This study developed risk equations from healthy (NGT) to pre-diabetic states and from pre-diabetes states to T2D using data of a Swedish population. In total, six risk equations were developed and validated. The equations can be used to identify individuals with increased risk to develop any of the three pre-diabetic states or T2D. In addition, the equations are useful for adapting prevention strategies to specific risk profiles. Risk models are widely used in clinical and public health practice . The six risk equations used in this study describe the risk of developing a pre-diabetic state from being healthy as well as developing T2D depending on modifiable and non-modifiable risk factors. The risk equations allow adjustment to a specific risk profile and thus give more precise risk estimates than general risk models.
We found that 49% of those with IFG&IGT at baseline developed T2D in comparison to 3%, 14% and 17% for those with NGT, IFG and IGT (Figure 1). Other prospective studies have also found that a combination of IFG and IGT increases the risk of developing T2D compared to subjects having either of the glycemic abnormalities [9, 21–23]. For example, de Vegt and colleagues estimated that the ORs for T2D were 10.1, 10.9 and 39.5 for those having IFG, IGT and IFG&IGT, respectively .
From the results of the logistic regression, it seems that the variables snus and sex as well as the variables education and smoking might show multicollinearity. However, we have examined the influence of each pair of risk factors and could not find that this first impression was true. All four potential risk factors were hence kept in the logistic regression estimations. We also compared whether OR coefficients and their equivalent bootstrap results in all logistic regression models were significantly different from 1. Besides the variables age and triglyceride in the transition from NGT to IFG as well as the variable “five a day” in the transitions from NGT to IFG&IGT and from IGT to T2D, all other variables could be validated with the bootstrap estimations.
In our study, sex had different influence on the progression to a pre-diabetic state. Whereas men have a higher risk to develop IFG, women have higher risk to develop IGT. As expected, the progression from NGT to IGT and/or IFG exhibits striking sex differences [24–26]. In most populations, IFG is substantially more common amongst men and IGT is slightly more common amongst women [26, 27]. In a study from Turkey, however, IFG and IGT were more common in women than in men . Meigs and colleagues reported that men in comparison to women had a higher risk to progress from NGT to IFG and/or IGT in the United States .
As expected, increasing age increases the likelihood to develop a worse diabetic state. Even though age is a non-modifiable risk factor, it needs to be included in all risk equations.
Our data suggest that lower education increases the risk to develop IFG from NGT, even tough the confidence interval of the odds ratio is quite close to one (non-significance). Nonetheless, education might be an important player in prevention of a pre-diabetic state.
Self-rated health was excluded from all six risk equations indicating that it did not add to the models. However, T2D is known to be related to low self-rated health [29, 30]. However, our results might be due to a small sample size for the equation to or from IFG&IGT or due to difficulties in measuring self-rated health.
As expected, high triglyceride levels, high blood pressure and high BMI are the strongest factors for a progression to a worse diabetic state. All are well-known risk factors for the development of T2D. In a study by Jauch-Chara and colleagues, low body weight was associated with increased risk to develop IGT from NGT . Underweight could not be examined separately in our study but was combined with normal weight due to the low number of cases in the underweight category. The influence of low body weight could thus not be estimated with our data. Our risk models assume a linear relationship looking only at increased body weight. In another study, BMI and waist circumference were higher in subjects with abnormalities of glucose metabolism compared to NGT . A study from the United States also found that higher BMI increased the rate to progress from NGT to IFG and/or IGT .
The odds ratio of smoking was above 1 for the development of NGT to IFG and below 1 for the development of NGT to IGT. In a study with American Indians, participants with pre-diabetes reported significantly less smoking than participants with NGT and were significantly more likely to be past smokers . However, in our study smokers and past-smokers were relatively evenly distributed at first examination. Our population included 24%, 29%, 19%, 21% and 29% smokers and 29%, 31%, 30%, 37% and 29% past-smokers among NGT, IFG, IGT, IFG&IGT and T2D respectively. In addition, smoking is related to lower BMI. Smoking prevalence has decreased and prevalence of high BMI has increased over time in this population. Individuals must have, therefore, quit smoking between baseline and follow-up. Smoking cessation might lead to an increased in weight and BMI . Possibly, BMI could also be the explanatory factor here.
Lower level of physical activity (vs. higher level) slightly increased the risk to develop IGT from NGT. This is consistent with the literature [4, 5].
The odds ratios of snus, “five a day” and marital status were all not significant. We aimed to describe diet with one simple variable in our model. However, the question what is healthy diet is difficult to answer. The purpose of the variable we created was that it needed to be simple and easy to understand. We decided to use the consumption of five portions of fruits and vegetables a day as a proxy of healthy diet, knowing that this is a simplification of reality. Diet is far more complex.
Marital status did not have any significant impact on the development of a worsened glucose status. It was only included in the model IFG to T2D but could not reach statistical significance.
In our population, individuals with a family history of diabetes developed a worsened glucose status more likely than those without a family history of diabetes. This factor was only excluded in the development from IGT to T2D. An evaluation of the Stockholm Diabetes Prevention Program, however, demonstrated that prevalence of IFG, similarly to the prevalence of IGT, IFG and IGT combined and T2D, was nearly twofold higher in those who had a family history of T2D compared to those without family history of T2D . It needs to be kept in mind that knowledge about family medical status and age of respondent might have been important influences on whether the study participant reported a family history of T2D. For example, it has been shown in our VIP population that knowledge about family history is rather low, in particular among younger men . In addition, parents of young study participants might not have been diagnosed with T2D yet .
In consequence, the influence of specific risk factors on the transition to worse states towards the development of T2D is diverse. Different risk factors have different impacts on the development of IFG, IGT, IFG&IGT and T2D.
Use of results in practice
Once glucose status has been estimated, information used to perform risk equations are relatively easy to obtain, for example age, smoking status or measurement of BMI. For the classification of the glucose status, an oral glucose tolerance test is needed. This need of a test is a challenge, because in a practical setting the individual or their physician rarely knows the patient’s glucose status. The advantage of our risk equations over similar risk tools is that four different glucose states can be represented, establishing six unique risk equations . Other risk equations focused on the development of T2D only [35–37].
As Noble and colleagues pointed out, caution is needed when extrapolating risk models to a different population . The models, therefore, best describe the Swedish population. Risk equations should be evaluated to determine whether they are also valid in other populations and in other prospective cohorts.
Comparing our results without looking at risk factors leads us back to Figure 1. Among individuals with IFG or IGT at baseline, 14% and 17%, respectively, developed T2D 10 years later. For those with a combination of IFG and IGT the risk was much higher. Almost half (49%) of our population with IFG&IGT at baseline developed T2D within 10 years. Our overall one-year risks for T2D (estimated based on 10-year changes within our cohort ) were 1.5%, 1.8% and 6.5% for IFG, IGT and IFG&IGT respectively.
In other studies, the annual progression rates to T2D were 1-5% for individuals with IFG and 3-11% for those with IGT [21, 23, 39, 40]. The Hoorn study estimated that the annual rate of developing T2D from IFG alone was 5.5%, and from IFG together with IGT 10.8% . The Paris Prospective Study reported a much lower annual rate, with 1% for individuals with IFG and 6% for individuals with IFG and IGT . An Italian study estimated an annual rate of 0.8% to develop diabetes from IFG alone and an annual rate of 3.9% to develop diabetes from a combination of IFG and IGT . A study from Iran showed that patients with first-degree relatives with T2D have a risk of 8.6% to progress to IFG and a risk of 3.7% to progress to IGT . Our results are comparable with the Paris Prospective Study and the Italian study. Nonetheless, these different risks point out that the risk profile of different populations is quite diverse. We examined a Swedish population from a population-based perspective, meaning that we did not aim at any high-risk profile population but intended to investigate the general public. This might be one reason why progression rates in our study are rather low in comparison with other studies. The highest risk to develop T2D was presented by combined IFG and IGT . Further, VIP participants have participated in interventions that aimed at the reduction of T2D and cardiovascular disease. They had experienced motivational counseling regarding life style modification. This might have contributed to comparable low rates of progression.
Even though a high number of individuals could be enclosed in this modeling study, some risk equations only included a small number of study participants, such as in the risk equation from IFG&IGT to T2D. This definitely hampered the ability to depict variables as potential risk factors.
Due to the design of the VIP, the number of people with T2D is likely to be underestimated. We included only panel data with information at baseline and 10 years follow-up. If a person was diagnosed with T2D at first or between first and second examination, he or she was less likely to participate in the VIP or in the second VIP examination. As a consequence, more people can be expected to be in T2D at baseline and at follow-up. However, any change in status between first and second examination of those individuals who participated twice were caught by the DiabNorth register. Also, our risk equations only consider worsening of the glucose state, not the reverse direction towards a less deteriorated glucose state.
In addition, we could only calculate risk over a 10-year period. Many events can happen during such a long time period. For example, individuals who would have developed any pre-diabetic state after some years could have progressed to T2D within the 10 years or someone has been in a pre-diabetic state and returned to NGT after 10 years. Also, a substantial number of individuals with IFG or IGT revert back to NGT . These changes between the states within the 10-year time frame could not be traced in our study, unless the participant was included in the diabetes registry. The progression to another state, nonetheless, takes many years. Meigs and colleagues suggest that subjects with IFG and IGT are already close to transitioning to T2D and underline that T2D develops slowly over many years, transitioning through a prolonged state of impaired glycemia . Also, individuals who were registered in the DiabNorth register would have been traced and re-sorted according to information in the DiabNorth.