasting glucose has the strongest predictive value and not necessarily with diagnosed GDM. This may help to explain why some of these apparently normal women, but with risk factors for DM, later are diagnosed with DM (14). At follow-up, the fasting glucose during pregnancy was lower in the normal OGTT group without DM and HbA1c similar compared to the other groups, i.e. the borderline and manifest DM group (Table 4). This indicates what is in wait for the borderline group as the fasting glucose is an indicator for beta cell function ‘at rest’ like a thermostat set at a particular homeostasis level, while HbA1c is a mirror of fluctuating glucose levels over days and weeks. The slightly higher fasting glucose indicates a higher risk of manifest DM.
As some of the women had previous GDM as indication for screening with OGTT this could cause some misclassification (Table 1); however, we decided not to exclude them from the cohort, as they were screened diabetes-negative at OGTT in the index period and had similar risk of DM displaying the variation on this issue. The other morbidities present at follow-up point at a potential concern of the endocrine health in these women. However, we have no record on pre-pregnant morbidity like hypertension or obstructive sleep apnea, which are associated with the development of diabetes. While the prevalence of the latter probably is negligible in our women with a mean of 29 years, the blood pressure may exert an impact as a risk factor. The prevalence of 35 to 77% cardiovascular disorders indicates that this is in fact an issue.
The major strengths of this study are a long follow-up and few missed cases. Virtually no DM was detected in the first 10 years of follow-up after borderline OGTT and almost 15 years after the initial screen-negative OGTT (Fig 2), both of which may be due to detection bias in a low-to medium risk population. Correspondingly, studies with shorter follow-up period show higher incidences after normal pregnancies (13). Even though the diagnosis itself was ascertained in repeated blood analyses and medication we cannot rule out that a certain delay has occurred, either in registration or diagnostic work-up as our results depend on the electronic medical charts and registries with no patient contact. We presume that this delay is a systematic occurrence in these women with risk factors after a normal pregnancy. When diabetes was first ruled out during pregnancy, any health professional would not expect to apply diagnostic tests for diabetes in women less than 40 years old with ‘just’ risk factors. Additionally, our primary selection of women for the cohort could be affected by several factors. The delivery rate in the years 1990/91 was 6000 and 406 women (6.8%) were screened due to risk for diabetes. This resulted in a GDM rate of only 0.9% (n=54), which is a low incidence rate by current standards. This is, in part, due to different diagnostic criteria or fewer risk factors 30 years ago as the most important factor for screening is body weight. Obesity rates have steadily increased in a few past decades and 17 % of adults in Denmark were obese in 2017 (17). Obesity, thus, has more than doubled; up from 7.5 in 1994 and 9.5 % in 2000 (18). This would, too, affect incidence rates of DM, as obesity increased during the follow up period and added new risks, for which we have no data and, thus, cannot substantiate. In comparison to this mega-trend in the Western world, other risk factors seem less important, like the diagnostic criteria, efficacy in the diagnostic work-up, breast feeding, smoking and life style changes. One tenth of the women did not have a second OGTT to confirm that they did not have GDM despite having diabetes risk factors. Thus, in these women GDM may have been overlooked and biased the follow-up with an increased the risk of manifest DM. The issue of non-compliance within screening programs is a well-known, even to-day when screen positive rates are 3-5 times higher (1,2,6,10,11,13).
In conclusion, the incidence of diabetes is higher in women with a previous normal OGTT in pregnancy compared to the background incidence. Follow-up on women with risk factors for diabetes, even after a pregnancy without GDM, may identify women with manifest diabetes earlier to prevent possible complications. On this background our results are useful in identifying the time where women may benefit from the effective implementation of evidence-based treatment to postpone and advert manifest DM, even though they had a normal OGTT during pregnancy.