Risk of all-cause and CHD mortality in women versus men with type 2 diabetes: a systematic review and meta-analysis

Objective Previous studies have shown sex-specific differences in all-cause and CHD mortality in type 2 diabetes. We performed a systematic review and meta-analysis to provide a global picture of the estimated influence of type 2 diabetes on the risk of all-cause and CHD mortality in women vs men. Methods We systematically searched PubMed, EMBASE and Web of Science for studies published from their starting dates to Aug 7, 2018. The sex-specific hazard ratios (HRs) and their pooled ratio (women vs men) of all-cause and CHD mortality associated with type 2 diabetes were obtained through an inverse variance-weighted random-effects meta-analysis. Subgroup analyses were used to explore the potential sources of heterogeneity. Results The 35 analyzed prospective cohort studies included 2 314 292 individuals, among whom 254 038 all-cause deaths occurred. The pooled women vs men ratio of the HRs for all-cause and CHD mortality were 1.17 (95% CI: 1.12–1.23, I2 = 81.6%) and 1.97 (95% CI: 1.49–2.61, I2 = 86.4%), respectively. The pooled estimate of the HR for all-cause mortality was approximately 1.30 in articles in which the duration of follow-up was longer than 10 years and 1.10 in articles in which the duration of follow-up was less than 10 years. The pooled HRs for all-cause mortality in patients with type 2 diabetes was 2.33 (95% CI: 2.02–2.69) in women and 1.91 (95% CI: 1.72–2.12) in men, compared with their healthy counterparts. Conclusions The effect of diabetes on all-cause and CHD mortality is approximately 17 and 97% greater, respectively, for women than for men.


Introduction
Diabetes is recognized as the world's fastest growing chronic condition.Due to rapid increases in the prevalence of physical inactivity, overweight and obesity, the number of people with diabetes is projected to rise to 592 million by 2035 (1).In particular, type 2 diabetes (T2D) has attained the status of a global pandemic, with the total number of patients with T2D estimated at 425 million in 2015 (2).
Accumulating evidence documents T2D as an independent risk factor for all-cause mortality (3,4,5,6).The risk of all-cause mortality in persons with T2D is approximately doubled (7).However, these conclusions are mainly based on the assumption that the risk of diabetes in women is the same as in men (8).However, evidence is accruing that the detrimental effects of diabetes are higher among women than among men (9).The sex-based difference in the risk of diabetes would not only result from patient management and treatment (10) but also from the diversity of biological factors (11).Epidemiological studies reported that women with T2D had a higher risk of mortality from cardiovascular diseases (10,12) and cancer (13) than men.
However, the differences between men and women in the risk of all-cause and CHD mortality is still unclear.Moreover, none of the relevant meta-analyses summarized the differences in risk of all-cause and CHD mortality between men and women.Therefore, we performed a comprehensive meta-analysis to estimate reliably the effect of T2D on all-cause and CHD risk among women in comparison with men.

Search strategy
We systematically searched the PubMed (www.ncbi.nlm.nih.gov),EMBASE and Web of Science databases (from their starting dates to Aug 7, 2018) with the limitations object human and language English.A combined text word and medical subject heading (MeSH) search strategy was applied with the terms 'mortality', 'death', 'Diabetes Mellitus, Type 2', 'Adult-Onset Diabetes', 'Non-Insulin-Dependent Diabetes', 'Gender', 'Sex', 'Cohort', 'Prospective' and 'Longitudinal'.We also scanned the reference lists of relevant reviews and meta-analyses to discern additional potentially relevant literature.

Inclusion and exclusion criteria
We included articles only when they had clearly reported hazard ratios (HRs) or equivalents for all-cause or CHD mortality in both genders (T2D patients vs healthy counterparts).We also included articles that did not report HRs for each gender directly, but from which we could calculate it.Studies in which the enrolled participants had stroke, coronary heart disease or other cardiovascular diseases (myocardial infarction, atherosclerosis etc.) were excluded.If more than two articles had been published about the same cohort, we enrolled the one with the longer follow-up period or a larger sample size.The search strategy and inclusion criteria were defined and agreed upon by all the authors.The quality of the included studies was evaluated by the NOS (Newcastle-Ottawa Scale) (14) (Supplementary data, see section on supplementary data given at the end of this article).Our meta-analysis was performed in accordance with the PRISMA statement (15) and registered at the International Prospective Register of Systematic Reviews (Prospero) (http://www.crd.york.ac.uk /PROSPERO, registration number: CRD42017074187).

Data extraction
For each study, we extracted the following variables: name of first author and study, baseline years of study, country of study, duration of follow-up, mean ages of participants, sample size, death count, adjusted variables, HRs and their 95% CIs in men and women and NOS score.Two authors (Guodong Xu and Dingyun You) independently extracted the data.If there was controversy, the discrepancy was resolved by an arbitrator (Liyuan Han).

Statistical analysis
We extracted gender-specific HRs and 95% CIs from each study (T2D patients vs healthy counterparts).Subsequently, gender-specific HRs and 95% CIs were used to estimate the pooled ratio of HR and the corresponding 95% CIs.Subgroup analyses were performed by year of study baseline (before 1980, 1980 to 1990 and after 1990); region (America, Europe, Asia, and Australia, Canada, New Zealand or Pacific); duration of study (<10, 10-14, >14 years); study quality (NOS score) (≥6 vs <6) (14) and adjusted status (unadjusted vs adjusted).Sensitivity analysis was conducted to ascertain the stability of the pooled results after removing one study at a time.The I 2 value was used to estimate heterogeneity.An I 2 value of 25, 50 and 75% represented a low, middle and high degree of heterogeneity, respectively (16).Metaregression analyses were also performed to estimate the source of heterogeneity.We used funnel plots to estimate publication bias.Egger's and Begg's test were also applied to quantitatively estimate publication bias.Additionally, to explore the possible effect of publication bias, we employed trim-and-fill method (17) in our meta-analyses for more reliable estimates.All P values were two sided and P values less than 0.05 were considered as statistically significant.Software Stata 12.0 (StataCorp) was used to perform statistical analyses.

Study characteristics
We systematically searched PubMed, EMBASE and Web of Science (from their starting dates to Aug 7, 2018).A total of 3907 articles were identified by assessment of titles and abstracts, and eight additional records were identified from the reference lists (Supplementary Fig. 1).

Subgroup analysis
We used subgroup analysis to explore the sources of heterogeneity.The year in which the study began did not explain the possible heterogeneity in our study (P = 0.072) (Fig. 3).Different quality scores between articles also did not account for the heterogeneity (P = 0.388).The presence or absence of adjustment for confounding factors in the articles also not the cause of heterogeneity (P = 0.729).The different regions and follow-up duration explained some of the heterogeneities (P = 0.035 and P = 0.014, respectively).
However, the result of meta-regression indicated that the different regions had no effect on the pooled estimate of the HR (All P > 0.05), and study duration less than 10 years may be a source of heterogeneity (P = 0.035).The pooled estimate of the HR was 1.30 (95% CI: 1.12-1.52) in the articles with a follow-up duration longer than 15 years, compared with 1.33 (95% CI: 1.15-1.53)and 1.10  Pooled women-to-men ratios of HRs for all-cause mortality, comparing people with type 2 diabetes vs those without the disorder.https://eje.bioscientifica.com(95% CI: 1.05-1.15) in those with a follow-up duration between 10 and 15 years and less than 10 years, respectively.

Sensitivity analysis and publication bias
Sensitivity analysis did not change the results of this metaanalysis.The Begg's funnel plot was used to estimate the potential for publication bias (Supplementary Fig. 2).The result of Begg's test (P = 0.001) indicated the possibility of publication bias in all-cause mortality.The trim-andfill method was applied to test and adjust for publication bias, which conservatively hypothesized publication bias is the reason for funnel plot asymmetry.Some negative unpublished studies were used to produce a symmetrical funnel plot (Supplementary Fig. 2), which continued to reveal a statistically significant sex-specific association between T2D and all-cause mortality (HR: 1.08.(95% CI: 1.03-1.13)).

Discussion
In this meta-analysis of 35 prospective cohort studies, which included data for more than 2 314 292 individuals and 254 038 all-cause mortality events, T2D was demonstrated as one of the risk factor for allcause and CHD mortality in both women and men.Diabetes-related mortality was higher in women than men, and the effect of diabetes on all-cause and CHD mortality was 17 and 97% higher in women than men, respectively.

NOTE: Weights are from random effects analysis
Overall (I-squared = 86.4%,p<0.001) Finland diabetes study 54   MESA 57   Diabetes Melanesian Fijians Cohort study 26   Swedish Annual Level of Living Survey 25 JACC study 58   The Rancho Bernardo Study 64   Diabetes Asian Indian Cohort study 26   The Adventist Health Study 66   The Reykjavik Study 60

Study
GeneSTAR 57   Finnish Diabetes study 56   Finland national hospital discharge register 55 Finland diabetes study 29   Hawaii-Los Angeles-Hiroshima study 65   NHANES I 62   NHANES III 57   Framingham heart study 63   South Tees Diabetes Mortality Study 43   ET-CHD Registry Similarly, the effect of type 1 diabetes on mortality was 37% higher in women than men (67).In a recent collaborative meta-analysis, diabetes was associated with all-cause mortality, and the relative risks were 1.59 in men and 2.00 in women, respectively (68).It is not clear why the rate of diabetes-related death was higher in women than in men, but several mechanisms could help to explain.One hypothesis is that the excess risk of mortality in women is due to the combined effect of greater deterioration and more prolonged exposure to cardiovascular risk profile during their prediabetic period (12,67,69).Due to their poorer glycemic control, women with diabetes had an overall greater cumulative lifetime exposure to hyperglycemia (12).Several studies have suggested that women with diabetes had higher BMI (70,71,72) and were more insulin resistant (72) than their men counterparts.And they also had significantly higher blood pressure and lipid levels than in men (12).
Research indicates that men with diabetes were diagnosed earlier than women (73), as the early symptoms of diabetes in man may be more likely to be recognized by physicians (74).Men with diabetes may receive better therapeutic interventions and more comprehensive care (75,76,77).Alternatively, sex differences in the management and treatment of diabetes may play a crucial role in the disparity in mortality rates between men and women (74).In addition, men with diabetes are more likely to use aspirin (78), which was proven to decrease the risk of stroke and myocardial infarction (79,80).Furthermore, it has been reported that more diabetic men than women received recommended care processes (62 vs 58%) (77).Notably, even under the same treatment regimen, women were less likely than men to achieve treatment targets for controlling mortality risk factors (81,82).Moreover, it has been reported that women were less likely to achieve glycemic targets with insulin glargine and exhibited significantly less reductions in fasting blood glucose levels (83).Previous research also observed that women were more likely to experience hypoglycemia during insulin treatment (84,85).Therefore, differences in treatment and management may explain a large component of the excess risk associated with diabetes in women.In addition, women have less stroke risk factors compared with men (86), so the effect of adding one risk factor (such as diabetes) on women may be more serious.
The other potential mechanisms for sex-specific differences in mortality may result from the differences in biological factors.A recent study (87) suggested that diabetic women had higher levels of endogenous testosterone, which could predict incident CHD risk (88,89).Women with diabetes also had a greater change in insulin resistance than men (90).Mansfield and colleagues also found sex-based differences in the level of coagulation and fibrinolysis in individuals with diabetes (91,92) and reported that factor VII and plasminogen activator inhibitor 1 activity levels were significantly higher in women than in men, contributing to the increased cardiovascular risk.Furthermore, higher levels of adiponectin were associated with all-cause mortality in people with T2D (92,93), and diabetic women were found to have higher levels of adiponectin (92).
The large sample size is one of the strengths of this meta-analysis.We are also the first study to estimate reliably the effect of T2D on CHD mortality risk among women in comparison with men.Additionally, the included studies were limited to prospective cohort studies, which eliminated the possible recall and selection bias.The subgroup and sensitivity analysis was used to explore the possible heterogeneity and ensure the reliability of the results.The trim-and-fill method was applied to adjust the potential publication bias.For the quality control of this meta-analysis, we also registered it at Prospero and performed the study in accordance with the PRISMA statement.
However, there were several limitations in our metaanalysis.Firstly, the standard definition of diabetes and https://eje.bioscientifica.comconfounding variables adjusted varied across studies, which may have resulted in inconsistent estimation of mortality risks.Secondly, the follow-up duration of T2D was not directly reported in some studies.Moreover, in most studies, diabetic status was mainly based on selfreport or past medical history; therefore, there was a higher probability of underestimation of the number of patients with T2D.Although we performed a range of sensitivity analyses, we were also unable to explain most of the heterogeneity among the studies for the outcome of all-cause mortality.In addition, some articles lacked specific data on patient's age; therefore, we could not perform age-specific subgroup analysis.
Taken together, we found that the relative effect of diabetes on all-cause and CHD mortality was significantly greater in women than in their men counterparts.For future, we should avoid sexual prejudice in diabetes, take all necessary steps to diagnose early and control risk factors comprehensively to guarantee the most suitable treatments in women patients.Besides, it is necessary to perform further studies to determine the actual mechanisms that account for sex-based difference in diabetes-related mortality risk.

Figure 3 Forest
Figure 3Forest plot of the subgroup analyses with pooled women-tomen ratios of HRs for all-cause mortality.HR, hazard ratio.

% deaths in women Ascertainment of diabetes Variables used to standardize HR Causes of death NOS score
https://eje.bioscientifica.com
Chronic Diseases: Surveillance System; DISS, Diabetes Incidence Study in Sweden; ET-CHD, Eastern Taiwan integrated health care delivery system of Coronary Heart Disease; FRESCO, Función de Riesgo ESpañola de acontecimientos Coronarios y Otros; GPRD, General Practice Research Database; GR1C, Gila River Indian Community; HR, hazard ratio; MSVDD, Multinational Study of Vascular Disease in Disease in Diabetes; NHANES, the National Health and Nutrition Examination Survey; NHISNSC, National Health Insurance Service-National Sample Cohort.