There is accumulating evidence on differences in the link between types of dairy products and cardio-metabolic health, but inconsistent findings limit the field. In my PhD project, I undertook an epidemiological investigation comprising inter-related but distinct themes evaluating aspects of nutritional, molecular and genetic epidemiology to advance scientific understanding. I undertook research to describe dairy consumption patterns over time by evaluating nationally-representative data of the United Kingdom National Diet and Nutrition Survey. I observed significant time trends for specific dairy types and groups, which were different among different groups of people e.g. adults younger than 65 years or elderly people. Using data from the large Fenland (n~12,000) and EPIC Norfolk (n~25,000) studies, I investigated associations of total and types of dairy consumption with markers of metabolic risk and adiposity as potential pathways to cardio-metabolic disease. The analyses showed differential associations of dairy types and groups mainly with markers of adiposity and lipidaemia. I explored the potential of objective markers to assess dairy consumption, by examining metabolomics profiles and blood fatty acids to identify a set of biomarkers predicting dairy consumption and prospective associations of the identified biomarkers with type 2 diabetes risk. I was able to develop and validate metabolite scores reflecting consumption of some dairy products and observed inverse associations between some of these scores and type 2 diabetes incidence. I analysed genetic determinants of dairy consumption, using a genome-wide association study in the UK Biobank (n~500,000) and identified single nucleotide polymorphisms predicting milk, cheese and total dairy consumption. Overall, this PhD work contributed towards (1) a more precise description of dairy consumption patterns in the UK, (2) hypothesis formulation for potential biological pathways linking to cardio-metabolic disease, (3) discovery of metabolite scores as potential dairy biomarkers and (4) hypothesis formulation for potential genetic predictors of dairy consumption.

Background and Aims: Type 2 diabetes mellitus (T2DM) is a chronic progressive condition characterised by hyperglycaemia due to insulin deficiency with or without insulin resistance. The prevalence of diabetes is increasing rapidly worldwide and it has a significant burden on health care resources with estimated costs of up to 10% of health expenditure in the UK. With an ageing population, people are now living longer with diabetes which consequently leads to increased multimorbidity and polypharmacy. Some previous studies have not assessed the effect of demographic or geographic factors (such as age, gender, UK nation and social deprivation) on the incidence and prevalence of T2DM in the UK over the past decade. In addition, detailed reports on the patterns of comorbidities in T2DM patients are sparse. Patients with T2DM are at a two-fold higher risk for cardiovascular (CV) disease. Some earlier studies assessing the CV risk associated with available therapies have been inconclusive in determining the preferred regimens. This thesis aimed to: i) assess the incidence and prevalence of T2DM in the UK; ii) investigate and compare mortality risk between T2DM patients and patients without diabetes and explore if it explains the observed prevalence rates; iii) examine the patterns of comorbidities in T2DM patients and matched comparators without diabetes; and iv) assess the comparative CV risk associated with second-line diabetes therapies. Methods: All the studies in this thesis used data from the UK Clinical Practice Research Datalink. Access to linked national hospitalisation, deprivation and mortality data was obtained for the individual studies. Annual overall and gender-specific incidence and prevalence of T2DM were calculated for the study period 2004-2014. Rates were standardised by age bands, gender, neighbourhood social deprivation, and UK nation and expressed per 10,000 person-years (PYRs) with 95% confidence intervals (95% CI). For the mortality analysis, T2DM patients (cases) were matched to patients without diabetes (controls) on age, gender and general practice. Annual mortality rates for the matched T2DM and patients without diabetes were calculated and compared. Cox regression analysis was used to examine the effect of important covariates on the risk for all-cause mortality in the matched cohort and calculate hazard ratios (HR) and 95% CI. The multimorbidity profile in T2DM cases and matched controls, registered in English general practices, were also examined. Annual prevalence rates of 18 physical and mental health comorbidities were determined between 2004 and 2014 using linked primary care and hospitalisation records. For the CV risk analysis, patients prescribed a second-line medication after greater than or equal to90days of metformin monotherapy between 1998 and 2011 were identified. Using a retrospective cohort study design, inverse probability of treatment-weighted time-varying Cox regression models were used to estimate HRs and 95% CI for developing a major CV event (myocardial infarction, stroke, acute coronary syndrome, unstable angina, coronary revascularisation, or CV death) associated with second-line therapies after adjusting for clinically important CV risk factors. Results: The prevalence of T2DM nearly doubled from 320.62 (95% CI: 318.83; 322.41) in 2004 to 526.36 per 10,000 PYR (95% CI: 523.81; 528.91) in 2014, whereas the incidence was relatively stable with overall rate of 43.07 per 10,000 PYR (95% CI: 40.06; 46.09). Gender-specific incidence and prevalence rates were markedly higher in men than women. Between 2004 and 2014, the prevalence increased from 380.31 (95% CI: 377.48; 383.13) to 625.45 (95% CI: 621.37; 629.52) in men and from 268.56 (95% CI: 266.22; 270.90) to 437.28 (95% CI: 433.94; 440.62) in women. Overall, older individuals, men, and residents in the most deprived locations were more likely to have T2DM. Wales and Northern Ireland had higher prevalence rates than the other UK nations. In the all-cause mortality analysis, 20,312 (11.5%) patients with T2DM died, as compared with 79,951 (9.1%) controls. The adjusted survival model showed that patients with T2DM were at significantly greater risk for mortality in comparison with patients without diabetes (HR: 1.26, 95% CI: 1.20; 1.32). Mortality rates decreased over time in both cases and controls. The multimorbidity study showed that comorbidities were more prevalent in patients diagnosed with T2DM in comparison with patients without diabetes. The number of patients with two comorbidities increased between 2004 and 2014. The prevalence of cardiovascular disease (CVD) in T2DM patients was double that of matched control patients. DPP-4 inhibitors, thiazolidinediones and sulphonylureas add-on therapies to metformin were the most commonly-prescribed second-line therapies. The time-varying survival models showed that DPP-4 inhibitors (HR: 0.78, 95% CI: 0.55; 1.11) and thiazolidinediones (HR: 0.68, 95% CI: 0.54; 0.85) add-on therapies were associated with lower risk for major CVD compared to sulphonylurea add-on therapy when all added to initial metformin. Conclusions: The prevalence of T2DM is increasing rapidly in the UK. Patients with T2DM are at significantly greater risk for mortality than patients without diabetes. However, with the declining mortality rates over the past decade, patients are now living longer to develop comorbidities. CVD was the most prevalent comorbidity in T2DM cases in comparison with people without diabetes. This is important as CVD is the main cause of mortality in patients with T2DM. Thiazolidinedione combination with metformin was associated with significantly lower CV risk in comparison with sulphonylurea add-on therapies to metformin. Lower, but non-statistically significant, risks were also found with DPP-4 inhibitors add-on therapies. These real-world findings add to the existing knowledge on the epidemiology of T2DM, provide novel insight on the patterns of multimorbidity in these patients and clinically relevant evidence on the CV risk associated with commonly-prescribed second-line regimens. Future larger studies are needed to confirm the observed CV benefits associated with antidiabetic therapies.

Background The Chinese famine of 1959–61 has been widely interpreted as an important driver of current and future epidemics of type 2 diabetes (T2D). We conducted a systematic review and meta-analysis of prenatal famine exposure and type 2 diabetes (T2D) in China to summarize study characteristics, examine impacts of control selections on study results, and identify whether study results can be related to any characteristics. Methods We searched PubMed, Embase, Wanfang Data, and CNKI databases for studies that examined the relationship between T2D and prenatal exposure to the Chinese famine up to January, 2021. From included studies, we abstracted information on the number of T2D cases and populations at risk among individuals born during the famine (famine births), born before the famine (pre-famine births), and born after the famine (post-famine births). We examined the quality of studies with the modified Newcastle–Ottawa scale. We compared T2D in famine births to different controls: post-famine births, pre-famine and post-famine births combined (age-balanced), and pre-famine births. Fixed-effects models and random-effects models were used to calculate summary estimates. Heterogeneity across studies was assessed, and subgroup analyses were performed using sex, age at the survey, T2D measurements, famine intensity, residence, and publication language as possible effect modifiers. Findings Of 5,363 studies identified, 18 studies met our inclusion criteria. All studies defined famine exposure based on participants' years and/or months of births. Sample sizes ranged between around 300 and over 80,000 across studies. When post-famine births were used as controls, we found an increased risk of T2D (OR 1.48, 95% CI 1.30–1.68) among famine births based on the random-effects model. Using age-balanced controls, we did not find any increased risk of T2D (1.09, 95% CI 0.98–1.22). When pre-famine births were used as controls, famine was associated with a reduction in risk (0.85, 95% CI 0.75–0.97). Large variations and inconsistencies were identified in study design, famine intensity assessment, and covariate adjustment across studies. Conclusion Our analysis shows that it still remains an open question whether the Chinese famine has contributed substantially to the current T2D epidemic in China. Studies with more rigorous methods will need to quantify this relationship. The four-level famine intensity instrument developed in this study is a robust measure to identify a potential dose-response relation between famine exposure and health outcomes. It is likely to improve the reproducibility and generalizability of future studies.

Contexto: O diabetes mellitus (DM) é uma causa importante de morbimortalidade nas sociedades ocidentais devido à carga de sofrimento, incapacidade, perda de produtividade e morte prematura que provoca. No Brasil, seu impacto econômico é desconhecido. Objetivos: Dimensionar a participação do DM nas hospitalizações da rede pública brasileira (1999-2001), colaborando na avaliação dos custos diretos. Especificamente, analisar as hospitalizações (327.800) e os óbitos hospitalares (17.760) por DM como diagnóstico principal (CID-10 E10-E14 e procedimento realizado) e estimar as hospitalizações atribuíveis ao DM, incluindo as anteriores e aquelas por complicações crônicas (CC) e condições médicas gerais (CMG). Métodos: A partir de dados do Sistema de Informação Hospitalar do Sistema Único de Saúde (SIH/SUS) (37 milhões de hospitalizações), foram calculados indicadores por região de residência do paciente e sexo (ajustados por idade pelo método direto, com intervalos de confiança de 95%), faixas etárias, médias de permanência e de gastos por internação e populacional em US$. Realizou-se regressão logística múltipla para o desfecho óbito. As prevalências de DM foram combinadas aos riscos relativos de hospitalização por CC e CMG (metodologia do risco atribuível) e somadas às internações por DM como diagnóstico principal. Utilizou-se análise de sensibilidade para diferentes prevalências e riscos relativos. Resultados: Os coeficientes de hospitalizações e de óbitos hospitalares e a letalidade por DM como diagnóstico principal atingiram respectivamente 6,4/104hab., 34,9/106hab. e 5,4%. As mulheres apresentaram os coeficientes mais elevados, porém os homens predominaram na letalidade em todas as regiões. O gasto médio (US$ 150,59) diferiu significativamente entre as internações com e sem óbito, mas a média de permanência (6,4 dias) foi semelhante. O gasto populacional equivaleu a US$ 969,09/104hab. As razões de chances de óbito foram maiores para homens, pacientes ≥75 anos, e habitantes das regiões Nordeste e Sudeste. As hospitalizações atribuíveis ao DM foram estimadas em 836,3 mil/ano (49,3/104hab.), atingindo US$ 243,9 milhões/ano (US$ 14,4 mil/104hab.). DM como diagnóstico principal (13,1%), CC (41,5%) e CMG (45,4%) responderam por 6,7%, 51,4% e 41,9% respectivamente dos gastos. O valor médio das internações atribuíveis (US$ 292) situou-se 36% acima das não-atribuíveis. As doenças vasculares periféricas apresentaram a maior diferença no valor médio entre hospitalizações atribuíveis e não-atribuíveis (24%), porém as cardiovasculares destacaram-se em quantidade (27%) e envolveram os maiores gastos (37%). Os homens internaram menos (48%) que as mulheres, porém com gasto total maior (53%). As internações de pacientes entre 45-64 anos constituíram o maior grupo (45%) e gastos (48%) enquanto os pacientes com ≥75, os maiores coeficientes de hospitalização (350/104hab.) e de despesa (US$ 93,4 mil/104hab.). As regiões mais desenvolvidas gastaram o dobro (/104hab.) em relação às demais. Considerações Finais e Recomendações: As configurações no consumo de serviços hospitalares foram semelhantes às de países mais desenvolvidos, com importantes desigualdades regionais e de sexo. O gasto governamental exclusivamente com hospitalizações atribuíveis ao DM foi expressivo (2,2% do orçamento do Ministério da Saúde). A ampliação de atividades preventivas poderia diminuir a incidência do DM, reduzir a necessidade de internações, minimizar as complicações e minorar a severidade de outras condições médicas mais gerais. / Background: Diabetes mellitus (DM) is one of the main causes of morbi/mortality in western societies due to the burden of suffering, disabilities, loss of productivity and premature death that encompasses. Its economic impact is unknown in Brazil. Objectives: To dimension the share of DM hospitalizations on the Brazilian national health system (1999-2001), helping on evaluating direct costs. Specifically, to analyze hospitalizations (327.800) and hospitalization deaths (17.760) caused by DM as first-listed diagnosis (ICD-10 E10-E14 and procedure done) and to estimate the magnitude of DM attributable hospitalizations, including DM itself, chronic complications (CC) and general medical conditions (GMC). Methods: Data from the Hospital Information System of the National Health System (SIH/SUS) (37 millions of hospitalizations). Indicators were calculated by residence region of the patients and sex (adjusted by direct method for age with 95% confidence intervals), age intervals, average length of stay and expenditure by admission and population in US$. Multiple logistic regression was performed for death as outcome. Combinations of DM prevalence and hospitalization relative risks for CC and GMC were added to DM first-listed hospitalizations (attributable risk methodology). Sensitivity analyze was used for different prevalences and relative risks. Results: Hospitalizations and hospitalization deaths coefficients and lethality by DM as first-listed diagnosis were 6.4/104inhab., 34.9/106inhab. and 5.4% respectively. Coefficients were higher for women, although lethality was for men in every five region. Average expenditure (US$150,59) differed significantly between those with/without death but presented equal average length of stay (6.4). Population expenditure was US$ 969.09/104inhab. Odds-ratio for dying were larger for men, patients 75 yrs, and inhabitants of northeast and southeast. Hospitalizations attributable to DM were estimated at 836.3 thousand/year (49.3/104inhab.) reaching US$ 243.9 millions/year (US$ 14.4 thousand/104inhab.). DM as fist-list diagnosis (13.1%), CC (41.5%) and GMC (45.4%) depicted 6.7%, 51.4% e 41.9% respectively of annual expenditures. Average value of attributable hospitalizations (US$ 292) was 36% higher than non-attributable. Peripheral vascular diseases posed the largest excess based on average values (24%) although cardiovascular ones represented the major quantity (27%) and expenditure group (37%). Men were less admitted (48%) than women, but incurred more expenditure (53%). People 45-64 years old consisted the largest (45%) and most expensive (48%) group while 75+ generated the highest coefficients of hospitalization (350/104inhab.) and expenditure (US$ 93.4 thousands/104inhab.). Most developed regions accounted for nearly twice expenses than other regions. Conclusions and Recommendations: Patterns of hospitalization were similar to those most developed countries. Important regional and gender inequalities did exist. Governmental expenditures related exclusively to DM attributable hospitalizations was meaningful (2.2% of the budget of the Ministry oh Health). Broadening preventive health care actions could diminish the incidence of DM, reduce the need for hospitalizations, minimize complications and minors the severity of general medical conditions.

Contexto: O diabetes mellitus (DM) é uma causa importante de morbimortalidade nas sociedades ocidentais devido à carga de sofrimento, incapacidade, perda de produtividade e morte prematura que provoca. No Brasil, seu impacto econômico é desconhecido. Objetivos: Dimensionar a participação do DM nas hospitalizações da rede pública brasileira (1999-2001), colaborando na avaliação dos custos diretos. Especificamente, analisar as hospitalizações (327.800) e os óbitos hospitalares (17.760) por DM como diagnóstico principal (CID-10 E10-E14 e procedimento realizado) e estimar as hospitalizações atribuíveis ao DM, incluindo as anteriores e aquelas por complicações crônicas (CC) e condições médicas gerais (CMG). Métodos: A partir de dados do Sistema de Informação Hospitalar do Sistema Único de Saúde (SIH/SUS) (37 milhões de hospitalizações), foram calculados indicadores por região de residência do paciente e sexo (ajustados por idade pelo método direto, com intervalos de confiança de 95%), faixas etárias, médias de permanência e de gastos por internação e populacional em US$. Realizou-se regressão logística múltipla para o desfecho óbito. As prevalências de DM foram combinadas aos riscos relativos de hospitalização por CC e CMG (metodologia do risco atribuível) e somadas às internações por DM como diagnóstico principal. Utilizou-se análise de sensibilidade para diferentes prevalências e riscos relativos. Resultados: Os coeficientes de hospitalizações e de óbitos hospitalares e a letalidade por DM como diagnóstico principal atingiram respectivamente 6,4/104hab., 34,9/106hab. e 5,4%. As mulheres apresentaram os coeficientes mais elevados, porém os homens predominaram na letalidade em todas as regiões. O gasto médio (US$ 150,59) diferiu significativamente entre as internações com e sem óbito, mas a média de permanência (6,4 dias) foi semelhante. O gasto populacional equivaleu a US$ 969,09/104hab. As razões de chances de óbito foram maiores para homens, pacientes ≥75 anos, e habitantes das regiões Nordeste e Sudeste. As hospitalizações atribuíveis ao DM foram estimadas em 836,3 mil/ano (49,3/104hab.), atingindo US$ 243,9 milhões/ano (US$ 14,4 mil/104hab.). DM como diagnóstico principal (13,1%), CC (41,5%) e CMG (45,4%) responderam por 6,7%, 51,4% e 41,9% respectivamente dos gastos. O valor médio das internações atribuíveis (US$ 292) situou-se 36% acima das não-atribuíveis. As doenças vasculares periféricas apresentaram a maior diferença no valor médio entre hospitalizações atribuíveis e não-atribuíveis (24%), porém as cardiovasculares destacaram-se em quantidade (27%) e envolveram os maiores gastos (37%). Os homens internaram menos (48%) que as mulheres, porém com gasto total maior (53%). As internações de pacientes entre 45-64 anos constituíram o maior grupo (45%) e gastos (48%) enquanto os pacientes com ≥75, os maiores coeficientes de hospitalização (350/104hab.) e de despesa (US$ 93,4 mil/104hab.). As regiões mais desenvolvidas gastaram o dobro (/104hab.) em relação às demais. Considerações Finais e Recomendações: As configurações no consumo de serviços hospitalares foram semelhantes às de países mais desenvolvidos, com importantes desigualdades regionais e de sexo. O gasto governamental exclusivamente com hospitalizações atribuíveis ao DM foi expressivo (2,2% do orçamento do Ministério da Saúde). A ampliação de atividades preventivas poderia diminuir a incidência do DM, reduzir a necessidade de internações, minimizar as complicações e minorar a severidade de outras condições médicas mais gerais. / Background: Diabetes mellitus (DM) is one of the main causes of morbi/mortality in western societies due to the burden of suffering, disabilities, loss of productivity and premature death that encompasses. Its economic impact is unknown in Brazil. Objectives: To dimension the share of DM hospitalizations on the Brazilian national health system (1999-2001), helping on evaluating direct costs. Specifically, to analyze hospitalizations (327.800) and hospitalization deaths (17.760) caused by DM as first-listed diagnosis (ICD-10 E10-E14 and procedure done) and to estimate the magnitude of DM attributable hospitalizations, including DM itself, chronic complications (CC) and general medical conditions (GMC). Methods: Data from the Hospital Information System of the National Health System (SIH/SUS) (37 millions of hospitalizations). Indicators were calculated by residence region of the patients and sex (adjusted by direct method for age with 95% confidence intervals), age intervals, average length of stay and expenditure by admission and population in US$. Multiple logistic regression was performed for death as outcome. Combinations of DM prevalence and hospitalization relative risks for CC and GMC were added to DM first-listed hospitalizations (attributable risk methodology). Sensitivity analyze was used for different prevalences and relative risks. Results: Hospitalizations and hospitalization deaths coefficients and lethality by DM as first-listed diagnosis were 6.4/104inhab., 34.9/106inhab. and 5.4% respectively. Coefficients were higher for women, although lethality was for men in every five region. Average expenditure (US$150,59) differed significantly between those with/without death but presented equal average length of stay (6.4). Population expenditure was US$ 969.09/104inhab. Odds-ratio for dying were larger for men, patients 75 yrs, and inhabitants of northeast and southeast. Hospitalizations attributable to DM were estimated at 836.3 thousand/year (49.3/104inhab.) reaching US$ 243.9 millions/year (US$ 14.4 thousand/104inhab.). DM as fist-list diagnosis (13.1%), CC (41.5%) and GMC (45.4%) depicted 6.7%, 51.4% e 41.9% respectively of annual expenditures. Average value of attributable hospitalizations (US$ 292) was 36% higher than non-attributable. Peripheral vascular diseases posed the largest excess based on average values (24%) although cardiovascular ones represented the major quantity (27%) and expenditure group (37%). Men were less admitted (48%) than women, but incurred more expenditure (53%). People 45-64 years old consisted the largest (45%) and most expensive (48%) group while 75+ generated the highest coefficients of hospitalization (350/104inhab.) and expenditure (US$ 93.4 thousands/104inhab.). Most developed regions accounted for nearly twice expenses than other regions. Conclusions and Recommendations: Patterns of hospitalization were similar to those most developed countries. Important regional and gender inequalities did exist. Governmental expenditures related exclusively to DM attributable hospitalizations was meaningful (2.2% of the budget of the Ministry oh Health). Broadening preventive health care actions could diminish the incidence of DM, reduce the need for hospitalizations, minimize complications and minors the severity of general medical conditions.

Contexto: O diabetes mellitus (DM) é uma causa importante de morbimortalidade nas sociedades ocidentais devido à carga de sofrimento, incapacidade, perda de produtividade e morte prematura que provoca. No Brasil, seu impacto econômico é desconhecido. Objetivos: Dimensionar a participação do DM nas hospitalizações da rede pública brasileira (1999-2001), colaborando na avaliação dos custos diretos. Especificamente, analisar as hospitalizações (327.800) e os óbitos hospitalares (17.760) por DM como diagnóstico principal (CID-10 E10-E14 e procedimento realizado) e estimar as hospitalizações atribuíveis ao DM, incluindo as anteriores e aquelas por complicações crônicas (CC) e condições médicas gerais (CMG). Métodos: A partir de dados do Sistema de Informação Hospitalar do Sistema Único de Saúde (SIH/SUS) (37 milhões de hospitalizações), foram calculados indicadores por região de residência do paciente e sexo (ajustados por idade pelo método direto, com intervalos de confiança de 95%), faixas etárias, médias de permanência e de gastos por internação e populacional em US$. Realizou-se regressão logística múltipla para o desfecho óbito. As prevalências de DM foram combinadas aos riscos relativos de hospitalização por CC e CMG (metodologia do risco atribuível) e somadas às internações por DM como diagnóstico principal. Utilizou-se análise de sensibilidade para diferentes prevalências e riscos relativos. Resultados: Os coeficientes de hospitalizações e de óbitos hospitalares e a letalidade por DM como diagnóstico principal atingiram respectivamente 6,4/104hab., 34,9/106hab. e 5,4%. As mulheres apresentaram os coeficientes mais elevados, porém os homens predominaram na letalidade em todas as regiões. O gasto médio (US$ 150,59) diferiu significativamente entre as internações com e sem óbito, mas a média de permanência (6,4 dias) foi semelhante. O gasto populacional equivaleu a US$ 969,09/104hab. As razões de chances de óbito foram maiores para homens, pacientes ≥75 anos, e habitantes das regiões Nordeste e Sudeste. As hospitalizações atribuíveis ao DM foram estimadas em 836,3 mil/ano (49,3/104hab.), atingindo US$ 243,9 milhões/ano (US$ 14,4 mil/104hab.). DM como diagnóstico principal (13,1%), CC (41,5%) e CMG (45,4%) responderam por 6,7%, 51,4% e 41,9% respectivamente dos gastos. O valor médio das internações atribuíveis (US$ 292) situou-se 36% acima das não-atribuíveis. As doenças vasculares periféricas apresentaram a maior diferença no valor médio entre hospitalizações atribuíveis e não-atribuíveis (24%), porém as cardiovasculares destacaram-se em quantidade (27%) e envolveram os maiores gastos (37%). Os homens internaram menos (48%) que as mulheres, porém com gasto total maior (53%). As internações de pacientes entre 45-64 anos constituíram o maior grupo (45%) e gastos (48%) enquanto os pacientes com ≥75, os maiores coeficientes de hospitalização (350/104hab.) e de despesa (US$ 93,4 mil/104hab.). As regiões mais desenvolvidas gastaram o dobro (/104hab.) em relação às demais. Considerações Finais e Recomendações: As configurações no consumo de serviços hospitalares foram semelhantes às de países mais desenvolvidos, com importantes desigualdades regionais e de sexo. O gasto governamental exclusivamente com hospitalizações atribuíveis ao DM foi expressivo (2,2% do orçamento do Ministério da Saúde). A ampliação de atividades preventivas poderia diminuir a incidência do DM, reduzir a necessidade de internações, minimizar as complicações e minorar a severidade de outras condições médicas mais gerais. / Background: Diabetes mellitus (DM) is one of the main causes of morbi/mortality in western societies due to the burden of suffering, disabilities, loss of productivity and premature death that encompasses. Its economic impact is unknown in Brazil. Objectives: To dimension the share of DM hospitalizations on the Brazilian national health system (1999-2001), helping on evaluating direct costs. Specifically, to analyze hospitalizations (327.800) and hospitalization deaths (17.760) caused by DM as first-listed diagnosis (ICD-10 E10-E14 and procedure done) and to estimate the magnitude of DM attributable hospitalizations, including DM itself, chronic complications (CC) and general medical conditions (GMC). Methods: Data from the Hospital Information System of the National Health System (SIH/SUS) (37 millions of hospitalizations). Indicators were calculated by residence region of the patients and sex (adjusted by direct method for age with 95% confidence intervals), age intervals, average length of stay and expenditure by admission and population in US$. Multiple logistic regression was performed for death as outcome. Combinations of DM prevalence and hospitalization relative risks for CC and GMC were added to DM first-listed hospitalizations (attributable risk methodology). Sensitivity analyze was used for different prevalences and relative risks. Results: Hospitalizations and hospitalization deaths coefficients and lethality by DM as first-listed diagnosis were 6.4/104inhab., 34.9/106inhab. and 5.4% respectively. Coefficients were higher for women, although lethality was for men in every five region. Average expenditure (US$150,59) differed significantly between those with/without death but presented equal average length of stay (6.4). Population expenditure was US$ 969.09/104inhab. Odds-ratio for dying were larger for men, patients 75 yrs, and inhabitants of northeast and southeast. Hospitalizations attributable to DM were estimated at 836.3 thousand/year (49.3/104inhab.) reaching US$ 243.9 millions/year (US$ 14.4 thousand/104inhab.). DM as fist-list diagnosis (13.1%), CC (41.5%) and GMC (45.4%) depicted 6.7%, 51.4% e 41.9% respectively of annual expenditures. Average value of attributable hospitalizations (US$ 292) was 36% higher than non-attributable. Peripheral vascular diseases posed the largest excess based on average values (24%) although cardiovascular ones represented the major quantity (27%) and expenditure group (37%). Men were less admitted (48%) than women, but incurred more expenditure (53%). People 45-64 years old consisted the largest (45%) and most expensive (48%) group while 75+ generated the highest coefficients of hospitalization (350/104inhab.) and expenditure (US$ 93.4 thousands/104inhab.). Most developed regions accounted for nearly twice expenses than other regions. Conclusions and Recommendations: Patterns of hospitalization were similar to those most developed countries. Important regional and gender inequalities did exist. Governmental expenditures related exclusively to DM attributable hospitalizations was meaningful (2.2% of the budget of the Ministry oh Health). Broadening preventive health care actions could diminish the incidence of DM, reduce the need for hospitalizations, minimize complications and minors the severity of general medical conditions.