Objective The prevalence of obesity and related diseases such as type 2 diabetes mellitus (T2DM) is increasing worldwide, and the Asian Indian population seems to be particularly susceptible to developing T2DM, even at a low body mass index (BMI). In Sweden, the age-adjusted prevalence of diabetes has not increased despite increasing self-reported obesity. However, modern data on the prevalence of obesity and T2DM in Scandinavia are absent.The biochemical links between obesity and subsequent T2DM are unknown, but the adipocyte-derived hormones leptin and adiponectin (adipokines) have been suggested as potential links because they both are related to insulin and glucose physiology. Some studies have found leptin to be an independent predictor of T2DM in men but not in women, although these results are inconsistent. In contrast, adiponectin has more consistently been linked to development of T2DM in both men and women. Furthermore, the leptin–adiponectin ratio may predict incident T2DM better than either of the two hormones separately.The aims of this thesis were to describe time trends in obesity and T2DM in northern Sweden, to evaluate leptin and adiponectin as predictors of deterioration in glucose metabolism including T2DM, and to evaluate leptin as a risk marker regarding ethnic differences, circ-annual variation, and intra-individual stability. Materials and methods Three large population surveys were used, the Northern Sweden MONICA (MONitoring of Trends and Determinants in CArdiovascular Disease) study, the Västerbotten Intervention Programme (VIP), and the Mauritius Non-Communicable Disease Study. Within the MONICA study, six cross-sectional surveys were performed in Sweden’s two northernmost counties, Norrbotten and Västerbotten, between 1986 and 2009. A total of 1000 men and 1000 women ages 25–64 years, also including from 1994 250 men and 250 women ages 65–74 years, were independently chosen for each survey. The overall participation rate was 75%. In 1999, a reinvestigation was performed in 74% of all participants from the three first surveys. Data from the MONICA surveys were used in papers I and IV and data from the reinvestigation survey in paper II. VIP is an ongoing population intervention program that started in the mid-eighties targeting cardiovascular risk factors and has covered the whole county of Västerbotten since 1991. Inhabitants are invited the years they turn 40, 50, and 60 years old, and the annual participation rate has varied between 48% and 67%. A subset (n=1780) from VIP was used in paper II for the circ-annual leptin analysis, and VIP data linked to the diabetes register in Västerbotten (DiabNorr) were used in a case referent study (640 patients with T2DM) in paper III. The Mauritius Non-Communicable Disease Study was performed in 1987 in 10 randomly selected (with probability proportional to size) population clusters. All eligible adults ages 25–74 years were invited, and the participation rate was 86% (n=5083). In 1992, a follow-up survey was performed in 49% of the initial participants. The Mauritius survey data were used in paper II. Results I. BMI increased in men ages 25–74 years and in women ages 25–44 years in northern Sweden between 1986 and 2004. The prevalence of obesity (BMI 30) increased in men ages 25–44 and 55–74 years and in women ages 25–44 years. The prevalence of obesity increased from 10.4% to 19.1% in men and from 12.9% to 17.9% in women ages 25–64 years. Waist circumference (WC) decreased in women of all ages and in men ages 55–64 years between 1986 and 1990. After 1990, WC increased again, and the prevalence of abdominal obesity rose markedly in women ages 25–64 years. II. Differences in circulating levels of leptin, leptin per BMI unit (leptin/BMI), and leptin per cm in WC (leptin/waist) were tested in men and women of Asian Indian, Creole (African), and Caucasian ethnicity. Asian Indian men and women had the highest leptin concentrations and Caucasian men and women the lowest while Creole men and women had intermediate values for leptin, leptin/BMI, and leptin/waist. No circ-annual variation in leptin concentrations was seen in Caucasians. The intra-individual test– retest stability for leptin was equal in men and women of different ethnicities, over 5–13 years, with an intra-class correlation of 0.65–0.82. III. High adiponectin concentrations predicted decreased risk of T2DM in both insulin-sensitive and insulin-resistant men and women, whereas high leptin levels predicted increased risk for T2DM only in insulinsensitive men. A high leptin–adiponectin ratio predicted T2DM in both men and women, and men with a high ratio had a shorter time to diagnosis than those with a low ratio. IV. In northern Sweden, fasting and post-load glucose increased in women ages 24–65 years with 0.2 mmol/l and 0.7 mmol/l, respectively, between 1990 and 2009. Consequently, the prevalence of impaired fasting glucose and impaired glucose tolerance (IGT) rose from 4.5% to 7.7%, and from 7.8% to 14.5%, respectively. In men, post-load glucose increased at 0.5 mmol/l, and the prevalence of IGT rose from 3.5% to 10.1%. The prevalence of diabetes did not increase. An independent relationship between leptin and changes in fasting and post-load glucose was seen in men but not in women. Conclusion An increasing obesity and concomitant deterioration in glucose metabolism was seen in northern Sweden in the period studied. High adiponectin concentrations predicted a decreased risk of T2DM in both men and women, whereas high leptin concentrations predicted an increase in fasting and post-load glucose as well as an increased risk of T2DM in men but not in women. Individual insulin resistance status modified the association between leptin and T2DM, and the leptin–adiponectin ratio may add further predictive information beyond the measures of the separate hormones. In relation to traditional anthropometric measures of obesity, Asian Indian men and women had the highest and Caucasians the lowest concentrations of leptin while Creole (African) men and women had intermediate levels. As a risk marker, leptin has a high intra-individual stability, equal in men and women and among different ethnicities over 5–13 years with no circ-annual variation.

Abstract Type 2 diabetes is increasing overall in the world and is mostly associated with Western lifestyle including replete, unhealthy diet, sedentary life style and growing obesity. In the future the highest prevalence will be seen among older people due to longer life expectancy and changes in demography. Because diabetes is associated with increased morbidity, especially from cardiovascular causes, and a shortened life expectancy, an important aspect in the older population is the impact of diabetes on health related quality of life (HRQoL) and development of disability. To measure HRQoL we have many valid instruments, one of them the widely used RAND-36 survey. The 2-hour glucose value is important in screening subjects at high risk for diabetes, but it is time consuming and costly. Recently, 1-hour post load glucose has aroused interest in the prediction of diabetes. Few studies have focused on the effect of the age of onset of diabetes and how it effects on the HRQoL at an older age. The aim of this study was to investigate the risk factors for future diabetes in men healthy in midlife; the interest was especially focused on 1-hour post load glucose. The other objective was to estimate the HRQoL in men who develop diabetes at an old age. The present prospective study is based on the Helsinki Businessmen Study originally consisting of 3.490 men, born in 1919–1934 and followed since the 1960´s. All the men had socioeconomically similar status and belonged to the highest social group. The extensive baseline examinations were conducted in 1974, when the men were in midlife, mean age 48. At that time the men´s health, medication and cardiovascular risk factors were observed and self-related health (SRH) was rated on a five-step scale. The men who were healthy without medications were included in the follow-up group. The men were later investigated with postal questionnaires (1985/86, 2000, 2002/03, 2007). In 2000, at mean age of 73, the HRQoL of the survivors was examined using the RAND-36 instrument and was replicated in 2002/03, 2005 and 2007. Development of diabetes was evaluated using the National Drug Reimbursement register and self-reported diabetes in questionnaires. Baseline 1-hour post-load blood glucose and weight gain from the age of 25 to midlife predicted future diabetes, and especially a combination of 1-hour glucose >8.9 mmol/L and body mass index (BMI) of ≥30 was associated with a 10.1-fold increase of diabetes risk independently of cardiovascular risk factors. Men with late-onset of diabetes (age >75) tented to be healthier in midlife. Diabetes did not affect the HRQoL until after diabetes onset. According to the evaluation in 2000, three RAND-36 scales, i.e. physical functioning, general health and social functioning, worsened already after 0–4 years from diabetes onset but did not deteriorate thereafter. There was no consistent impact on mental health. In conclusion, this study demonstrates that in men, who develop diabetes later in old age, cardiovascular risk factors in midlife and elevated 1-hour post-load glucose and weight gain up to midlife are important predictors for future diabetes. Developing diabetes exerts clear effects on HRQoL, measured with RAND-36 very early after diagnosis, but affects only some of the domains. / Tiivistelmä Tyypin 2 diabetes on lisääntymässä maailmanlaajuisesti ja on suurelta osin yhteydessä länsimaisen elämäntyylin yleistymiseen. Keskeisimpinä tekijöinä ovat muuttunut, yltäkylläinen ruokavalio, vähentynyt fyysinen aktiivisuus ja niiden myötä lihavuuden lisääntyminen. Diabeteksen lisääntyminen näyttäisi olevan suurinta yli 65-vuotiaiden keskuudessa. Tähän vaikuttavat mm. eliniän pidentyminen ja väestörakenteen muutokset. Diabetesta sairastavien riski sairastua ja kuolla sydän- ja verisuonisairauksien komplikaatioihin on suurempi kuin ei-diabeetikoilla ja sairauksien myötä toimintakyky sekä elämänlaatu heikkenevät. Henkilön oma arvio hyvinvoinnista on tärkeä mittari. Elämänlaadun luotettava mittaaminen on tullut mahdolliseksi testattujen kysymyssarjojen myötä. Yksi laajasti käytetty mittari on RAND-36. Diabetesriskissä olevien henkilöiden löytämiseksi on käytetty 2-tunnin glukoosirasitustestiä, joka on aikaa vievä. Viimeaikoina onkin herännyt kiinnostus glukoosirasituksen jälkeistä yhden tunnin sokeriarvoa kohtaan. Useimmat tyypin 2 diabetekseen liittyvät tutkimukset ovat keskittyneet työikäiseen väestöön, minkä vuoksi tutkimuksia vanhemmalla iällä puhjenneesta diabeteksesta ja sen vaikutuksesta elämänlaatuun on vähemmän. Tämän tutkimuksen tarkoituksena oli selvittää keski-iän riskitekijöitä, jotka ennustavat tulevaa tyypin 2 diabetesta myöhemmällä iällä. Yhtenä tutkimuskohteena oli yhden tunnin glukoosirasituksen jälkeinen sokeriarvo. Myöhemmällä iällä puhjenneen diabeteksen merkitystä elämänlatuun kartoitettiin RAND-36-mittarilla. Tutkimusaineistona oli ns. Helsingin Johtajatutkimuksen mieskohortti, jota on seurattu 1960-luvulta alkaen. Kaikki miehet (n=3.490) olivat alkuaan terveitä, vuosina 1919–1934 syntyneitä miehiä, jotka kuuluivat ylimpään sosiaaliryhmään. Heidän terveydentilaansa on kartoitettu nykypäiviin saakka. Laajempi perustutkimus tehtiin vuonna 1974, jolloin miehet olivat keski-iässä (48 v). Seurantaan valittiin terveet henkilöt, joilla ei ollut säännöllisiä lääkityksiä. Tällöin kartoitettiin sydän- ja verisuonisairauksien riskitekijöitä, mitattiin glukoosirasituksen jälkeinen yhden tunnin sokeriarvo ja selviteltiin elämäntapoja, minkä lisäksi sen hetkistä itse koettua hyvinvointia mitattiin 5-portaisella asteikolla. Postitse lähetettyjä kyselytutkimuksia on tehty tietyin väliajoin ja aineiston tiedot tähän tutkimukseen on kerätty 31.12.2007 mennessä. Vuonna 2000 kyselytutkimukseen liitettiin elämänlaatumittarina RAND-36-mittarin suomalainen versio. Diabeteksen puhkeamista on seurattu Kelan lääkekorvausrekisteristä sekä itse ilmoitettuna postikyselyiden kautta. Tässä keski-iässä terveessä mieskohortissa tulevan diabeteksen kannalta merkittävimmät ennusteelliset tekijät olivat painon nousu 25-vuoden iästä keski-ikään sekä keski-iässä mitattu glukoosirasituksen jälkeinen yhden tunnin sokeriarvo. Etenkin yhden tunnin sokeriarvo >8.9 mmol/L ja BMI ≥30 nostivat tulevan diabeteksen riskin 10-kertaiseksi. Näillä tekijöillä oli myös vahva yhteys sydän- ja verisuonisairauksien aiheuttamaan kuolleisuuteen. Glukoosirasituksen jälkeisellä yhden tunnin sokeriarvolla näyttäisi olevan merkitystä arvioitaessa tyypin 2 diabetesriskiä tulevaisuudessa, katkaisupisteenä plasman ≥7.8 mmol/L. Elämänlaatua diabetes näytti heikentävän pian diagnoosin jälkeen (0–4 vuotta diagnoosista), mutta sen jälkeen elämänlaatu ei huonontunut oleellisesti. RAND-36 mittarilla mitattuna elämänlaatu heikkeni merkittävästi diabeetikoilla ei-diabeetikoihin verrattuna fyysisen toimintakyvyn, yleisen elämänlaadun ja sosiaalisten toimintojen osa-alueilla, mutta mielenterveyteen diabetes ei näyttänyt vaikuttavan.

La prévalence du diabète ne cesse d’augmenter et la détection de sujets à risque de développer cette maladie reste une préoccupation importante afin qu’un programme de prévention leur soit proposé. Le dosage de l’HbA1c est utilisé pour l’évaluation de l’efficacité du traitement pris par les diabétiques. Il n’était pas utilisé pour le diagnostic des dysglycémies car les méthodes de ce dosage n’étaient pas standardisées. Actuellement l’IFCC (International Federation of Clinical Chemistry) propose une nouvelle méthode de référence pour cette mesure, et l’OMS (Organisation Mondiale de la Santé) a intégré ce paramètre dans les critères diagnostiques du diabète. but Evaluer la capacité de l’HbA1c à prédire un diabète incident, chercher des seuils, d’HbA1c, de GAJ et de glycémie deux heures (G2H) après un test d’hyperglycémie provoquée par voie orale (HGPO), au delà desquels les sujets sont plus à risque de développer un diabète incident et, enfin, étudier l’influence du tabac sur les variations de ces mesures. Méthodes Pour l’évaluation de la prédiction du diabète et la recherche de seuils, nous avons utilisé les données de l’étude australienne AusDiab, de l’étude danoise Inter99 et de l’étude française D.E.S.I.R. avec plus de 5500, 4500 et 3550 sujets, respectivement. Dans la troisième partie, nous avons utilisé les données du projet DETECT-2 (12 études, 26 000 sujets), ainsi que celles des études françaises D.E.S.I.R. et TELECOM (3700 sujets). Les distributions de l’HbA1c dans les trois premières études étaient différentes, nous avons donc, dans les deux premières parties d’analyses, ajusté les moyennes d’HbA1c à l’inclusion et après le suivi. Nous avons utilisé un modèle logistique pour la comparaison du pouvoir prédictif de l’HbA1c et de GAJ ; l’intervalle de confiance des Odds Ratios (ORs) à été obtenu par bootstrap. Pour rechercher les seuils de prédiction du diabète, nous avons comparé le modèle logistique (avec la variable glycémique) sans seuil au modèle avec seuil. Enfin, nous avons utilisé un modèle linéaire mixte pour évaluer la différence entre les moyennes d’HbA1c, de GAJ et de G2H en fonction du tabagisme, en attribuant un effet aléatoire à la variable ‘centre’. Résultats Avant ajustement des moyennes d’HbA1c, l’incidence du diabète (défini par la prise de traitement antidiabétique, une HbA1c≥6.5% ou une GAJ≥7mmol/l) était de 3.1% dans AusDiab, 2.7% dans Inter99 et 2.5% dans D.E.S.I.R. Les sujets dépistés comme étant diabétiques par l’HbA1c et par la GAJ n’étaient pas toujours les mêmes. L’incidence estimée du diabète augmente bien avec l’augmentation du taux de l’HbA1c et de la GAJ à l’inclusion. Pour chaque critère, l’aire sous la courbe de ROC (Receiver Operating Characteristic) était supérieure à 0.80 témoignant d’une bonne discrimination des deux tests entre les diabétiques incidents et les non diabétiques et le test de Hosmer-Lemeshow témoigne d’une bonne adéquation des modèles utilisés (p>0.05). Dans les trois populations, les ORs qui mesurent l’association entre les taux d’HbA1c et de GAJ et la survenue du diabète étaient presque toujours supérieurs pour l’HbA1c. Les seuils d’HbA1c et de GAJ au-delà desquels les sujets étaient plus à risque de développer un diabète variaient en fonction de la définition du diabète incident sauf pour l’HbA1c dans l’étude D.E.S.I.R. (5.3%) et la GAJ dans l’étude AusDiab (5.5mmol/l). Enfin, la moyenne d’HbA1c chez les fumeurs actuels était 0.10%(0.08,0.12) plus élevée que chez ceux qui n’ont jamais fumé ; la moyenne de G2H était -0.44(-0.51,-0.36) moins élevée chez les fumeurs actuels que chez ceux qui n’ont jamais fumé.Conclusion Ces résultats soulignent : 1) l’importance de l’utilisation de l’HbA1c comme critère diagnostique de dysglycémies, 2) la nécessité de mieux explorer les limites inférieures des stades intermédiaires qui précèdent la survenue du diabète, 3) l’importance de prendre en considération les facteurs qui peuvent influencer les taux d’HbA1c / The increasing prevalence of diabetes worldwide makes the detection of people at risk of developing diabetes a major concern, so that they can benefit from diabetes prevention programs. HbA1c is used to evaluate the effectiveness of treatment taken by diabetic patients. HbA1c had not been used to diagnose dysglycemia because the assay methods were not standardized. The International Federation of Clinical Chemistry has proposed a reference method, and in 2011 the World Health Organization included HbA1c as one of the criteria for the diagnosis of diabetes. aims: 1) To evaluate the ability of HbA1c to predict incident diabetes compared with fasting plasma glucose (FPG); 2) to find thresholds for HbA1c, FPG and two hour plasma glucose (G2H) after an oral glucose tolerance test (OGTT) beyond which subjects are more at risk for developing incident diabetes and finally; 3) to study the influence of smoking on HbA1c, FPG and G2H.Methods: Several populations were studied. To evaluate the prediction of diabetes and the search for thresholds, we used data from the Australian study (AusDiab), a Danish study (Inter99) and a French study (D.E.S.I.R.) with respectively more then 5500, 4500 and 3550 participants. In the third part, we used data from the DETECT-2 consortium (12 studies with more than 26 000 men and women) and from two French studies: D.E.S.I.R. and TELECOM (with more than 3700 participants). The distribution of HbA1c in AusDiab, Inter99 and D.E.S.I.R. differed, so in the first two parts of this thesis, we adjusted HbA1c so that all three studies had the same mean HbA1c at baseline and the same mean HbA1c at follow-up. We used a logistic model to quantify the predictive ability of HbA1c and FPG for diabetes, and then derived confidence intervals for the difference in Odds Ratios (ORs) by bootstrap. To search for thresholds to predict incident diabetes, based on HbA1c, FPG and G2H at inclusion, we compared logistic regression models that were linear in the glycaemic variable, without a threshold, with a spline model with a threshold. Linear mixed models with ‘centre’ as a random variable, were used to assess the difference between the means of HbA1c, FPG and G2H in current-, ex- and never-smokers.Results: With unadjusted HbA1c data, the incidence of diabetes (defined by treatment, HbA1c≥6.5% or FPG≥7 mmol/l) was 3.1% in AusDiab, 2.7% in Inter99 and 2.5% in D.E.S.I.R.. Subjects detected as having diabetes by HbA1c and FPG were not always the same. The incidence of diabetes increased with increasing HbA1c and FPG at baseline. For each test, the area under the Receiver Operating Characteristic curve was greater than 0.80, indicating good discrimination for these two measures between those with and without incident diabetes, and the Hosmer-Lemeshow test indicated that the models fitted well (p>0.05). In all three populations, the ORs measuring the association between HbA1c and FPG and the development of diabetes were almost always higher for HbA1c than for FPG. The thresholds of HbA1c and FPG above which the incidence of diabetes were higher, varied according to the definition of incident diabetes - except for HbA1c in D.E.S.I.R. (always 5.3%) and for FPG in AusDiab (always 5.5mmol/l). Finally, in current-smokers, the mean HbA1c was 0.10%(0.08,0.12) higher than in never-smokers; the mean G2H was 0.44( 0.51,-0.36) lower in current-smokers than in never-smokers. Conclusion: The results that we found emphasize: 1) the importance of using HbA1c as a diagnostic criterion for dysglycemia, as those diagnosed diabetic by HbA1c did not have always an FPG ≥ 7 mmol/l, 2) the need to better explore the lower limits of the “pre-diabetic” stage as the thresholds of HbA1c, FPG and 2H-PG that we found were lower than those used in clinical practice, 3) the importance to consider factors that may influence HbA1c and G2H, such as smoking.