Introduction
Indigenous peoples worldwide are experiencing an epidemic of type 2 diabetes (T2DM)Reference Chen, Magliano and Zimmet 1 , Reference Yu and Zinman 2 and its complications.Reference Naqshbandi, Harris, Esler and Antwi-Nsiah 3 While genetic factors likely play a role in the disproportionate emergence of this chronic disease among Indigenous populations,Reference Barroso 4 , Reference McCarthy 5 the rapid appearance of T2DM since the end of the Second World War attests to the importance of environmental factors, most notably the rise in obesity rates, that parallel unprecedented changes in caloric intake and physical activity.Reference McCarthy 5 In addition, prenatal factors related to intrauterine exposures and stressors are also increasingly implicated in an inter-generational amplification of T2DM risk.Reference Whincup, Kaye and Owen 6 , Reference Dabalea and Pettitt 7
Both low birth weight (LBW) and high birth weight (HBW) are associated with an increased risk for T2DM, but a systematic review carried out in 2008 showed that the most consistent relationship was between LBW and T2DM.Reference Whincup, Kaye and Owen 6 This has been reported primarily in middle aged and older Europeans, and also in Chinese, Japanese and South Asian populations.Reference Whincup, Kaye and Owen 6 With respect to HBW and T2DM, the above review found that the observed relationship was largely due to findings within two North American Indigenous populations, the Pima Indians of Arizona and First Nations peoples from Saskatchewan (see below). A recent PubMed search revealed only two additional studies since 2008 reporting on a significant relationship between HBW and T2DM in adult cohorts from Denmark and Sweden.Reference Johnsson, Haglund, Ahlsson and Gustafsson 8 , Reference Zimmermann, Gamborg, Sorensen and Baker 9
In 2001, we published the results of a case-control study in which we compared birth weights of diabetic First Nations and non-First Nations people with those of their age- and sex-matched counterparts who did not have diabetes.Reference Dyck, Klomp and Tan 10 A significantly larger proportion of First Nations individuals with diabetes had HBW (>4000 g) than we observed in the other three groups. We coined the term ‘hefty fetal phenotype’ as a descriptor for macrosomic neonates at risk for the later development of T2DM and speculated that they were primarily the offspring of diabetic mothers (ODM). We further hypothesized that the relationship between gestational diabetes (GDM), macrosomia and increased T2DM risk among ODM was a significant factor in the epidemic of T2DM among First Nations people resulting from a currently maladaptive chain of events superimposed on the physiologic insulin resistance of normal pregnancies. With the recent acquisition of a multi-decade maternal–infant database (MIDB) in Saskatchewan, the purpose of this study was to re-examine the relationship between birth weight and the subsequent development of T2DM, as well as GDM, among as many of the female cases and controls as possible from our original study. By conducting a survival analysis rather than a case-control study, and by employing a more rigorous algorithm to identify diabetes cases from health administrative data, we aimed to determine if the results of our original study would be confirmed and possibly expanded.
Methods
Source population and health care system administrative data
Saskatchewan is a mid-western Canadian province with a stable population of approximately one million people for several decades. About 99% of the total population are beneficiaries of a universal health care system and their health care information is housed within Ministry of Health administrative databases.Reference Downey, Stang, Beck, Osei and Nichol 11 Three groups of Indigenous people are constitutionally recognized in Canada under the descriptor Aboriginal: First Nations people who are indigenous to Canada, Metis people who are of mixed European and First Nations heritage, and Inuit people who originate from the arctic regions of Canada. Although very few Inuit live in Saskatchewan, about 10–12% of the total provincial population are First Nations people and 5–6% are Metis. For this and the previous study on which this research are based,Reference Dyck, Klomp and Tan 10 we sub-divided the provincial population into First Nations people who are identifiable within Ministry of Health databases as individuals registered under Section 6 of the Indian Act of Canada (Registered Indians), and non-First Nations people who are mainly of European origin but also include Metis. Both the current and previous study described below were approved by the University of Saskatchewan Ethics Review Board (ERB number 11–95).
Summary of 2001 birth weight/diabetes case-control study
Our original 2001 study aimed to determine the relationship between birth weight and T2DM.Reference Dyck, Klomp and Tan 10 We used a 1:1 age- and sex-matched case-control design in which the cases were First Nations people with diabetes and the three control groups were First Nations people without diabetes, and non-First Nations people with and without diabetes. Each group was comprised of 1366 individuals who were identified in 1996 by linking Saskatchewan Ministry of Health administrative databases. People with diabetes were those with at least one service code for diabetes (ICD-9-250) and/or diabetic retinopathy (ICD-362) from the physicians services database between January 1, 1991 and December 31, 1994. Age- and sex-matched individuals without diabetes were selected from the Health Insurance Registration file.
Birth registration information was sought for all 5464 subjects (1366×4 groups of cases and controls) born between January 1, 1950 (birth weight was not routinely collected before then) and January 31, 1984. For the final analysis, birth weight was available for 3992 individuals (2557 females) including 846 First Nations people with diabetes (554 females), 882 First Nations people without diabetes (571 females), 1164 non-First Nations people with diabetes (737 females) and 1100 non-First Nations people without diabetes (695 females).
Design of current study
For this paper, we conducted a survival analysis to investigate the relationship between birth weight and time to occurrence of first episode of GDM and diagnosis of T2DM. Only female subjects from our 2001 studyReference Dyck, Klomp and Tan 10 were evaluated (n=2557), as well as a small and partially overlapping group of women from a birth weight and end stage renal disease (ESRD) case-control study (n=~400) that was published in 2003.Reference Dyck, Klomp, Tan and Stang 12 Men were not included in the current study because of the GDM focus. All women from the above two studies were re-identified within a larger maternal cohort which has been compiled to examine the impact of GDM on the intra- and inter-generational risk for T2DM (yet unpublished). This larger cohort, contained in a MIDB, includes all mothers eligible for Saskatchewan Health Benefits between January 1, 1980 and December 31, 2009 and is linked with all live born or stillborn babies delivered during that time period. All 174,273 mothers (by diabetes in pregnancy status) and their offspring have been followed up to March 31, 2013 for the development of T2DM and, in the case of female offspring (n=176,867), the occurrence of GDM.
For the purposes of this current study, Ministry of Health personnel linked birth registration information (birth weight, gestational age, maternal parity and age, and previous sibling stillbirths) from the births of the women included in the 2001 case-control study (and the birth weight/ESRD study) with the same women who were in the current 1980–2009 MIDB. Each group of First Nations and non-First Nations women were sub-divided by birth weight: LBW (<2500 g), normal birth weight (NBW) (2500–4000 g) and HBW (>4000 g). They were then followed for up to 60+ years for the development of T2DM and/or GDM.
Using hospital separation and physicians services records, we identified all women who met the case definition for diabetes using a validated algorithmReference Hux, Ivis, Flintoft and Bica 13 that required at least one inpatient hospital service record with a diagnostic code of ICD-8 or ICD-9 250.x or ICD-10-CA E10-E14.xxx OR two or more physician service claims on different days with a diagnosis of ICD-8 250 or ICD-9 250 within any 730-day period OR an above physician service claim followed by an above hospital service record within any 730-day period. Diabetes records related to a gestational event were excluded (see below) to ensure that GDM cases were not counted as diabetes cases. The diabetes case date (diabetes incident date) was the date of the first record of a diabetes code.
All women were also assessed for the occurrence of GDM, which could occur more than once, but not after an individual developed diabetes if diagnosed. Women were deemed to have met the case definition for GDMReference Feig, Zinman, Wang and Hux 14 if, within 150 days preceding or 90 days following a delivery date, they had one or more inpatient hospital separation records with a diagnostic code of ICD-9 250.x, 648.0 or 648.8 or ICD-10-CA E10-E14.xxx or O24.xxx OR one or more physician service claims on different days with a diagnosis of ICD-8 250 or ICD-9 250. The GDM case date was the delivery date.
Statistical analysis
We determined the number of women by ethnic group and the three birth weight categories. We then calculated the proportion who ever developed diabetes and/or GDM. Differences between the two ethnic groups and birth weight categories were evaluated using the χ2 test.
Using Cox regression, we carried out time to event (or survival) analysis for First Nations women and non-First Nations women for whom the outcome was either diabetes or GDM. Women in each ethnic group were divided into the three birth weight groups (LBW, NBW and HBW) and followed from birth (as early as 1950) to either death, loss of health care coverage (e.g. moving to another province or country) or end of study (March 31, 2013). Overall and ethnicity-specific hazard ratios (HRs) were calculated for HBW and LBW using NBW as the reference.
The level of significance was set at P<0.05 (two-tailed). Statistical analyses were conducted using SPSS version 24 (IBM SPSS Statistics for Windows; IBM Corp., Armonk, NY, 2015).
Results
Of the 174,273 (17,115 First Nations; 157,158 non-First Nations) mothers in the 1980–2010 Saskatchewan MIDB. In total, 2003 subjects had been in the birth weight/diabetes and birth weight/ESRD case-control studies, from which birth weight information was available. This represented about 2/3 of a maximum of 3000 women who had been in the original studies (2557 in the birth weight/diabetes study and about 400 in the birth weight/ESRD study.
Table 1 shows study populations by ethnicity, birth weight and diabetes outcomes. Among 971 First Nations women, 384 (39.5%) developed diabetes and 80 (8.2%) had at least one episode of GDM. Among 1032 non-First nations women, 302 (29.3%) developed diabetes and 95 (9.2%) had an episode of GDM. Over half (50.8%) of First Nations women born with HBW developed diabetes compared with 37.7% of non-First nations women (P=0.082), whereas 44.9% of First Nations women compared with 30.7% of non-First Nations women born with LBW developed diabetes (P=0.096). Both First Nations and non-First Nations women with HBW were more likely to develop GDM than those with NBW or LBW, but the difference between the two ethnic groups was not significant (P=0.379).
Table 1 Study populations by ethnicity, birth weight and diabetes outcomes
LBW, low birth weight; NBW, normal birth weight; HBW, high birth weight.
Table 1 also shows that, within ethnic groups, those with LBW were not more likely to develop diabetes or GDM than those with NBW. However, a significantly larger proportion of First Nations women with HBW developed diabetes compared with those with NBW (50.8 v. 37.6%; P=0.006). Similarly, First Nations women with HBW were also more likely to develop GDM but the difference was not quite significant (12.7 v. 7.6%; P=0.059). Among non-First Nations women, a larger proportion of those with HBW also developed diabetes (37.7 v. 28.5%; P=0.105) and GDM (17.4 v. 8.5%; P=0.013) compared with those with NBW, but only the GDM findings were statistically significant.
Table 2 summarizes the results of Cox regression analysis and shows the HRs of developing diabetes by ethnicity and birth weight. Overall, there was an increased risk of developing both diabetes and GDM among First Nations and non-First Nations women who were born with HBW. This was primarily due to the findings among First Nations women who were almost 60% more likely to develop diabetes [HR 1.568; 95% confidence interval (CI): 1.188, 2.069] and almost 50% more likely to develop GDM (HR 1.468; 95% CI: 1.016, 2.121) if they were born with HBW rather than NBW. Being born with HBW did not have a significant impact on diabetes or GDM risk among non-First Nations women but those with LBW experienced a higher risk for GDM.
Table 2 Hazard ratios (HR) of developing diabetes and gestational diabetes by ethnicity and birth weightFootnote a
Bold indicates significant P values.
CI, confidence interval.
a Reference: Normal Birth Weight.
Figure 1 shows the cumulative survival of women until development of diabetes by birth weight and ethnicity. For both First Nations and non-First Nations women, those born with HBW experienced the greatest risk for diabetes followed by those who were born with LBW. By age 60, almost 70% of First Nations women with HBW had developed diabetes compared to about 50% of non-First Nations women.
Fig. 1 Cumulative survival of women until development of diabetes by birth weight and ethnicity.
Figure 2 shows the cumulative survival of women until development of first episode of GDM by birth weight and ethnicity. Once again, both First Nations and non-First Nations women born with HBW experienced the greatest risk for GDM followed closely by those who were born with LBW. By age 40+, almost 30% of First Nations women with HBW had developed GDM compared to just over 20% of non-First Nations women.
Fig. 2 Cumulative survival of women until development of first episode of gestational diabetes by birth weight and ethnicity.
Discussion
Using survival analysis techniques and a more rigorous and validated algorithmReference Hux, Ivis, Flintoft and Bica 13 for identifying diabetes cases in health administrative databases, we have now confirmed our previous case-control findingsReference Dyck, Klomp and Tan 10 that Saskatchewan First Nations women (specifically mothers) who were born with HBW were significantly more likely to develop diabetes than those with NBW or LBW when followed for up to 60 years. Although there was a trend for non-First Nations women to demonstrate the same findings, the differences between birth weight groups were not significant. In addition to the diabetes findings, we have now shown for the first time that First Nations mothers born with HBW are also at increased risk of developing GDM when followed for up to 40+ years. Once again, there was a similar but not significant association for non-First Nations mothers born with HBW. Finally, Cox regression analysis showed a consistent trend for LBW mothers of both ethnic groups to be more likely to develop diabetes or GDM compared to those with NBW. However, the only relationship that was significant in this latter analysis was between LBW and GDM among non-First Nations mothers.
Because of the manner in which subjects were selected for this study from two previous case-control studies, the proportions of women who developed GDM and diabetes in the current analysis are higher (particularly for non-First Nations women) than those found in the total populations of First Nations and non-First Nations women in Saskatchewan from the same time period (see limitations below).
A systematic review of birth weight and T2DM showed that the most commonly reported relationship was between LBW and T2DM.Reference Whincup, Kaye and Owen 6 Hales and Barker have suggested that this is due to a ‘thrifty phenotype’ undergoing a maladaptive response to fetal undernutrition,Reference Hales and Barker 15 initially mediated through impaired islet cell development with the later development of insulin resistance. Although the findings were not significant, both First Nations and non-First Nations women with LBW in this study exhibited a trend for higher T2DM risk than those with NBW – this is in keeping with this known relationship. Furthermore, one would expect that women with LBW would also have a higher risk for GDM, as we found in non-First Nations women and which has been reported elsewhere.Reference Innes, Byers, Marshall, Barón, Orleans and Hamman 16
Less clear is the relationship between HBW and T2DM. Indeed, in the 2008 systematic review above,Reference Whincup, Kaye and Owen 6 our study of Saskatchewan First Nations people was the only one to show a significant relationship between HBW and T2DM although it had been observed earlier among Pima Indians.Reference Mccance, Pettitt, Hanson, Jacobsson, Knowler and Bennett 17 While more recent publications have also demonstrated this relationship,Reference Johnsson, Haglund, Ahlsson and Gustafsson 8 , Reference Zimmermann, Gamborg, Sorensen and Baker 9 it is somewhat surprising that it has not been more widely reported given the known association between diabetic pregnancies and macrosomia,Reference Cunningham, Lenevo, Bloom, Hauth, Gilstrap and Wenstrom 18 as well as the now widely held view that GDM increases the inter-generational risk for T2DM. Increasing insulin resistance is a physiological response to normal pregnancy that contributes to optimal fetal nutrition.Reference Cunningham, Lenevo, Bloom, Hauth, Gilstrap and Wenstrom 18 Pederson proposed over 60 years ago that if excessive demands for insulin production tipped this normal balance – for example, a woman with pre-pregnancy obesity – it could result in GDM and maternal hyperglycemia, fetal hyperinsulinemia and macrosomia.Reference Pederson 19 Freinkel suggested that this could contribute to the later development of insulin resistance in the ODM.Reference Freinkel 20
Building upon the above themes, the seminal findings among Pima Indians on the inter-generational impact of diabetic pregnancies on T2DM risk for ODM,Reference Dabalea and Pettitt 7 and our own observations of the HBW and T2DM relationship in particular,Reference Dyck, Klomp and Tan 10 there are two main elements to the hefty fetal phenotype hypothesis. The first is that GDM, and its effect on the fetus and their later risk for T2DM, is the consequence of a maladaptive chain of events triggered by maternal adiposity superimposed on the physiologic insulin resistance of normal pregnancies. The second element is that this phenomenon has become the overriding intrauterine factor in the pathogenesis and epidemic of T2DM among Indigenous peoples in Canada. To summarize the rationale for the latter, First Nations women have higher GDM rates than non-Indigenous women in Canada; this is at least partly due to an interaction of pre-gravid obesity with Indigenous heritage.Reference Dyck, Klomp, Tan, Turnell and Boctor 21 Elevated GDM rates preceded the significant appearance of T2DM in remote Saskatchewan First Nations communities in the early 1990sReference Dyck, Tan and Hoeppner 22 and increasing HBW ratesReference Dyck and Tan 23 have paralleled rising GDM and T2DM incidence over several decades.Reference Aljohani, Rempel and Ludwig 24 , Reference Dyck, Osgood, Lin, Gao and Stang 25 Finally, First Nations females demonstrate a stronger relationship between HBW and T2DM than males,Reference Dyck, Klomp and Tan 10 and we have now shown that HBW appears to predispose to GDM as well. Thus, both the inter- and intra-generational impact of GDM on the overall epidemic of T2DM among First Nations peoples is substantialReference Osgood, Dyck and Grassmann 26 and is likely a key factor in the disproportionately higher incidence of T2DM in First Nations females compared with males.Reference Dyck, Osgood, Lin, Gao and Stang 25
The most important limitation of this research is that our study populations originate from two earlier case-control studies,Reference Dyck, Klomp and Tan 10 , Reference Dyck, Klomp, Tan and Stang 12 the first of which selected subjects born between 1950 and 1984 based on their diabetes status (yes/no) from 1991 to 1994. Therefore, the overall proportions of women with diabetes and GDM in this study exceed their population-wide values and this is particularly so for non-First Nations women who experience markedly lower diabetesReference Dyck, Osgood, Lin, Gao and Stang 25 and GDMReference Dyck, Klomp, Tan, Turnell and Boctor 21 incidence than First Nations women. Nonetheless, the First Nations diabetes cases from the 2001 case-control studyReference Dyck, Klomp and Tan 10 represented all 1991–1994 prevalent diabetes cases born from 1950 to 1984 and the three control groups were randomly selected age- and sex-matched individuals. Thus, the birth weight distributions are likely a true reflection of those of diabetic and non-diabetic females in the total First Nations and non-First Nations populations during the period in question. Other limitations include an inability to differentiate between type 1 diabetes and T2DM from administrative data. However, over 90–95% of diabetes cases in adults are T2DM so our findings are primarily a result of the impact of birth weight on T2DM (and GDM) risk. Finally, we could not identify Metis within the non-First Nations population so our findings probably minimize true differences between Indigenous and non-Indigenous people in Saskatchewan.
The main strengths of the study include its longitudinal design and our ability to follow birth cohorts for up to 60 years for development of diabetes. In contrast to our 2001 case-control study, we used a validated and more rigorous algorithm for the identification of diabetes in health care system administrative data. Finally, we were able to sub-divide the Saskatchewan population by ethnicity into First Nations and non-First Nations women.
Conclusions
First Nations mothers born with HBW are at increased risk of developing both GDM and T2DM. This is likely due to exposure to GDM or gestational impaired glucose tolerance in their own mothers who in turn may well have been born with LBW. Poor maternal/fetal nutrition was commonplace in Saskatchewan as late as the 1970sReference Eduard, Gillis and Habbick 27 and has been related in part to Residential School attendance.Reference Hackett, Abonyi and Dyck 28 , Reference Mosby and Galloway 29 We suggest that this inter-generational amplification of T2DM risk mediated through prenatal exposures plays a substantial role in the epidemic of T2DM among First Nations peoples. Furthermore, with global increases in levels of obesity,Reference McCarthy 5 the phenomenon that we are observing among First Nations peoples – HBW has become a more important influence on T2DM risk than LBW – may be a harbinger of similar patterns to come in other populations.
Acknowledgments
The authors wish to thank personnel from the Ministry of Health, province of Saskatchewan, for preparing the database files on which this study is based.
Financial Support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Conflicts of Interest
None.
