OBJECTIVE

Oxidative stress plays an important role in the pathophysiology of type 2 diabetes mellitus (T2DM). However, associations of biomarkers of oxidative stress with diabetes complications have not yet been addressed in large cohort studies.

RESEARCH DESIGN AND METHODS

Derivatives of reactive oxygen metabolites (d-ROMs) levels, a proxy for the reactive oxygen species burden, and total thiol levels (TTLs), a proxy for the reductive capacity, were measured in 2,125 patients with T2DM from two German cohort studies of almost equal size at baseline and 3–4 years later. Multivariable adjusted Cox proportional hazards models with time-dependent modeled d-ROMs levels and TTLs were used to assess the associations with incident major cardiovascular events (MCE), cancer incidence, and all-cause mortality.

RESULTS

In total, 205, 179, and 394 MCE, cancer, and all-cause mortality cases were observed during 6–7 years of follow-up, respectively. Both oxidative stress biomarkers and the d-ROMs-to-TTL ratio were statistically significantly associated with all-cause mortality in both cohorts, and the pooled hazard ratios (HRs) and 95% CIs for top versus bottom tertiles were 2.10 (95% CI 1.43, 3.09) for d-ROMs levels, 0.59 (0.40, 0.87) for TTLs, and 2.50 (1.86, 3.36) for d-ROMs-to-TTL ratio. The d-ROMs-to-TTL ratio was also statistically significantly associated with incident MCE for top versus bottom tertile (1.65 [1.07, 2.54]), but this association did not persist after additional adjustment for chronic diseases. No associations with cancer were detected.

CONCLUSIONS

The observed strong associations of both biomarkers with mortality suggest an important contribution of an imbalanced redox system to the premature mortality of patients with diabetes.

Previous studies have consistently shown increased levels of oxidative stress in subjects with type 2 diabetes mellitus (T2DM) (1). This could in part be explained by the generally higher caloric intake of the generally more obese individuals with T2DM, resulting in a higher mitochondrial energy synthesis rate and subsequently production of reactive oxygen species (ROS) (1). Furthermore, hyperglycemia can induce oxidative stress by multiple mechanisms (1,2). Hyperglycemia and associated metabolic abnormalities of obese people can cause an increased production of mitochondrial superoxide molecules in endothelial cells of both large and small vessels, as well as in the myocardium (3,4). Thus, oxidative stress may contribute to microvascular and macrovascular diabetes complications (5). Furthermore, higher oxidative stress of subjects with T2DM may also explain the observed higher cancer incidence and shorter life expectancy in subjects with T2DM than in diabetes-free individuals of comparable age (6).

Nevertheless, large cohort studies linking biomarkers of oxidative stress to diabetes complications in humans are sparse because ROS are highly reactive, have a very short half-life, and, thus, are difficult to measure directly in biological samples (1). Recently, assays that indirectly measure ROS burden and control have been developed and operationalized for high-throughput measurement techniques: derivatives of reactive oxygen metabolites (d-ROMs) and total thiol levels (TTLs). d-ROMs can be regarded as a proxy for ROS production (7) and TTL as a proxy for the redox control status of blood (8). The d-ROMs assay detects hydroperoxide metabolites (chemical: R-O-O-H), mainly of lipids, but also of glycosides, amino acids, and proteins in the serum sample (9). The TTL assay detects free thiol groups of the amino acids cysteine or methionine (chemical: R-S-H), which can be reversibly oxidized to disulfide bridges (chemical: R-S-S-R) (10). The most frequent molecules in human blood with free thiol groups are albumin, glutathione, α-lipoic acid, and members of the thioredoxin protein family (11). These molecules act as an antioxidant defense system by their ability to oppose the propagation phase of the peroxidation processes in order to maintain intracellular redox environment, which plays a key role in regulating endothelial cell function (12).

A recent application of these assays in large population-based cohort studies showed that d-ROMs levels and TTLs were independent predictors of mortality (11). The aim of this investigation is to explore the associations of serum d-ROMs and TTL with incident major cardiovascular events (MCE), total cancer incidence, and cause-specific and all-cause mortality in patients with T2DM from two large-scale cohort studies from Germany and show results summarized by meta-analyses.

Study Population and Data Collection

This investigation is based on the diabetes subcohort of the 8-year follow-up of the population-based ESTHER cohort (German name: “Epidemiologische Studie zu Chancen der Verhütung, Früherkennung und optimierten Therapie chronischer Erkrankungen in der älteren Bevölkerung”) and the DIANA cohort of patients with T2DM (German name: “Diabetes Mellitus: Neue Wege der Optimierung der allgemeinärztlichen Betreuung”).

The ESTHER cohort is an ongoing population-based cohort study, conducted in the federal state of Saarland, located in Southwest Germany. Details of the study design have been reported previously (13). Briefly, 9,940 male and female residents of Saarland, 50–75 years of age and with a sufficient knowledge of the German language were recruited in 2000–2002 by their general practitioners (GPs) during a general health checkup. The distribution of major sociodemographic characteristics and the prevalence of major diseases of the ESTHER study population closely resembled that of a representative German National Health Survey of 1998 within the corresponding age range (13). The participants have been recontacted ∼2, 5, 8, 11, and 14 years after baseline so far. In the current investigation, the fourth ESTHER contact with 6,018 study participants (8 years after baseline recruitment; 2008–2010) was used as the baseline. We measured the d-ROMs levels and TTLs at the fourth contact in 4,345 participants who donated blood samples. Of these study participants, 3,171 donated another blood sample at the fifth contact, 3 years later (2011–2013), during which we measured the two biomarkers once more. However, in this investigation, we only included study participants with physician-diagnosed diabetes (ascertained by medical records of the GP or antidiabetic medication reported by the GP or study participant). We excluded persons with diabetes diagnosed at the age of ≤40 years in order to focus on patients with T2DM (14). The prevalence of T2DM at the fourth ESTHER contact was 23.7% and n = 1,029 patients with T2DM with d-ROMs and TTL measurements could be included in this analysis (among whom 720 had repeated mea-surements).

The DIANA study is a prospective cohort study of patients with T2DM conducted in the Ludwigsburg-Heilbronn area, located in South-West Germany (15,16). The study was initiated to address (short- and long-term) diabetes-related outcomes and to evaluate potentials for health care improvements in people with T2DM. Overall, 1,146 ptients with a physician-diagnosed T2DM, ≥18 years of age, were recruited consecutively according to a standardized protocol by 38 GPs during regular practice visits between October 2008 and March 2010. Follow-up contacts have been performed after 4 and 7 years so far. The GPs were explicitly asked to recruit only patients with T2DM and no patients with type 1 diabetes. Notwithstanding, we excluded seven persons younger than 40 years from the cohort, which could potentially be patients with type 1 diabetes (14). The d-ROMs and TTL measurements could be performed for 1,096 baseline study participants, and d-ROMs were again measured for 738 participants of the 4-year follow-up. Repeated TTL measurements were not conducted in the DIANA cohort due to a lack of funding.

Ethics, Consent, and Permissions

The ESTHER study has been approved by the responsible ethics committees of the Medical Faculty of the University of Heidelberg and the Medical Association of Saarland. The DIANA study was approved by the ethics committees of the Medical Faculty of the University of Heidelberg and of the Chamber of Physicians of Baden-Württemberg. Written informed consent was obtained from all subjects, and the studies are being conducted in accordance with the Declaration of Helsinki.

Oxidative Stress Serum Marker Measurements

For the ESTHER study, the d-ROMs test from Diacron (Grossetto, Italy) and the Total Thiol assay from Rel Assay Diagnostics (Gaziantep, Turkey) were used, and measurements were performed at two time points (June 2012 and September 2014) because funding was not available to measure all samples in June 2012. To correct for potential shifts in the assays, 80 samples from time point 1 were measured again at time point 2, and linear regression equations were obtained and used to standardize the results from time point 2 to the levels of time point 1. Agreement of d-ROMs measurements was high (Spearman correlation coefficient, r = 0.92; difference of means = 15 Carratelli Units [Carr U]), and therefore no corrections were applied. For TTLs, an obvious shift in the assay results (r = 0.88; difference of means = 81 μmol/L) was corrected by applying the following equation to all measurements of time point 2: TTLcorrected = 0.746 × TTLuncorrected.

For the DIANA study, the d-ROMs test and –SHp test by Diacron were used to measure d-ROMs concentration and TTL, respectively, and all measurements were performed in February and March 2015. Serum samples from the two cohorts were delivered to the Laboratory for Health Protection Research (Bilthoven, the Netherlands) where the d-ROMs assay and –SHp assay were adapted to an automatic analyzer (UniCel DxC 800; Beckman-Coulter, Woerden, the Netherlands), as described previously (17).

Covariate Assessment

In both studies, information on sociodemographic factors, lifestyle factors, and year of diabetes diagnosis was taken from the participant’s questionnaire, and information on the medical history was taken from the physicians’ questionnaires. Diabetes medication information was obtained from the study participants’ and physicians’ questionnaires. The laboratory methods used to measure HbA1c from full blood as well as to measure creatinine, C-reactive protein (CRP), and total and HDL cholesterol from serum samples in both cohorts are shown in Supplementary Table 1.

Ascertainment and Definition of Outcomes

In both cohorts, deaths were obtained by inquiry at the residents’ registration offices and information about the vital status of almost all study participants could be obtained. Additionally, death certificates were provided by local health authorities for deceased study participants. Mortality follow-up was conducted until the end of 2015 in the ESTHER study and until the end of 2017 in the DIANA study.

Cancer incidence was reported by the Saarland cancer registry in the ESTHER study for the time until the end of 2014. In the DIANA study, cancer incidence was reported by physicians or has been ascertained by cancer deaths ascertained in the mortality follow-up. In a few cases, in which no information from the physicians could be obtained, self-reported information from the study participants was used. All cancers (ICD-10 codes I00–I99) except for nonmelanoma skin cancers (C44) were considered for the outcome “total cancer incidence.”

The composite end point “MCE” consisted of the three end points myocardial infarction (MI), stroke, and cardiovascular mortality. In both studies, self-reported MI and stroke cases were validated by medical records from the study participants’ physicians. Underlying causes of death with ICD-10 codes I00–I99 were classified as cardiovascular deaths. For the ESTHER study, the fifth (3 years after the baseline of this analysis) and sixth ESTHER contact (6 years after the baseline for this analysis) were used to ascertain incident MCE events. The follow-up of the DIANA study was longer because recontacts were conducted 4 and 7 years after baseline. In the ESTHER study, the median follow-up times for MCE, cancer, and all-cause mortality were 5.6, 5.3, and 6.3 years, respectively, and in the DIANA study the median follow-up times were 7.4, 7.4, and 8.5 years, respectively.

Statistical Analyses

Differences in median serum d-ROMs levels and TTLs according to characteristics of the study population were assessed with the Wilcoxon rank sum test. Cox proportional hazards models were used to derive hazard ratios (HRs) and 95% CIs for the association of d-ROMs level, TTL, and the d-ROMs-to-TTL ratio with incident MCE, total cancer incidence, and all-cause and cause-specific mortality. The ratio of the two biomarkers was also used in the analyses because their correlation was low (Spearman correlation coefficients in the ESTHER and DIANA cohort were −0.09 and −0.15, respectively). In analyses for MCE and cancer, subjects with a history of events before baseline were excluded, and a competing risks analysis was performed modeling the competing risk of death. d-ROMs levels and TTLs were modeled time dependently to make use of repeated measurements during follow-up. Three statistical models were used with an increasing level of adjustment. Model 1 adjusted for the potential confounders age, sex, HbA1c, BMI, education, smoking, alcohol consumption, physical activity, and diabetes medication. Model 2 was further controlled for potential intermediating conditions, such as history of MCE, history of coronary heart disease (CHD), history of heart failure, history of hypertension, time since diabetes diagnosis, dyslipidemia (assessed by total and HDL cholesterol), and renal function (measured with serum creatinine). Model 3 was additionally adjusted for the inflammatory marker CRP, which is highly correlated with d-ROMs and TTL. Model 1 was used to show the main results because it is not overadjusted by potential intermediate factors or highly correlated factors. Covariates were modeled categorically with the definitions shown in Table 1, except for age, which was used as a continuous variable. A priori defined subgroup analyses were conducted for subgroups stratified by age, sex, BMI, HbA1c, history of CHD, and history of cancer. In sensitivity analyses, d-ROMs and TTL were not modeled time dependently. The analyses were first performed separately for the ESTHER study and the DIANA study, and results were then pooled by random effects meta-analyses.

Table 1

Baseline characteristics of the ESTHER diabetes subcohort and the DIANA cohort

Baseline characteristicsCohorts, N (%)a
ESTHER (N = 1,029)DIANA (N = 1,096)
Age (years)   
 <60 55 (5.3) 202 (18.4) 
 60 to <70 434 (42.2) 312 (28.5) 
 70 to <80 466 (45.3) 455 (41.5) 
 ≥80 74 (7.2) 127 (11.6) 
Sex   
 Male 535 (52.0) 600 (54.7) 
 Female 494 (48.0) 496 (45.3) 
Educationb   
 Low 786 (78.4) 777 (73.3) 
 Medium 120 (12.0) 185 (17.5) 
 High 97 (9.7) 98 (9.3) 
BMI (kg/m2  
 <25 136 (13.8) 147 (13.4) 
 25 to <30 444 (44.9) 430 (39.3) 
 30 to <35 284 (28.7) 338 (30.9) 
 35 to <40 87 (8.8) 108 (9.9) 
 ≥40 37 (3.7) 72 (6.6) 
Smoking   
 Never smoker 490 (48.1) 528 (48.5) 
 Former smoker 441 (43.3) 433 (39.8) 
 Current smoker, 0–15 g tobacco/day 44 (4.3) 63 (5.8) 
 Current smoker, >15 g tobacco/day 43 (4.2) 64 (5.9) 
Alcohol consumptionc   
 Abstainer 436 (44.0) 400 (36.6) 
 Low consumption 509 (51.4) 625 (57.1) 
 Moderate consumption 32 (3.2) 47 (4.3) 
 High consumption 13 (1.3) 22 (2.0) 
Physical activityd   
 Inactive 549 (55.3) 265 (24.8) 
 Active 444 (44.7) 802 (75.2) 
Diabetes medication   
 None 335 (32.7) 252 (24.9) 
 Oral antidiabetics 495 (48.3) 539 (53.2) 
 Insulin 194 (19.0) 222 (21.9) 
Antihypertensive drugs   
 No 149 (14.6) 174 (17.2) 
 Yes 875 (85.4) 839 (82.8) 
Lipid-lowering drugs   
 No 522 (51.0) 562 (55.5) 
 Yes 502 (49.0) 451 (44.5) 
Anticoagulant drugs   
 No 518 (50.6) 618 (61.0) 
 Yes 506 (49.4) 395 (39.0) 
Time since diabetes diagnosis (years)   
 <5 554 (53.8) 357 (32.6) 
 5 to <10 244 (23.7) 297 (27.1) 
 10 to <20 170 (16.5) 345 (31.5) 
 ≥20 61 (5.9) 97 (8.9) 
HbA1c (%/mmol/mol)   
 <6/<42 222 (21.6) 151 (13.8) 
 6 to <7/42 to <53 482 (46.8) 572 (51.2) 
 7 to <8/53 to <64 249 (24.2) 241 (22.0) 
 ≥8/≥64 76 (7.4) 132 (12.0) 
Total cholesterol (mg/dL)   
 <200 377 (36.6) 387 (35.6) 
 200 to <280 557 (54.1) 561 (51.6) 
 ≥280 95 (9.2) 139 (12.8) 
HDL cholesterol (mg/dL)   
 <40 156 (15.2) 261 (24.0) 
 40 to <80 828 (80.5) 765 (70.4) 
 ≥80 45 (4.4) 60 (5.5) 
CRP (mg/L)   
 ≤3 596 (57.9) 651 (60.0) 
 >3 to ≤10 337 (32.8) 342 (31.5) 
 >10 96 (9.3) 92 (8.5) 
eGFR (mL/min/1.73 m2)e   
 ≥60 739 (71.8) 797 (72.7) 
 30 to <60 278 (27.0) 264 (24.1) 
 <30 12 (1.2) 35 (3.2) 
History of CHD   
 No 724 (70.4) 843 (81.5) 
 Yes 305 (29.6) 191 (18.5) 
History of MI   
 No 901 (87.6) 944 (91.4) 
 Yes 128 (12.4) 89 (8.6) 
History of stroke   
 No 914 (88.8) 971 (93.9) 
 Yes 115 (11.2) 63 (6.1) 
History of heart failure   
 No 813 (79.0) 897 (87.6) 
 Yes 216 (21.0) 127 (12.4) 
History of hypertension   
 No 292 (28.4) 221 (21.4) 
 Yes 737 (71.6) 812 (78.6) 
History of cancer   
 No 893 (86.8) 927 (89.9) 
 Yes 136 (13.2) 104 (10.1) 
Baseline characteristicsCohorts, N (%)a
ESTHER (N = 1,029)DIANA (N = 1,096)
Age (years)   
 <60 55 (5.3) 202 (18.4) 
 60 to <70 434 (42.2) 312 (28.5) 
 70 to <80 466 (45.3) 455 (41.5) 
 ≥80 74 (7.2) 127 (11.6) 
Sex   
 Male 535 (52.0) 600 (54.7) 
 Female 494 (48.0) 496 (45.3) 
Educationb   
 Low 786 (78.4) 777 (73.3) 
 Medium 120 (12.0) 185 (17.5) 
 High 97 (9.7) 98 (9.3) 
BMI (kg/m2  
 <25 136 (13.8) 147 (13.4) 
 25 to <30 444 (44.9) 430 (39.3) 
 30 to <35 284 (28.7) 338 (30.9) 
 35 to <40 87 (8.8) 108 (9.9) 
 ≥40 37 (3.7) 72 (6.6) 
Smoking   
 Never smoker 490 (48.1) 528 (48.5) 
 Former smoker 441 (43.3) 433 (39.8) 
 Current smoker, 0–15 g tobacco/day 44 (4.3) 63 (5.8) 
 Current smoker, >15 g tobacco/day 43 (4.2) 64 (5.9) 
Alcohol consumptionc   
 Abstainer 436 (44.0) 400 (36.6) 
 Low consumption 509 (51.4) 625 (57.1) 
 Moderate consumption 32 (3.2) 47 (4.3) 
 High consumption 13 (1.3) 22 (2.0) 
Physical activityd   
 Inactive 549 (55.3) 265 (24.8) 
 Active 444 (44.7) 802 (75.2) 
Diabetes medication   
 None 335 (32.7) 252 (24.9) 
 Oral antidiabetics 495 (48.3) 539 (53.2) 
 Insulin 194 (19.0) 222 (21.9) 
Antihypertensive drugs   
 No 149 (14.6) 174 (17.2) 
 Yes 875 (85.4) 839 (82.8) 
Lipid-lowering drugs   
 No 522 (51.0) 562 (55.5) 
 Yes 502 (49.0) 451 (44.5) 
Anticoagulant drugs   
 No 518 (50.6) 618 (61.0) 
 Yes 506 (49.4) 395 (39.0) 
Time since diabetes diagnosis (years)   
 <5 554 (53.8) 357 (32.6) 
 5 to <10 244 (23.7) 297 (27.1) 
 10 to <20 170 (16.5) 345 (31.5) 
 ≥20 61 (5.9) 97 (8.9) 
HbA1c (%/mmol/mol)   
 <6/<42 222 (21.6) 151 (13.8) 
 6 to <7/42 to <53 482 (46.8) 572 (51.2) 
 7 to <8/53 to <64 249 (24.2) 241 (22.0) 
 ≥8/≥64 76 (7.4) 132 (12.0) 
Total cholesterol (mg/dL)   
 <200 377 (36.6) 387 (35.6) 
 200 to <280 557 (54.1) 561 (51.6) 
 ≥280 95 (9.2) 139 (12.8) 
HDL cholesterol (mg/dL)   
 <40 156 (15.2) 261 (24.0) 
 40 to <80 828 (80.5) 765 (70.4) 
 ≥80 45 (4.4) 60 (5.5) 
CRP (mg/L)   
 ≤3 596 (57.9) 651 (60.0) 
 >3 to ≤10 337 (32.8) 342 (31.5) 
 >10 96 (9.3) 92 (8.5) 
eGFR (mL/min/1.73 m2)e   
 ≥60 739 (71.8) 797 (72.7) 
 30 to <60 278 (27.0) 264 (24.1) 
 <30 12 (1.2) 35 (3.2) 
History of CHD   
 No 724 (70.4) 843 (81.5) 
 Yes 305 (29.6) 191 (18.5) 
History of MI   
 No 901 (87.6) 944 (91.4) 
 Yes 128 (12.4) 89 (8.6) 
History of stroke   
 No 914 (88.8) 971 (93.9) 
 Yes 115 (11.2) 63 (6.1) 
History of heart failure   
 No 813 (79.0) 897 (87.6) 
 Yes 216 (21.0) 127 (12.4) 
History of hypertension   
 No 292 (28.4) 221 (21.4) 
 Yes 737 (71.6) 812 (78.6) 
History of cancer   
 No 893 (86.8) 927 (89.9) 
 Yes 136 (13.2) 104 (10.1) 

aNumbers shown were drawn from the data set without imputed missing values. Therefore, numbers do not always add up to the total because of missing values.

bDefinition of low, medium, and high education levels were ≤9, 10–11, and ≥12 years of school education, respectively.

cDefinition of low alcohol consumption: women >0 to <20 and men >0 to <40 g ethanol/day. Definition of moderate alcohol consumption: women ≥20 to <40 and men ≥40 to <60 g ethanol/day. Definition of high alcohol consumption: women ≥40 and men ≥60 g ethanol/day.

dInactive: 0 h of vigorous physical activity/week; active: >0 h of vigorous physical activity/week.

eThe eGFR was calculated with the creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation (35).

Multiple imputation was used to impute the number of missing baseline covariate values. The proportion of missing values was <5% for all variables with the exception physical activity, which had up to 16% of missing values. To the best of our knowledge, data were missing at random, which is the assumption of the multiple imputation. Separately by cohort and sex, 20 complete data sets were imputed with the SAS 9.3 procedure “PROC MI,” using the Markov chain Monte Carlo method. The variables of model 3 were used for the imputation model. All multivariable analyses were performed in the 20 imputed data sets, and results of the individual data sets were combined by the SAS 9.3 procedure “PROC MIANALYZE.” All analyses were performed with SAS version 9.3 (SAS Institute, Cary, NC), and all statistical tests were two sided using an α level of 0.05.

Baseline characteristics of the included patients with T2DM of the ESTHER and DIANA study are shown in Table 1. The distribution of most baseline characteristics was comparable in the two studies, but there were also differences. The DIANA study participants were more frequently younger than 60 years of age, and were more frequently physically active and had a lower burden of comorbidities (with the exception of hypertension). Patients with T2DM from the ESTHER study had more frequently received a diagnosis of diabetes in the last 5 years, more frequently were taking no diabetes medication, and more frequently had a low HbA1c of <6% (42 mmol/mol).

The correlation of d-ROMs levels and TTLs was low in both cohorts (Spearman correlation coefficients in the ESTHER and DIANA cohorts were −0.09 and −0.15, respectively). Median d-ROMs levels were slightly higher in the DIANA cohort than in the ESTHER cohort and did not vary much by age (Supplementary Table 2). However, consistently statistically significant in both cohorts, d-ROMs levels were increased in females, current smokers, and in patients with T2DM with BMI ≥40 kg/m2, without any antidiabetic medication, with insulin therapy, without lipid-lowering medication, with high total cholesterol levels, or with high CRP levels. Median TTLs were lower in the DIANA cohort than in the ESTHER cohort and decreased with age in both cohorts (Supplementary Table 3). Furthermore, consistently statistically significant in both cohorts, low TTLs were observed in females, alcohol abstainers, and patients with T2DM with BMI ≥40 kg/m2, without any antidiabetic medication, with insulin therapy, with antihypertensive therapy, with anticoagulant medication, with high CRP levels, with estimated glomerular filtration rate (eGFR), or with a history of MI, heart failure, or hypertension.

The associations of preventive drug use and d-ROMs levels as well as TTLs were followed up in detail in Supplementary Tables 4 and 5. The favorable relatively low d-ROMs levels and high TTLs from patients with T2DM who use only oral antidiabetic drugs originates from biguanides use (mainly metformin). The relatively low d-ROMs levels of patients with T2DM who are receiving lipid-lowering therapy can be ascribed to statin use because use of other lipid-lowering medication was rare. Unfavorable low TTLs among antihypertensive drug users was mainly related to diuretics use, which may be explained by low TTLs among heart failure patients that frequently use this drug group. Low TTLs among anticoagulant drug users may also be explained by a severe underlying cardiovascular condition because only users of vitamin K antagonists or heparin had substantially low TTLs.

During follow-up, 96 MCE cases, 97 cancer cases, and 155 deaths were recorded in the ESTHER diabetes subcohort. In the DIANA study, 109 MCE cases, 82 cancer cases, and 239 deaths were identified. The associations of d-ROMs levels, TTLs, and the d-ROMs-to-TTL ratios with these outcomes in the individual cohorts and in random-effects meta-analyses of both cohorts are provided in Table 2. Both oxidative biomarkers and the d-ROMs-to-TTL ratios were strongly and statistically significantly associated with all-cause mortality in both cohorts. In the meta-analysis, a remarkably strong 2.1-fold increased mortality for the comparison of top versus bottom d-ROMs tertile and a 41% reduced all-cause mortality for the comparison of top versus bottom TTL tertile were observed. An even 2.5-fold mortality was observed when comparing patients with T2DM in the top and bottom tertile of the d-ROMs-to-TTL ratio. This strong association for the d-ROMs-to-TTL ratio with all-cause mortality mainly originated from a similarly strong association with cardiovascular mortality, whereas the association with cancer mortality was weaker and not statistically significant (Supplementary Table 6). Only d-ROMs levels, but not TTLs, were statistically significantly associated with cancer mortality.

Table 2

Associations of d-ROMs levels, TTLs, and d-ROMs-to-TTL ratios with all-cause mortality, MCE, and cancer incidence in the ESTHER diabetes subcohort, the DIANA cohort, and in a meta-analysis of both studies

OutcomeOxidative stress markerESTHER
DIANA
Meta-analysis
ModelingNtotalaNcasesHR (95% CI)bModelingNtotalaNcasesHR (95% CI)bHR (95% CI)b
All-cause mortality d-ROMs <320 343 41 Ref <387 365 63 Ref Ref 
  320 to <368 343 54 1.26 (0.79, 2.01) 387 to <450 365 73 1.22 (0.84, 1.76) 1.24 (0.92, 1.65) 
  ≥368 343 60 1.67 (1.05, 2.67) ≥450 366 103 2.49 (1.74, 3.55) 2.10 (1.43, 3.09) 
 TTL <236 342 71 Ref <224 365 113 Ref Ref 
  236 to <297 345 46 0.62 (0.41, 0.94) 224 to <269 364 67 0.69 (0.50, 0.94) 0.66 (0.52, 0.85) 
  ≥297 342 38 0.46 (0.28, 0.75) ≥269 367 59 0.69 (0.49, 0.96) 0.59 (0.40, 0.87) 
 d-ROMs- <1.11 343 34 Ref <1.50 365 51 Ref Ref 
 to-TTL- 1.11 to <1.51 343 55 1.58 (0.93, 2.68) 1.50 to <1.94 365 80 1.52 (1.04, 2.22) 1.54 (1.13, 2.10) 
 ratio ≥1.51 343 66 2.56 (1.56, 4.20) ≥1.94 366 108 2.47 (1.71, 3.57) 2.50 (1.86, 3.36) 
MCE d-ROMs <317 271 30 Ref <387 302 30 Ref Ref 
  317 to <367 270 30 0.93 (0.53, 1.64) 387 to <449 303 38 1.20 (0.71, 2.03) 1.07 (0.73, 1.57) 
  ≥367 271 36 1.08 (0.58, 2.00) ≥449 304 41 1.56 (0.85, 2.87) 1.30 (0.84, 2.01) 
 TTL <239 271 31 Ref <226 300 49 Ref Ref 
  239 to <297 271 35 1.22 (0.72, 2.05) 226 to <271 306 35 0.83 (0.53, 1.29) 0.98 (0.67, 1.43) 
  ≥297 270 30 0.89 (0.50, 1.59) ≥271 303 25 0.71 (0.42, 1.20) 0.79 (0.53, 1.16) 
 d-ROMs- <1.10 271 28 Ref <1.50 302 27 Ref Ref 
 to-TTL- 1.10 to <1.47 271 39 1.50 (0.82, 2.72) 1.50 to <1.93 305 33 1.27 (0.73, 2.21) 1.37 (0.91, 2.06) 
 ratio ≥1.47 270 29 1.32 (0.73, 2.38) ≥1.93 302 49 2.05 (1.15, 3.65) 1.65 (1.07, 2.54) 
Cancer d-ROMs <320 298 34 Ref <387 322 28 Ref Ref 
  320 to <367 298 37 0.95 (0.56, 1.64) 387 to <449 324 27 1.32 (0.73, 2.37) 1.10 (0.74, 1.64) 
  ≥367 297 26 0.83 (0.47, 1.48) ≥449 324 27 1.56 (0.86, 2.80) 1.13 (0.61, 2.10) 
 TTL <238 297 34 Ref <225 323 23 Ref Ref 
  238 to <296 299 31 1.10 (0.65, 1.86) 225 to <270 323 25 1.25 (0.68, 2.30) 1.16 (0.78, 1.73) 
  ≥296 297 32 0.79 (0.43, 1.45) ≥270 324 34 1.89 (1.02, 3.52) 1.22 (0.52, 2.87) 
 d-ROMs- <1.12 298 38 Ref <1.49 324 28 Ref Ref 
 to-TTL- 1.12 to <1.47 297 29 0.85 (0.47, 1.54) 1.49 to <1.93 323 32 1.19 (0.67, 2.14) 1.01 (0.67, 1.53) 
 ratio ≥1.47 298 30 0.89 (0.50, 1.58) ≥1.93 323 22 0.98 (0.52, 1.86) 0.93 (0.61, 1.42) 
OutcomeOxidative stress markerESTHER
DIANA
Meta-analysis
ModelingNtotalaNcasesHR (95% CI)bModelingNtotalaNcasesHR (95% CI)bHR (95% CI)b
All-cause mortality d-ROMs <320 343 41 Ref <387 365 63 Ref Ref 
  320 to <368 343 54 1.26 (0.79, 2.01) 387 to <450 365 73 1.22 (0.84, 1.76) 1.24 (0.92, 1.65) 
  ≥368 343 60 1.67 (1.05, 2.67) ≥450 366 103 2.49 (1.74, 3.55) 2.10 (1.43, 3.09) 
 TTL <236 342 71 Ref <224 365 113 Ref Ref 
  236 to <297 345 46 0.62 (0.41, 0.94) 224 to <269 364 67 0.69 (0.50, 0.94) 0.66 (0.52, 0.85) 
  ≥297 342 38 0.46 (0.28, 0.75) ≥269 367 59 0.69 (0.49, 0.96) 0.59 (0.40, 0.87) 
 d-ROMs- <1.11 343 34 Ref <1.50 365 51 Ref Ref 
 to-TTL- 1.11 to <1.51 343 55 1.58 (0.93, 2.68) 1.50 to <1.94 365 80 1.52 (1.04, 2.22) 1.54 (1.13, 2.10) 
 ratio ≥1.51 343 66 2.56 (1.56, 4.20) ≥1.94 366 108 2.47 (1.71, 3.57) 2.50 (1.86, 3.36) 
MCE d-ROMs <317 271 30 Ref <387 302 30 Ref Ref 
  317 to <367 270 30 0.93 (0.53, 1.64) 387 to <449 303 38 1.20 (0.71, 2.03) 1.07 (0.73, 1.57) 
  ≥367 271 36 1.08 (0.58, 2.00) ≥449 304 41 1.56 (0.85, 2.87) 1.30 (0.84, 2.01) 
 TTL <239 271 31 Ref <226 300 49 Ref Ref 
  239 to <297 271 35 1.22 (0.72, 2.05) 226 to <271 306 35 0.83 (0.53, 1.29) 0.98 (0.67, 1.43) 
  ≥297 270 30 0.89 (0.50, 1.59) ≥271 303 25 0.71 (0.42, 1.20) 0.79 (0.53, 1.16) 
 d-ROMs- <1.10 271 28 Ref <1.50 302 27 Ref Ref 
 to-TTL- 1.10 to <1.47 271 39 1.50 (0.82, 2.72) 1.50 to <1.93 305 33 1.27 (0.73, 2.21) 1.37 (0.91, 2.06) 
 ratio ≥1.47 270 29 1.32 (0.73, 2.38) ≥1.93 302 49 2.05 (1.15, 3.65) 1.65 (1.07, 2.54) 
Cancer d-ROMs <320 298 34 Ref <387 322 28 Ref Ref 
  320 to <367 298 37 0.95 (0.56, 1.64) 387 to <449 324 27 1.32 (0.73, 2.37) 1.10 (0.74, 1.64) 
  ≥367 297 26 0.83 (0.47, 1.48) ≥449 324 27 1.56 (0.86, 2.80) 1.13 (0.61, 2.10) 
 TTL <238 297 34 Ref <225 323 23 Ref Ref 
  238 to <296 299 31 1.10 (0.65, 1.86) 225 to <270 323 25 1.25 (0.68, 2.30) 1.16 (0.78, 1.73) 
  ≥296 297 32 0.79 (0.43, 1.45) ≥270 324 34 1.89 (1.02, 3.52) 1.22 (0.52, 2.87) 
 d-ROMs- <1.12 298 38 Ref <1.49 324 28 Ref Ref 
 to-TTL- 1.12 to <1.47 297 29 0.85 (0.47, 1.54) 1.49 to <1.93 323 32 1.19 (0.67, 2.14) 1.01 (0.67, 1.53) 
 ratio ≥1.47 298 30 0.89 (0.50, 1.58) ≥1.93 323 22 0.98 (0.52, 1.86) 0.93 (0.61, 1.42) 

Boldface type indicates statistically significant heterogeneity (P < 0.05). Ncases, incident case numbers; Ntotal, total study population; Ref, reference category.

aNumbers do not always add up to the total because study participants lost to follow-up and subjects with a history of MCE before baseline or a cancer diagnosis before baseline were excluded in the respective analyses on incident MCE and incident cancer.

bAdjusted for age, sex, HbA1c, BMI, education, smoking, alcohol consumption, physical activity, and diabetes medication.

Associations of d-ROMs levels and TTLs with incident MCE pointed in the same direction as for all-cause mortality in both cohorts but were much weaker and not statistically significant (Table 2). However, a statistically significant 1.65-fold increased MCE incidence was observed for the comparison of top versus bottom d-ROMs-to-TTL ratio tertile. The associations of both biomarkers with total cancer incidence were inconsistent in the two cohorts, and no associations were observed in the meta-analyses.

Additional adjustment for diseases (model 2) and CRP (model 3) attenuated all observed effect estimates (Table 3). Although the associations of TTL and the d-ROMs-to-TTL ratio with all-cause mortality remained statistically significant, the associations of d-ROMs levels with all-cause mortality and the d-ROMs-to-TTL ratio with MCE lost statistical significance.

Table 3

Sensitivity analysis for results of the meta-analyses with extended model adjustment

OutcomeOxidative stress markerModelingNtotalaNcasesModel 1 HR (95% CI)bModel 2 HR (95% CI)cModel 3 HR (95% CI)d
All-cause mortality d-ROMs Tertile 1 708 104 Ref Ref Ref 
  Tertile 2 708 127 1.24 (0.92, 1.65) 1.17 (0.87, 1.57) 1.10 (0.81, 1.48) 
  Tertile 3 709 163 2.10 (1.43, 3.09) 1.89 (1.41, 2.55) 1.49 (0.92, 2.42) 
 TTL Tertile 1 707 184 Ref Ref Ref 
  Tertile 2 709 113 0.66 (0.52, 0.85) 0.75 (0.58, 0.97) 0.74 (0.57, 0.96) 
  Tertile 3 709 97 0.59 (0.40, 0.87) 0.65 (0.47, 0.91) 0.70 (0.52, 0.95) 
 d-ROMs- Tertile 1 708 85 Ref Ref Ref 
 to-TTL- Tertile 2 708 135 1.54 (1.13, 2.10) 1.41 (1.02, 1.93) 1.31 (0.95, 1.81) 
 ratio Tertile 3 709 174 2.50 (1.86, 3.36) 2.15 (1.57, 2.94) 1.86 (1.35, 2.58) 
MCE d-ROMs Tertile 1 573 60 Ref Ref Ref 
  Tertile 2 573 68 1.07 (0.73, 1.57) 0.99 (0.66, 1.49) 0.92 (0.61, 1.38) 
  Tertile 3 575 77 1.30 (0.84, 2.01) 1.25 (0.79, 1.96) 1.04 (0.65, 1.66) 
 TTL Tertile 1 571 80 Ref Ref Ref 
  Tertile 2 577 70 0.98 (0.67, 1.43) 1.13 (0.64, 2.00) 1.14 (0.66, 1.96) 
  Tertile 3 573 55 0.79 (0.53, 1.16) 0.98 (0.57, 1.69) 1.02 (0.57, 1.81) 
 d-ROMs- Tertile 1 573 55 Ref Ref Ref 
 to-TTL- Tertile 2 576 72 1.37 (0.91, 2.06) 1.23 (0.80, 1.89) 1.18 (0.77, 1.82) 
 ratio Tertile 3 572 78 1.65 (1.07, 2.54) 1.30 (0.64, 2.63) 1.13 (0.55, 2.31) 
Cancer d-ROMs Tertile 1 620 62 Ref Ref Ref 
  Tertile 2 622 64 1.10 (0.74, 1.64) 0.98 (0.64, 1.51) 0.88 (0.57, 1.37) 
  Tertile 3 621 53 1.13 (0.61, 2.10) 0.99 (0.59, 1.65) 0.79 (0.48, 1.31) 
 TTL Tertile 1 620 57 Ref Ref Ref 
  Tertile 2 622 56 1.16 (0.78, 1.73) 1.10 (0.72, 1.70) 1.11 (0.72, 1.70) 
  Tertile 3 621 66 1.22 (0.52, 2.87) 1.21 (0.50, 2.94) 1.23 (0.50, 3.00) 
 d-ROMs- Tertile 1 622 66 Ref Ref Ref 
 to-TTL- Tertile 2 620 61 1.01 (0.67, 1.53) 0.90 (0.58, 1.39) 0.82 (0.53, 1.27) 
 ratio Tertile 3 621 52 0.93 (0.61, 1.42) 0.76 (0.47, 1.24) 0.67 (0.40, 1.13) 
OutcomeOxidative stress markerModelingNtotalaNcasesModel 1 HR (95% CI)bModel 2 HR (95% CI)cModel 3 HR (95% CI)d
All-cause mortality d-ROMs Tertile 1 708 104 Ref Ref Ref 
  Tertile 2 708 127 1.24 (0.92, 1.65) 1.17 (0.87, 1.57) 1.10 (0.81, 1.48) 
  Tertile 3 709 163 2.10 (1.43, 3.09) 1.89 (1.41, 2.55) 1.49 (0.92, 2.42) 
 TTL Tertile 1 707 184 Ref Ref Ref 
  Tertile 2 709 113 0.66 (0.52, 0.85) 0.75 (0.58, 0.97) 0.74 (0.57, 0.96) 
  Tertile 3 709 97 0.59 (0.40, 0.87) 0.65 (0.47, 0.91) 0.70 (0.52, 0.95) 
 d-ROMs- Tertile 1 708 85 Ref Ref Ref 
 to-TTL- Tertile 2 708 135 1.54 (1.13, 2.10) 1.41 (1.02, 1.93) 1.31 (0.95, 1.81) 
 ratio Tertile 3 709 174 2.50 (1.86, 3.36) 2.15 (1.57, 2.94) 1.86 (1.35, 2.58) 
MCE d-ROMs Tertile 1 573 60 Ref Ref Ref 
  Tertile 2 573 68 1.07 (0.73, 1.57) 0.99 (0.66, 1.49) 0.92 (0.61, 1.38) 
  Tertile 3 575 77 1.30 (0.84, 2.01) 1.25 (0.79, 1.96) 1.04 (0.65, 1.66) 
 TTL Tertile 1 571 80 Ref Ref Ref 
  Tertile 2 577 70 0.98 (0.67, 1.43) 1.13 (0.64, 2.00) 1.14 (0.66, 1.96) 
  Tertile 3 573 55 0.79 (0.53, 1.16) 0.98 (0.57, 1.69) 1.02 (0.57, 1.81) 
 d-ROMs- Tertile 1 573 55 Ref Ref Ref 
 to-TTL- Tertile 2 576 72 1.37 (0.91, 2.06) 1.23 (0.80, 1.89) 1.18 (0.77, 1.82) 
 ratio Tertile 3 572 78 1.65 (1.07, 2.54) 1.30 (0.64, 2.63) 1.13 (0.55, 2.31) 
Cancer d-ROMs Tertile 1 620 62 Ref Ref Ref 
  Tertile 2 622 64 1.10 (0.74, 1.64) 0.98 (0.64, 1.51) 0.88 (0.57, 1.37) 
  Tertile 3 621 53 1.13 (0.61, 2.10) 0.99 (0.59, 1.65) 0.79 (0.48, 1.31) 
 TTL Tertile 1 620 57 Ref Ref Ref 
  Tertile 2 622 56 1.16 (0.78, 1.73) 1.10 (0.72, 1.70) 1.11 (0.72, 1.70) 
  Tertile 3 621 66 1.22 (0.52, 2.87) 1.21 (0.50, 2.94) 1.23 (0.50, 3.00) 
 d-ROMs- Tertile 1 622 66 Ref Ref Ref 
 to-TTL- Tertile 2 620 61 1.01 (0.67, 1.53) 0.90 (0.58, 1.39) 0.82 (0.53, 1.27) 
 ratio Tertile 3 621 52 0.93 (0.61, 1.42) 0.76 (0.47, 1.24) 0.67 (0.40, 1.13) 

Boldface type indicates statistically significant heterogeneity (P < 0.05). Ncases, incident case numbers; Ntotal, total study population; Ref, Reference category.

aNumbers do not always add up to the total because study participants lost to follow-up and subjects with a history of MCE before baseline or a cancer diagnosis before baseline were excluded in the respective analyses on incident MCE and incident cancer.

bAdjusted for age, sex, HbA1c, BMI, education, smoking, alcohol consumption, physical activity, and diabetes medication.

cAdjusted for variables of model 1, history of MI, history of stroke, history of CHD, history of heart failure, history of hypertension, eGFR, total and HDL cholesterol, and time since diabetes diagnosis.

dAdjusted for variables of model 2 and CRP.

Due to limited case numbers for the other outcomes, subgroup analyses were only conducted for the outcome “all-cause mortality,” and results are shown in Table 4. Associations of d-ROMs levels with all-cause mortality were particularly strong among males and among patients with T2DM with HbA1c <7% (53 mmol/mol), age <70 years, BMI <30 kg/m2, and a history of CHD. Results for TTLs did not differ much among the subgroups. The comparison of patients with T2DM within tertile 3 and tertile 1 of the d-ROMs-to-TTL ratio showed particularly strongly increased mortality among females (3.8-fold) and among subjects with a history of CHD (3.6-fold) and a history of cancer (3.3-fold). However, sex differences were largely driven by the generally higher d-ROMs levels among females and were less pronounced when sex-specific tertile cutoffs were applied (Supplementary Table 7).

Table 4

Subgroup analyses for the associations of d-ROMs levels, TTLs, and d-ROMs-to-TTL ratios with all-cause mortality: pooled results of the ESTHER diabetes subcohort and DIANA cohort

Stratified byModelingd-ROMs
TTL
d-ROMs-to-TTL ratio
NtotalNcasesHR (95% CI)aNtotalNcasesHR (95% CI)aNtotalNcasesHR (95% CI)a
Sex           
 Men Tertile 1 536 82 Ref 324 100 Ref 530 75 Ref 
 Tertile 2 392 86 1.41 (1.01, 1.98) 356 74 0.72 (0.52, 0.99) 364 92 1.56 (1.09, 2.23) 
 Tertile 3 207 78 2.46 (1.50, 4.03) 455 72 0.59 (0.42, 0.82) 241 79 2.39 (1.69, 3.38) 
 Women Tertile 1 172 22 Ref 383 84 Ref 178 10 Ref 
 Tertile 2 316 41 0.73 (0.41, 1.29) 353 39 0.54 (0.35, 0.83) 344 43 1.91 (0.83, 4.42) 
 Tertile 3 502 85 1.39 (0.85, 2.26) 254 25 0.56 (0.25, 1.24) 468 95 3.77 (1.14, 12.39) 
Age           
 <70 years Tertile 1 345 29 Ref 270 38 Ref 395 30 Ref 
 Tertile 2 338 40 1.19 (0.64, 2.20) 307 34 0.66 (0.40, 1.08) 326 42 1.41 (0.83, 2.42) 
 Tertile 3 320 41 2.52 (1.48, 4.30) 426 38 0.50 (0.30, 0.84) 282 38 2.59 (1.55, 4.32) 
 ≥70 years Tertile 1 363 75 Ref 437 146 Ref 313 55 Ref 
 Tertile 2 370 87 1.26 (0.89, 1.80) 402 79 0.59 (0.44, 0.78) 382 93 1.64 (1.12, 2.41) 
 Tertile 3 389 122 2.04 (1.30, 3.19) 283 59 0.50 (0.36, 0.70) 427 136 2.69 (1.86, 3.88) 
BMI (kg/m2          
 <30 Tertile 1 427 59 Ref 344 106 Ref 453 52 Ref 
 Tertile 2 396 80 1.44 (0.99, 2.10) 404 73 0.69 (0.50, 0.94) 381 82 1.68 (1.13, 2.48) 
 Tertile 3 354 97 2.51 (1.74, 3.63) 429 57 0.57 (0.40, 0.83) 343 102 2.49 (1.71, 3.63) 
 ≥30 Tertile 1 281 45 Ref 363 78 Ref 255 33 Ref 
 Tertile 2 312 47 1.15 (0.72, 1.82) 305 40 0.61 (0.41, 0.93) 327 53 1.47 (0.89, 2.41) 
 Tertile 3 355 66 1.70 (0.84, 3.44) 280 40 0.72 (0.46, 1.10) 366 72 2.45 (1.52, 3.95) 
HbA1c (%/mmol/mol)           
 <7/<53 Tertile 1 491 67 Ref 445 106 Ref 505 61 Ref 
 Tertile 2 469 81 1.37 (0.94, 2.01) 497 80 0.76 (0.56, 1.03) 465 84 1.73 (1.18, 2.52) 
 Tertile 3 466 102 2.36 (1.64, 3.40) 484 64 0.63 (0.44, 0.90) 456 105 2.40 (1.65, 3.49) 
 ≥7/≥53 Tertile 1 217 37 Ref 262 78 Ref 203 24 Ref 
 Tertile 2 239 46 1.16 (0.63, 2.13) 212 33 0.44 (0.28, 0.70) 243 51 1.30 (0.75, 2.25) 
 Tertile 3 243 61 1.84 (0.94, 3.59) 225 33 0.54 (0.34, 0.85) 253 69 2.82 (1.70, 4.69) 
History of CHD           
 No Tertile 1 516 69 Ref 493 109 Ref 548 59 Ref 
 Tertile 2 524 82 1.10 (0.77, 1.58) 506 64 0.66 (0.48, 0.91) 526 93 1.60 (1.11, 2.32) 
 Tertile 3 527 95 1.98 (1.38, 2.84) 568 73 0.66 (0.47, 0.92) 493 94 2.18 (1.51, 3.15) 
 Yes Tertile 1 172 35 Ref 186 67 Ref 146 26 Ref 
 Tertile 2 162 39 1.47 (0.87, 2.48) 181 48 0.66 (0.42, 1.03) 160 38 1.20 (0.62, 2.30) 
 Tertile 3 162 64 2.63 (1.41, 4.87) 129 23 0.45 (0.14, 1.43) 190 74 3.62 (2.05, 6.40) 
History of cancer           
 No Tertile 1 637 88 Ref 598 140 Ref 635 72 Ref 
 Tertile 2 622 102 1.28 (0.92, 1.76) 629 94 0.67 (0.51, 0.88) 625 111 1.43 (1.02, 2.00) 
 Tertile 3 603 126 2.07 (0.95, 4.52) 635 82 0.65 (0.48, 0.89) 602 133 2.38 (1.71, 3.29) 
 Yes Tertile 1 71 16 Ref 108 44 Ref 73 13 Ref 
 Tertile 2 86 25 1.00 (0.49, 2.03) 80 19 0.75 (0.38, 1.48) 83 24 1.89 (0.77, 4.62) 
 Tertile 3 105 37 1.60 (0.66, 3.87) 74 15 0.58 (0.23, 1.48) 106 41 3.32 (1.44, 7.63) 
Stratified byModelingd-ROMs
TTL
d-ROMs-to-TTL ratio
NtotalNcasesHR (95% CI)aNtotalNcasesHR (95% CI)aNtotalNcasesHR (95% CI)a
Sex           
 Men Tertile 1 536 82 Ref 324 100 Ref 530 75 Ref 
 Tertile 2 392 86 1.41 (1.01, 1.98) 356 74 0.72 (0.52, 0.99) 364 92 1.56 (1.09, 2.23) 
 Tertile 3 207 78 2.46 (1.50, 4.03) 455 72 0.59 (0.42, 0.82) 241 79 2.39 (1.69, 3.38) 
 Women Tertile 1 172 22 Ref 383 84 Ref 178 10 Ref 
 Tertile 2 316 41 0.73 (0.41, 1.29) 353 39 0.54 (0.35, 0.83) 344 43 1.91 (0.83, 4.42) 
 Tertile 3 502 85 1.39 (0.85, 2.26) 254 25 0.56 (0.25, 1.24) 468 95 3.77 (1.14, 12.39) 
Age           
 <70 years Tertile 1 345 29 Ref 270 38 Ref 395 30 Ref 
 Tertile 2 338 40 1.19 (0.64, 2.20) 307 34 0.66 (0.40, 1.08) 326 42 1.41 (0.83, 2.42) 
 Tertile 3 320 41 2.52 (1.48, 4.30) 426 38 0.50 (0.30, 0.84) 282 38 2.59 (1.55, 4.32) 
 ≥70 years Tertile 1 363 75 Ref 437 146 Ref 313 55 Ref 
 Tertile 2 370 87 1.26 (0.89, 1.80) 402 79 0.59 (0.44, 0.78) 382 93 1.64 (1.12, 2.41) 
 Tertile 3 389 122 2.04 (1.30, 3.19) 283 59 0.50 (0.36, 0.70) 427 136 2.69 (1.86, 3.88) 
BMI (kg/m2          
 <30 Tertile 1 427 59 Ref 344 106 Ref 453 52 Ref 
 Tertile 2 396 80 1.44 (0.99, 2.10) 404 73 0.69 (0.50, 0.94) 381 82 1.68 (1.13, 2.48) 
 Tertile 3 354 97 2.51 (1.74, 3.63) 429 57 0.57 (0.40, 0.83) 343 102 2.49 (1.71, 3.63) 
 ≥30 Tertile 1 281 45 Ref 363 78 Ref 255 33 Ref 
 Tertile 2 312 47 1.15 (0.72, 1.82) 305 40 0.61 (0.41, 0.93) 327 53 1.47 (0.89, 2.41) 
 Tertile 3 355 66 1.70 (0.84, 3.44) 280 40 0.72 (0.46, 1.10) 366 72 2.45 (1.52, 3.95) 
HbA1c (%/mmol/mol)           
 <7/<53 Tertile 1 491 67 Ref 445 106 Ref 505 61 Ref 
 Tertile 2 469 81 1.37 (0.94, 2.01) 497 80 0.76 (0.56, 1.03) 465 84 1.73 (1.18, 2.52) 
 Tertile 3 466 102 2.36 (1.64, 3.40) 484 64 0.63 (0.44, 0.90) 456 105 2.40 (1.65, 3.49) 
 ≥7/≥53 Tertile 1 217 37 Ref 262 78 Ref 203 24 Ref 
 Tertile 2 239 46 1.16 (0.63, 2.13) 212 33 0.44 (0.28, 0.70) 243 51 1.30 (0.75, 2.25) 
 Tertile 3 243 61 1.84 (0.94, 3.59) 225 33 0.54 (0.34, 0.85) 253 69 2.82 (1.70, 4.69) 
History of CHD           
 No Tertile 1 516 69 Ref 493 109 Ref 548 59 Ref 
 Tertile 2 524 82 1.10 (0.77, 1.58) 506 64 0.66 (0.48, 0.91) 526 93 1.60 (1.11, 2.32) 
 Tertile 3 527 95 1.98 (1.38, 2.84) 568 73 0.66 (0.47, 0.92) 493 94 2.18 (1.51, 3.15) 
 Yes Tertile 1 172 35 Ref 186 67 Ref 146 26 Ref 
 Tertile 2 162 39 1.47 (0.87, 2.48) 181 48 0.66 (0.42, 1.03) 160 38 1.20 (0.62, 2.30) 
 Tertile 3 162 64 2.63 (1.41, 4.87) 129 23 0.45 (0.14, 1.43) 190 74 3.62 (2.05, 6.40) 
History of cancer           
 No Tertile 1 637 88 Ref 598 140 Ref 635 72 Ref 
 Tertile 2 622 102 1.28 (0.92, 1.76) 629 94 0.67 (0.51, 0.88) 625 111 1.43 (1.02, 2.00) 
 Tertile 3 603 126 2.07 (0.95, 4.52) 635 82 0.65 (0.48, 0.89) 602 133 2.38 (1.71, 3.29) 
 Yes Tertile 1 71 16 Ref 108 44 Ref 73 13 Ref 
 Tertile 2 86 25 1.00 (0.49, 2.03) 80 19 0.75 (0.38, 1.48) 83 24 1.89 (0.77, 4.62) 
 Tertile 3 105 37 1.60 (0.66, 3.87) 74 15 0.58 (0.23, 1.48) 106 41 3.32 (1.44, 7.63) 

Boldface type indicates statistically significant heterogeneity (P < 0.05). Ncases, incident case numbers; Ntotal, total study population; Ref, reference category.

aAdjusted for age, sex, HbA1c, BMI, education, smoking, alcohol consumption, physical activity, and diabetes medication if not stratified for this variable.

In this analysis of two cohort studies of patients with T2DM from Germany, the serum oxidative stress markers d-ROMs levels, TTLs, and their ratios were consistently associated with all-cause mortality. Moreover, the d-ROMs-to-TTL ratio was significantly associated with incident MCE, but this association did not persist after adjustment for diseases and CRP. No association was observed between oxidative stress biomarkers and total cancer incidence.

Several observations in our study lend support to the interpretation that an imbalanced redox system is likely to be on the causal chain from morbidity to mortality in patients with T2DM. First, the associations of both oxidative stress markers with mortality were attenuated when models were adjusted for chronic diseases and observed associations with incident MCE disappeared (Table 2). This attenuation was even more prominent after adjusting for CRP, which is known to be increased in subjects with chronic diseases and was shown to be associated with both oxidative stress markers (Supplementary Tables 2 and 3). Second, associations of both oxidative stress markers with mortality were particularly strong among patients with CHD or a history of cancer (Table 4). Third, TTLs were particularly low in subjects with any kind of chronic disease that we assessed (Supplementary Table 3).

Contrary to TTLs, d-ROMs levels were not associated with diseases but rather with lifestyle factors (smoking, obesity, and total cholesterol) (Supplementary Table 3). This may explain the low correlation between the two oxidative stress markers and the independent associations of the two with mortality that enabled us to introduce the d-ROMs-to-TTL ratio, which showed stronger results compared with d-ROMs or TTL alone.

To the best of our knowledge, there is no previous study that assessed the predictive values of d-ROMs levels or TTLs for MCE, cancer or all-cause mortality in patients with T2DM. However, several cohort studies have investigated the associations of d-ROMs levels or TTLs with adverse end points in patients with cardiovascular disease (CVD) or in general population samples. Previously, our group explored the associations of d-ROMs levels and TTLs with all-cause mortality, CVD mortality, cancer mortality, MI, and stroke using pooled data from the ESTHER and HAPIEE (Health, Alcohol and Psychosocial Factors in Eastern Europe) studies (11,18). In agreement with the results reported now for patients with T2DM, we observed that both d-ROMs levels and TTLs were independently and strongly associated with all-cause mortality in the general population (11). Furthermore, in line with the results for patients with T2DM, results for incident CVD (MI and stroke) were much stronger for fatal than for nonfatal events (18). With respect to cancer incidence, we recently published pooled results for d-ROMs measurements from two population-based studies (ESTHER and Tromsø studies) (19). d-ROMs levels were associated with lung, colorectal, and breast cancer incidence but not with prostate cancer incidence. Low case numbers for site-specific cancers precluded such a detailed analysis in the two T2DM cohorts and the composite outcome total cancer incidence used was presumably too unspecific to observe any associations with oxidative stress biomarkers. With respect to cohorts with CVD patients, Masaki et al. (20) investigated the prognostic value of the d-ROMs test for CVD events in 265 patients. They reported that d-ROMs levels >395 Carr U were associated with a composite outcome of CVD events. Vassalle et al. (21) conducted a cohort analysis in 93 patients with coronary artery disease and observed increased risks for MCE and all-cause mortality at d-ROMs levels >481 Carr U.

In addition, several studies investigated the associations of oxidatively generated DNA damage (8-oxo-2′-deoxyguanine [8-oxodG] or 8-oxo-2′-deoxyguanosine [8-oxodGuo]) or RNA damage (8-oxo-7,8-dihydroguanosine [8-oxoGuo]) with diabetes complications. A large case-cohort study with prevalent patients with T2DM of the ADVANCE trial observed that plasma 8-oxodGuo levels were associated with all-cause and CVD mortality but not with nonfatal MI or nonfatal stroke (22). Moreover, several prospective cohort studies reported that urinary 8-oxoGuo levels were associated with all-cause mortality (2325) and CVD mortality (25) in patients with T2DM, and that changes in 8-oxoGuo levels during the first 6 years after the diagnosis of T2DM were associated with mortality. With respect to cancer, in accordance with our findings, a cohort study with 1,381 patients with diabetes observed no association of 8-oxodG or 8-oxodGuo with total cancer incidence (26). However, this study 8-oxodG and breast cancer, which deserves further investigation. Larger studies with patients with T2DM are needed to investigate the potential associations of oxidative stress biomarkers and site-specific cancer end points. Taken together, there is accumulating evidence for an important role of oxidative stress in the premature mortality of patients with T2DM, whereas it is still unclear whether oxidative stress is involved in the development of CVD and cancer in patients with T2DM.

Interestingly and possibly explaining this phenomenon, Vassalle et al. (27) observed that elevated levels of lipid peroxidation products and reduced antioxidant capacity were associated with the severity of coronary artery disease and the occurrence and number of different atherogenic risk factors. This is in line with our hypothesis that an imbalanced redox system is on the pathway from morbidity to mortality. Thus, patients with T2DM are at a particularly high risk of fatal events due to high oxidative stress because they often have multimorbidity. Furthermore, hyperglycemia can increase oxidative stress through several pathways. A major mechanism appears to be the overproduction of the superoxide anion (O2·) by the mitochondrial electron transport chain in patients with diabetes (28). In addition to enhancing ROS production, hyperglycemia may also weaken the antioxidant defense systems by reducing the number of antioxidant proteins with free thiol groups (29). Interestingly, TTLs were lower in patients with T2DM with HbA1c ≥8% (64 mmol/mol) than among those with HbA1c <8% (64 mmol/mol) in the ESTHER diabetes subcohort and the DIANA study (Supplementary Table 3).

Until now, pharmacological treatments for high oxidative stress among patients with T2DM are not available and classical antioxidants like vitamin E or C are not helpful (30). Our results for associations of preventive drugs with d-ROMs levels and TTLs suggest that metformin and statins may be potent drugs against high oxidative stress in patients with T2DM. Previous studies (31,32) showed that metformin is able to scavenge hydroxyl radicals (·OH) and enhance TTLs. Statins lower total cholesterol levels, and, as the lipid oxidation biomarker d-ROMs level strongly depends on the total cholesterol concentration (Supplementary Table 2), it is not surprising that statin users have lower d-ROMs levels. However, this striking relationship may not be transferred to other biomarkers of oxidative stress. Rasmussen et al. (33) did not observe any alteration in 8-oxoGuo levels after short-term simvastatin treatment in healthy volunteers, as did Kjaer et al. (24). In addition, Gaede et al. (34) did not observe differences in 8-oxoGuo levels in patients with T2DM with microalbuminuria after 7.8 years of an intensified or conventional multifactorial treatment. The intensified treatment regimen included medication to reach strict HbA1c, blood lipid, and blood pressure targets and motivation for lifestyle changes. However, 8-oxoGuo level was a useful biomarker for mortality prediction within the intensified intervention group (24). Although the search for effective treatments for high oxidative stress should go on, our study can recommend using d-ROMs and TTL measurements as end points in these trials. Especially a trial, which tests the effects of metformin therapy on oxidative stress biomarkers would be of interest. Furthermore, these biomarkers appear to be very useful for mortality prediction in patients with T2DM, and they should be used in further studies with personalized medicine approaches that aim to identify patients that profit most from specific diabetes treatments.

This study has several limitations and strengths. The large sample size and the repeated measurements of the oxidative stress biomarkers are strengths of this study. The observational nature of this study is a limitation, and residual confounding cannot be totally excluded. In addition, a low number of cancer cases did not allow cancer site–specific analyses. Moreover, we cannot exclude that there are some patients with type 1 diabetes in the ESTHER diabetes subcohort because 247 patients with diabetes did not report their diabetes type. Finally, we would like to state that our results can only be generalized to Caucasian populations.

Conclusion

In this meta-analysis of two cohort studies with patients with T2DM from Germany, strong associations of the serum oxidative stress markers d-ROMs levels and TTLs with mortality were observed. Several findings pointed in the direction that an imbalanced redox system is on the causal chain from morbidity to mortality in patients with T2DM. Future studies should evaluate interventions against high oxidative stress in patients with T2DM and may use these biomarkers as end points. Furthermore, they may be helpful for personalized medicine by identifying patients with diabetes with a particularly high risk of premature death.

Funding. This project was funded by the German Research Foundation (grant #SCHO 1545/3-1) and a scholarship from the China Scholarship Council (201606090184). The ESTHER study was funded by grants from the Saarland state Ministry for Social Affairs, Health, Women and Family Affairs (Saarbrücken, Germany), the Baden-Württemberg state Ministry of Science, Research and Arts (Stuttgart, Germany), the Federal Ministry of Education and Research (Berlin, Germany) (grant #01GX0746), and the Federal Ministry of Family Affairs, Senior Citizens, Women and Youth (Berlin, Germany). The DIANA study was funded by a grant from the Federal Ministry of Education and Research (Berlin, Germany).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.X. contributed to conception of the article and interpretation of the results, drafted the manuscript, and approved the final version of the manuscript. X.G. and A.A. contributed to statistical analysis and conception of the article and approved the final version of the manuscript. B.H., E.H.J.M.J., D.C.M., and H.B. contributed to acquisition of the data and conception of the article and approved the final version of the manuscript. B.S. contributed to conception of the article and interpretation of the results, drafted the manuscript, and approved the final version of the manuscript. B.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Stephens
JW
,
Khanolkar
MP
,
Bain
SC
.
The biological relevance and measurement of plasma markers of oxidative stress in diabetes and cardiovascular disease
.
Atherosclerosis
2009
;
202
:
321
329
[PubMed]
2.
Tangvarasittichai
S
.
Oxidative stress, insulin resistance, dyslipidemia and type 2 diabetes mellitus
.
World J Diabetes
2015
;
6
:
456
480
[PubMed]
3.
Paneni
F
,
Beckman
JA
,
Creager
MA
,
Cosentino
F
.
Diabetes and vascular disease: pathophysiology, clinical consequences, and medical therapy: part I
.
Eur Heart J
2013
;
34
:
2436
2443
[PubMed]
4.
Giacco
F
,
Brownlee
M
.
Oxidative stress and diabetic complications
.
Circ Res
2010
;
107
:
1058
1070
[PubMed]
5.
Negre-Salvayre
A
,
Salvayre
R
,
Augé
N
,
Pamplona
R
,
Portero-Otín
M
.
Hyperglycemia and glycation in diabetic complications
.
Antioxid Redox Signal
2009
;
11
:
3071
3109
[PubMed]
6.
Giovannucci
E
,
Harlan
DM
,
Archer
MC
, et al
.
Diabetes and cancer: a consensus report
.
Diabetes Care
2010
;
33
:
1674
1685
[PubMed]
7.
Kotani
K
,
Sakane
N
.
C-reactive protein and reactive oxygen metabolites in subjects with metabolic syndrome
.
J Int Med Res
2012
;
40
:
1074
1081
[PubMed]
8.
Marrocco
I
,
Altieri
F
,
Peluso
I
.
Measurement and clinical significance of biomarkers of oxidative stress in humans
.
Oxid Med Cell Longev
2017
;
2017
:6501046
[PubMed]
9.
Cesarone
MR
,
Belcaro
G
,
Carratelli
M
, et al
.
A simple test to monitor oxidative stress
.
Int Angiol
1999
;
18
:
127
130
[PubMed]
10.
Hu
ML
.
Measurement of protein thiol groups and glutathione in plasma
.
Methods Enzymol
1994
;
233
:
380
385
[PubMed]
11.
Schöttker
B
,
Brenner
H
,
Jansen
EH
, et al
.
Evidence for the free radical/oxidative stress theory of ageing from the CHANCES consortium: a meta-analysis of individual participant data
.
BMC Med
2015
;
13
:
300
[PubMed]
12.
Li
C
,
Zhang
WJ
,
Choi
J
,
Frei
B
.
Quercetin affects glutathione levels and redox ratio in human aortic endothelial cells not through oxidation but formation and cellular export of quercetin-glutathione conjugates and upregulation of glutamate-cysteine ligase
.
Redox Biol
2016
;
9
:
220
228
[PubMed]
13.
Schöttker
B
,
Haug
U
,
Schomburg
L
, et al
.
Strong associations of 25-hydroxyvitamin D concentrations with all-cause, cardiovascular, cancer, and respiratory disease mortality in a large cohort study
.
Am J Clin Nutr
2013
;
97
:
782
793
[PubMed]
14.
Laakso
M
,
Pyörälä
K
.
Age of onset and type of diabetes
.
Diabetes Care
1985
;
8
:
114
117
[PubMed]
15.
Krämer
HU
,
Raum
E
,
Rüter
G
, et al
.
Gender disparities in diabetes and coronary heart disease medication among patients with type 2 diabetes: results from the DIANA study
.
Cardiovasc Diabetol
2012
;
11
:
88
[PubMed]
16.
Raum
E
,
Krämer
HU
,
Rüter
G
, et al
.
Medication non-adherence and poor glycaemic control in patients with type 2 diabetes mellitus
.
Diabetes Res Clin Pract
2012
;
97
:
377
384
[PubMed]
17.
Jansen
E
,
Beekhof
P
,
Cremers
J
, et al
.
Short-term stability of biomarkers of oxidative stress and antioxidant status in human serum
.
ISRN Biomarkers
2013
;
2013
:316528
18.
Xuan
Y
,
Bobak
M
,
Anusruti
A
, et al
.
Association of serum markers of oxidative stress with myocardial infarction and stroke: pooled results from four large European cohort studies
.
Eur J Epidemiol
2019
;
34
:
471
481
[PubMed]
19.
Gào
X
,
Wilsgaard
T
,
Jansen
EH
, et al
.
Pre-diagnostic derivatives of reactive oxygen metabolites and the occurrence of lung, colorectal, breast and prostate cancer: an individual participant data meta-analysis of two large population-based studies
.
Int J Cancer
2019
;
145
:
49
57
[PubMed]
20.
Masaki
N
,
Sato
A
,
Horii
S
, et al
.
Usefulness of the d-ROMs test for prediction of cardiovascular events
.
Int J Cardiol
2016
;
222
:
226
232
[PubMed]
21.
Vassalle
C
,
Bianchi
S
,
Bianchi
F
,
Landi
P
,
Battaglia
D
,
Carpeggiani
C
.
Oxidative stress as a predictor of cardiovascular events in coronary artery disease patients
.
Clin Chem Lab Med
2012
;
50
:
1463
1468
[PubMed]
22.
Thomas
MC
,
Woodward
M
,
Li
Q
, et al.;
ADVANCE Collaborative Group
.
Relationship between plasma 8-OH-deoxyguanosine and cardiovascular disease and survival in type 2 diabetes mellitus: results from the ADVANCE trial
.
J Am Heart Assoc
2018
;
7
:e008226
[PubMed]
23.
Broedbaek
K
,
Siersma
V
,
Henriksen
T
, et al
.
Urinary markers of nucleic acid oxidation and long-term mortality of newly diagnosed type 2 diabetic patients
.
Diabetes Care
2011
;
34
:
2594
2596
[PubMed]
24.
Kjaer
LK
,
Oellgaard
J
,
Henriksen
T
,
Gaede
P
,
Pedersen
O
,
Poulsen
HE
.
Indicator of RNA oxidation in urine for the prediction of mortality in patients with type 2 diabetes and microalbuminuria: a post-hoc analysis of the Steno-2 trial
.
Free Radic Biol Med
2018
;
129
:
247
255
[PubMed]
25.
Kjær
LK
,
Cejvanovic
V
,
Henriksen
T
, et al
.
Cardiovascular and all-cause mortality risk associated with urinary excretion of 8-oxoGuo, a biomarker for RNA oxidation, in patients with type 2 diabetes: a prospective cohort study
.
Diabetes Care
2017
;
40
:
1771
1778
[PubMed]
26.
Broedbaek
K
,
Siersma
V
,
Henriksen
T
, et al
.
Urinary markers of nucleic acid oxidation and cancer in type 2 diabetes
.
Redox Biol
2015
;
4
:
34
39
[PubMed]
27.
Vassalle
C
,
Petrozzi
L
,
Botto
N
,
Andreassi
MG
,
Zucchelli
GC
.
Oxidative stress and its association with coronary artery disease and different atherogenic risk factors
.
J Intern Med
2004
;
256
:
308
315
[PubMed]
28.
Connelly
KA
,
Advani
A
,
Advani
SL
, et al
.
Impaired cardiac anti-oxidant activity in diabetes: human and correlative experimental studies
.
Acta Diabetol
2014
;
51
:
771
782
[PubMed]
29.
Dincer
A
,
Onal
S
,
Timur
S
,
Zeytunluoglu
A
,
Duman
E
,
Zihnioglu
F
.
Differentially displayed proteins as a tool for the development of type 2 diabetes
.
Ann Clin Biochem
2009
;
46
:
306
310
[PubMed]
30.
Ceriello
A
,
Testa
R
.
Antioxidant anti-inflammatory treatment in type 2 diabetes
.
Diabetes Care
2009
;
32
(
Suppl. 2
):
S232
S236
[PubMed]
31.
Bonnefont-Rousselot
D
,
Raji
B
,
Walrand
S
, et al
.
An intracellular modulation of free radical production could contribute to the beneficial effects of metformin towards oxidative stress
.
Metabolism
2003
;
52
:
586
589
[PubMed]
32.
Chakraborty
A
,
Chowdhury
S
,
Bhattacharyya
M
.
Effect of metformin on oxidative stress, nitrosative stress and inflammatory biomarkers in type 2 diabetes patients
.
Diabetes Res Clin Pract
2011
;
93
:
56
62
[PubMed]
33.
Rasmussen
ST
,
Andersen
JT
,
Nielsen
TK
, et al
.
Simvastatin and oxidative stress in humans: a randomized, double-blinded, placebo-controlled clinical trial
.
Redox Biol
2016
;
9
:
32
38
[PubMed]
34.
Gaede
P
,
Lund-Andersen
H
,
Parving
HH
,
Pedersen
O
.
Effect of a multifactorial intervention on mortality in type 2 diabetes
.
N Engl J Med
2008
;
358
:
580
591
[PubMed]
35.
Pugliese
G
,
Solini
A
,
Bonora
E
, et al
.
The Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation provides a better definition of cardiovascular burden associated with CKD than the Modification of Diet in Renal Disease (MDRD) Study formula in subjects with type 2 diabetes
.
Atherosclerosis
2011
;
218
:
194
199
[PubMed]
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.

Supplementary data