Strengths and weaknesses
To our knowledge, this is the first study specifically performing OGTTs among twins in Sub-Saharan Africa. The fact that it was carried out within a well-established cohort likely improves data quality and ensures good follow-up possibilities.
When performing OGTTs, there is considerable
intra-subject variability [
19]. This affects the precision of our estimates, though the effect would likely be the same for twins and singletons. The majority of participants were investigated using a HemoCue device. Yet, for a small number an AccuCheck apparatus was used. This may have affected comparisons slightly. However, a sub-analysis examining twins and singletons tested with HemoCue and AccuCheck separately did not change the result significantly with regard to the prevalence of dysglycaemia.
For practical reasons, we conducted point of care testing using capillary blood glucose. This could be considered a limitation, since venous plasma glucose is the golden standard of glucose measurement [
30]. According to a WHO report of 2006 capillary blood can, however, be used in under-resourced settings [
30]. The CV of the two glucose measuring devices used of 2–5 % was considered acceptable and was likely less than the
intra-subject variability.
The young age of the participants is also an important limitation as the risk of developing DM, IFG and IGT markedly increases with age. This was a consequence of the age of the overall HDSS. In the context of DM and age, it should be noted that many Type 1 DM patients die at an early age in Guinea-Bissau due to the cost and difficulties of insulin delivery.
Our main statistical analysis was not individually matched on age, but simply compared proportions of IGT in the two groups. However, in a sub-group analysis with exact 1:1 match on date of birth, a similar trend in glucose levels as in the overall sample was found.
Twins and singleton controls were not matched on sex either. Gender differences related to behavior and biology may affect glucose metabolism, particularly in adolescents [
31,
32]. Yet, the sex distribution was very similar for twins and singletons, and inclusion of gender in the adjusted model did not change the estimates significantly.
Information regarding chronic diseases as well as family history of diabetes was based on interviews alone which likely makes the estimates more uncertain. Hence, these parameters were not included in the regression model. The increased attention on DM in Guinea-Bissau is fairly new, and during the field work our impression was that many participants at the time had little or no knowledge of the disease.
Birth weight was only available for 26 % of the participants due to many being born outside the BHP area. Birth weight was also often missing in the HDSS registrations from the 1980s and 1990s, and birth weight could therefore not be introduced in our regression model. Birth weight is a marker of the fetal environment, albeit only a crude one. Gestational age is a likely confounder, since twin pregnancies are known to have shorter gestation than singletons, often resulting in LBW [
33]. Unfortunately, information regarding gestational age was unavailable. Previously, we have found that 41 % of twins in Guinea-Bissau were prematurely born (<37 weeks), compared to 23 % of singletons [
34].
We did not have data regarding insulin sensitivity or beta cell function during the OGTTs. In the previous study, no difference was found in fasting insulin or HOMA insulin resistance between twins and singletons [
26].
Finally, we did not have any data regarding maternal nutrition during pregnancy. This was a limitation, since factors such as maternal BMI and diet are likely of importance in fetal programming [
31,
35].
Consistency with previous findings
A number of investigations have assessed IGT in Africa, though with highly varying estimates, including between urban and rural areas. A study from neighboring Guinea-Conakry observed an IGT prevalence of 13 % in a middle-aged population [
21]. A recent study from Kenya found a quite similar IGT burden of 12 %, though in a considerably younger population with mean age 38.6 years [
22]. An older survey from Cameroon described an urban IGT prevalence of 2 % [
23]. The overall IGT prevalence for the Africa region has been estimated to 7.3 % for adults [
10]. In high income settings, IGT burdens of similar proportions have been observed [
36,
37].
We did not observe a higher IGT burden among twins compared to singletons. This could suggest that LBW in twins – as a marker of adverse fetal conditions - is not strongly associated with IGT in our setting. In this context, it is important to consider whether different mechanisms lead to the LBW in twins vs. singletons. Twins could often be born with LBW due to preterm delivery and spatial limitations in utero, rather than due to an actual adverse fetal environment [
13].
Our previous study found a prevalence of metabolic syndrome of 3.0 % for twins vs. 3.6 % of singletons in the same cohort, supporting that twins are not at higher risk of dysmetabolic disorders later in life. No difference was found in mean HbA1c either. Yet, as in the present investigation, higher fasting glucose and higher IFG prevalence were found in twins [
26]. Adjusting for potential confounders such as age, sex, WHR and season caused the association between twin status and IFG to become even stronger.
Apart from the studies by our group, data on twinning and dysglycaemic disorders in Sub-Saharan Africa are not available, though investigations from high income settings have examined the association. Several smaller clinical studies have shown an association between twin status and dysglycemia [
7,
12]. A Danish study found that MZ twins were particularly disadvantaged in terms of glucose metabolism [
38], as they may experience even more adverse fetal conditions, e.g., due to placental vascular anastomoses [
39]. However, the findings of the smaller clinical investigations have not been replicated in large scale Scandinavian register studies [
13,
14], which found no differences in the DM burden between twins and singletons in adulthood. Furthermore, a study observed no intra-pair differences in glucose metabolism among extremely birth weight discordant MZ twins, indicating common genetic factors in the association between LBW and adult metabolic disease [
40].
Importantly, we do not know how these findings relate to Guinea-Bissau, where twins have a very high perinatal mortality of 22 % [
34]. Infant twin mortality is also twofold elevated, compared to singletons [
41]. Birth weight below 2000 g thus remains an important risk factor for death during the first year of life [
41]. A recent study demonstrated that twins were not hospitalized more often in Guinea-Bissau during infancy, despite elevated mortality [
41]. A considerable “healthy survivor bias” is therefore likely, favoring twins in adulthood.
Also, adverse exposures during fetal life could be markedly different in Africa. Thus, in our setting around 13 % of all children are born with LBW [
26], with twins accounting for 20–23 % of those [
42]. This indicates that sub-optimal nutrition during pregnancy is common in Guinea-Bissau. Of particular concern in Africa is currently the combination of a resource poor fetal environment, followed by a nutrient rich diet later in life due to changes in type of food intake and availability in the populations, as this may predispose to DM and other metabolic disorders [
4]. Data regarding nutritional transition in relation to DM are not available from Guinea-Bissau, but the tendency would likely be similar. Moreover, malaria - which is endemic across Africa - has been associated with both reduced fetal growth and elevated post-natal blood pressure [
16], as has maternal HIV infection [
43], access to antenatal care and socio-economic factors [
44]. Such exposures could influence our glucose results, especially if correlated to twin status. Interestingly, singletons had higher blood pressure than twins.
In the subset of twins where zygosity status was known, MZ twins did not have higher glucose levels than DZ twins. This was somewhat surprising, as MZ twins in theory should experience an even more adverse fetal enviroment. A study from the main maternity ward in Bissau found a MZ twinning rate of 3.4/1000 [
34].
We collected basic information regarding the nutritional intake. We found that singletons had a more frequent intake of fish, potatoes and meat. Twins had a more frequent intake of deep fried and oily food on a weekly basis, whereas singletons had a higher intake on a daily basis. There was no difference in the intake of vegetables, sweets and sweet beverages. Thus, dietary differences were present, though we do not know to what extent this influenced the results (see Additional file
1). Therefore, though we observed significantly higher fasting and postprandial glucose levels for twins, we cannot ascertain whether this related specifically to the fetal environment or maternal and post-natal nutritional patterns.
In relation to the divergent IGT and IFG findings observed, it is important to note that they may have a different etiology. IFG has been related to raised hepatic glucose output and dysfunction in insulin secretion, whereas IGT is associated with peripheral insulin resistance [
20]. Thus, it is possible that twins are at an increased risk of IFG, but not IGT.
Finally, it has been proposed that age is important in unmasking the effect of adverse exposures during fetal life [
12]. As our cohort was young, we cannot exclude that over time regulatory deficits in twins - possibly expressed early by higher fasting and two hour glucose levels - will lead to an increased risk of IGT, and subsequently DM.