Conclusions
In this large prospective cohort of adults from Mexico City, previously diagnosed diabetes was associated with greater than fourfold higher risk of premature death from any infectious cause between 35 and 74 years of age. Death rates in diabetes were highest for urinary tract infection, skin, bone and connective tissue infection and septicemia, with up to 10-fold higher risks than participants without diabetes. Moreover, infectious disease mortality risks were higher among those with longer duration of diabetes or with higher baseline HbA1c levels. In this population with a high prevalence of frequently poorly controlled diabetes, the condition accounted for approximately one-third of all premature deaths due to infectious diseases.
The diabetes-associated risks of mortality due to infectious diseases observed in the present study are generally more extreme than those reported previously. Although infection-related mortality rates among patients with type 2 diabetes recruited from an outpatient clinic in Brazil were found to be six times higher than in the general adult population, this was based on a very small study population (n=471) and residual confounding may have explained some of the excess risk, since rates were standardized only for age and sex.12 In contrast, large-scale studies comprising populations from predominantly high-income countries have typically reported a more modest doubling of the risk of death due to infectious causes.5 7 15 23 For example, in the Emerging Risk Factors Collaboration’s individual participant data meta-analysis of 40 000 participants with previously or newly diagnosed diabetes and 675 000 participants without diabetes followed for an average of 14 years, diabetes was associated with a 2.4-fold higher risk of death due to all infections excluding pneumonia (n=1081) and a 1.7-fold higher risk of death due to pneumonia (n=2893).5 Differences between studies in the definitions of diabetes used may have contributed to differences in reported risks, as illustrated by the lower risks in undiagnosed, than diagnosed, diabetes in the present study. However, these would not explain the magnitude of differences observed, which could also reflect the influence of glycemic control or differing age distributions of study populations. Moreover, disparities between countries in access to healthcare interventions (eg, newer generation antibiotics, intensive care) with potential to differentially impact on population subgroups with greater susceptibility to severe infections, including individuals with diabetes,13 may have contributed to the comparably high diabetes-associated infectious disease mortality risks in this Mexican study population.
There is limited evidence available on the relevance of glycemic control for infectious disease mortality.13 24 Studies examining infectious disease outcomes more generally (ie, both fatal and non-fatal infections) have shown mixed findings, but many have found J-shaped or U-shaped associations of HbA1c levels with risk of both infectious disease incidence and mortality.10 24–26 For example, apparent J-shaped associations were observed between HbA1c levels and risks of both hospitalization for, and death from, infection among 85 000 patients with diabetes recorded in English primary care data, with lowest risks observed at levels of 6%–7%.24 In contrast, we found a strong positive association of baseline HbA1c levels with risk of death due to infectious causes among participants with previously diagnosed diabetes, with no apparent threshold in the association. The absence of higher risks at the lowest HbA1c levels in the present study likely reflects exclusion of participants with prior chronic diseases (in contrast with the other studies described), reducing potential for reverse causality and residual confounding, while the comparably clear positive association at higher HbA1c levels may reflect relatively poor glycemic control (mean HbA1c was 9.1% in the Mexico City study population, but 7.4% and 8.3% among participants with type 2 and type 1 diabetes, respectively, in the English study population24). Higher glucose concentrations may increase the risk of infectious diseases, as well as contributing to adverse outcomes following infection, through multiple mechanisms. These include impaired immune function, adversely impacting both humoral and cell-mediated immunity, and promotion of the growth of some microorganisms.27 However, trials of intensive glycemic control in diabetes have generally not investigated the impact on infections,14 and better evidence is needed to understand whether the association is causal. Evidence is also limited on the relevance of non-diabetic glycemia for infectious disease risk.28 Although we observed a higher risk of death due to any infectious cause among participants with pre-diabetes than among those with HbA1c levels in the truly ‘normoglycaemic’ range, this may simply reflect subsequent development of diabetes and associated infectious disease mortality risks in this group.
Duration of diagnosed diabetes showed a strong positive association with the risk of mortality due to infectious causes in the Mexico City population, independent of its association with glycemic control. This is consistent with the observed lower risks among individuals with undiagnosed, than with diagnosed, diabetes, despite similar mean HbA1c levels. Given the average 9-year duration of diabetes diagnosis, which is higher than in several previous studies,10 29 this may have contributed to the higher diabetes-associated risks of mortality due to infectious causes in this study population. Both longer diabetes duration11 and higher levels of glycemia30 31 are established risk factors for vascular complications of diabetes, which previous studies suggest may play a role in determining infectious disease risks and prognosis.15 16 In contrast, the associations presented herein differed little according to participants’ history of cardiovascular (ie, macrovascular) diseases. However, we were unable to account for macrovascular complications developed during follow-up in the study, or for microvascular complications. Few studies have simultaneously examined the association of diabetes with risks of death due to the full range of site-specific infections. However, this has been explored for infection-related outcomes more generally (including both primary care recorded diagnoses and hospitalizations),10 16 27 29 in many instances showing the highest diabetes-associated risks for the same, or closely related, infections as in the present study, although more modest than those observed in this Mexico City population.16 27 29 The notably strong associations of diabetes, glycemic control and diabetes duration with risk of death due to skin, bone and connective tissue infection may, at least in part, reflect the influence of chronic microvascular and macrovascular complications of diabetes, as well as the predominance of bacterial infections at these sites,29 and similar factors may explain the comparatively strong associations with death due to urinary tract infections and septicemia.
In the present study, we estimated that one-third of infectious disease deaths before age 75 could be attributed to diabetes. This clearly highlights the need for efforts to prevent infectious diseases among this Mexican population, particularly given the high prevalence of diabetes and poor glycemic control. Despite lower diabetes-associated relative mortality risks, respiratory infections accounted for the greatest proportion of infectious disease deaths among individuals with diabetes in the present study, highlighting the potential value of vaccination against respiratory pathogens in this population. Although pneumococcal and influenza vaccinations are recommended for all adults with diabetes in Mexico,32 uptake is reported to be low,33 but there is clear value in ensuring effective implementation of these existing vaccination guidelines.
Our study has certain limitations. First, diabetes may influence both the onset of infectious diseases and their course, and the focus on mortality prevented differentiation between these. However, mortality outcomes would be expected to be less susceptible to misclassification and to potential diagnostic biases, and readily permitted investigation of a wide spectrum of infectious diseases. Assessment of the relevance of glycemic control for infectious disease mortality risks was based on single HbA1c measurements, which may not reflect longer term trends. However, there is arguably clinical value in understanding the relevance of single measurements for future infectious disease mortality risks. More women than men were recruited into the study (because women were more likely to be at home when the fieldworkers’ visit was during standard working hours). However, the size of the study meant that large numbers of deaths were observed in both men and women, leading to reliable sex-specific estimates. The two study districts are also not representative of the overall Mexican population, or even the overall Mexico City population. However, prospective studies of non-representative cohorts of individuals can provide reliable evidence about the associations of risk factors with disease that are widely generalizable.34 35 Finally, the observational study design precludes assessment of the likely causality of the observed associations.
Diabetes is highly prevalent in Mexico and is associated with very high risks of infectious disease mortality, particularly among those with longer duration of diabetes diagnosis and poorer glycemic control. The findings presented clearly highlight the need for an increased focus on prevention of infectious diseases in the care of individuals with diabetes, including through effective implementation of existing vaccination policies, with potential for significant reductions in premature mortality. Moreover, prevention (or delay) of diabetes onset, including through prevention and management of the high levels of adiposity in the Mexican population, will be essential for reducing diabetes-associated infectious disease mortality.