Effect of diabetes duration and glycaemic control on 14-year cause-specific mortality in Mexican adults: a blood-based prospective cohort study

Abstract

Background: Diabetes is a cause of at least a third of all deaths in Mexican adults aged 35-74 years, with the excess mortality due mainly to vascular disease, renal disease, infection, and acute diabetic crises. We aimed to analyse the effect of diabetes duration and glycaemic control on death rate ratios (RRs) for these causes and to assess the relevance to cause-specific mortality of undiagnosed diabetes.

Methods: About 100 000 women and 50 000 men aged 35 years or older from Mexico City were recruited into a blood-based prospective study between April 14, 1998, and Sept 28, 2004, and followed up until Jan 1, 2016, for cause-specific mortality. Participants who, at recruitment, reported any chronic disease other than diabetes and those who had missing data for HbA_1c or diabetes duration were excluded. We used Cox models to estimate the associations of undiagnosed or previously diagnosed diabetes (almost all type 2) with risk of mortality from vascular disease, renal disease, and infection, exploring among those with previously diagnosed diabetes the independent relevance of diabetes duration (<5 years, ≥5 to <10 years, or ≥10 years) and HbA_1c (<9%, ≥9% to <11%, or ≥11%). We also estimated the association of HbA_1c with mortality in participants without diabetes at recruitment.

Findings: 133 662 participants were aged 35-74 years and had complete data and no other chronic disease. 16 940 (13%) had previously diagnosed diabetes, 6541 (5%) had undiagnosed diabetes, and 110 181 (82%) had no diabetes. Among participants with previously diagnosed diabetes, glycaemic control was poor (median HbA_1c 8·9% [IQR 7·0-10·9]), and was worse in those with longer duration of disease at recruitment. Compared with participants without diabetes, the death RRs at ages 35-74 years for the combination of vascular, renal, or infectious causes were 3·0 (95% CI 2·7-3·4) in those with undiagnosed diabetes, 4·5 (4·0-5·0) for the 5042 participants with a diabetes duration of less than 5 years, 6·6 (6·1-7·1) for the 7713 participants with a duration of 5 years to less than 10 years, and 11·7 (10·7-12·7) for the 4185 participants with a duration of at least 10 years. Similarly, the death RRs were 5·2 (4·8-5·7) for those with HbA_1c less than 9%, 6·8 (6·2-7·4) for those with HbA_1c of 9% to less than 11%, and 10·5 (9·7-11·5) for those with HbA_1c of at least 11%. Diabetes was not strongly associated with the combination of deaths from other causes apart from acute glycaemic crises. Among participants without diabetes, higher HbA_1c was not positively related to mortality.

Interpretation: In Mexico, the rates of death from causes strongly associated with diabetes increased steeply with duration of diabetes and were higher still among people with poor glycaemic control. Delaying the onset of type 2 diabetes, as well as improving its treatment, is essential to reduce premature adult mortality in Mexico.

Funding: Wellcome Trust, the Mexican Health Ministry, the Mexican National Council of Science and Technology, Cancer Research UK, British Heart Foundation, and the UK Medical Research Council Population Health Research Unit.

Figure 1

Relevance of previously diagnosed and undiagnosed diabetes to mortality from vascular, renal, or infectious causes, by age and sex (A), duration of diabetes (B), and glycaemic control (C) (A) Death rate ratios (RRs) by age and sex for deaths at ages 35–84 years, for patients with diagnosed diabetes versus those with no diabetes. Diamonds show values for men and women combined. The RRs for participants with undiagnosed diabetes (ie, no previous diagnosis but baseline HbA_1c ≥6·5%) compared with participants without diabetes were 4·0 (95% CI 3·3–4·9) at ages 35–59 years, 2·6 (2·3–3·0) at ages 60–74 years, and 1·4 (1·2–1·6) at ages 75–84 years, and were similar in men and women. (B) Death RRs by duration of diabetes, at ages 35–74 years. (C) Death RRs by glycaemic control, at ages 35–74 years. RRs in all panels are adjusted for age at risk, smoking status, district, educational level, height, weight, and waist and hip circumferences. In (B) and (C), RR estimates are additionally adjusted for sex, and the estimates for those with previously diagnosed diabetes are also adjusted, respectively, for any HbA_1c or diabetes duration differences between the groups (to the average HbA_1c or duration seen for all those with previously diagnosed diabetes) in such a way that their information-weighted average equals the overall RR estimate for all those with previously diagnosed diabetes versus those with no diabetes. The numbers above the squares are the RRs and the numbers below the squares are the number of deaths in that group. In all panels, the size of each square is proportional to the amount of statistical information.

Figure 2

Relevance of previously diagnosed and undiagnosed diabetes to mortality from vascular (A), renal (B), and infectious (C) causes at ages 35–74 years, by duration of diabetes Rate ratios (RRs) are adjusted for age at risk, sex, smoking status, district, educational level, height, weight, and waist and hip circumferences, as well as HbA_1c for participants with previously diagnosed diabetes. The numbers above the squares are the RRs and the numbers below the squares are the number of deaths in that group. In all panels, the size of each square is proportional to the amount of statistical information.

Figure 3

Relevance of previously diagnosed and undiagnosed diabetes to mortality from vascular (A), renal (B), and infectious (C) causes at ages 35–74 years, by glycaemic control Rate ratios (RRs) are adjusted for age at risk, sex, smoking status, district, educational level, height, weight, and waist and hip circumferences, as well as diabetes duration for participants with previously diagnosed diabetes. The numbers above the squares are the RRs and the numbers below the squares are the number of deaths in that group. In all panels, the size of each square is proportional to the amount of statistical information.