Differences in baseline clinical characteristics among the MACE and non‑MACE groups of the study population
Demographics and clinical characteristics of 10,251 T2DM patients were shown in Table 1. Patients enrolled in current study were 62.81 ± 6.65 years old of age on average. Among the 10,251 individuals, 61.45% was male and 38.55% was female. 1826 patients (17.8%) developed MACEs after a median follow-up of 9.7 years.
Table 1
Characteristics of patients among the non-MACE and MACE group.
Characteristics | Total (n = 10251) | Non-MACE(n = 8425) | MACE (n = 1826) | P |
Age, (year) | 62.81 ± 6.65 | 62.52 ± 6.51 | 64.15 ± 7.13 | ༜0.001 |
Gender (%) | | | | ༜0.001 |
Female | 3952 (38.55) | 3389 (40.23) | 563(30.83) | |
Male | 6299 (61.45) | 5036 (59.77) | 1263(61.45) | |
Live alone | 8171 (79.72) | 6735 (79.96) | 1436(78.64) | 0.211 |
Race/ethnicity, n (%) | | | | ༜0.001 |
White | 6393 (62.36) | 5128 (60.87) | 1265(69.28) | |
Non-White | 3858 (37.64) | 3297 (39.13) | 561(30.72) | |
Education, n (%) | | | | 0.002 |
Less than high school | 1521 (14.85) | 1214 (14.42) | 307 (16.84) | |
High school graduate or GED | 2704 (26.40) | 2223 (26.40) | 481 (26.39) | |
Some college | 3357 (32.77) | 2740 (32.54) | 617 (33.85) | |
College degree or higher | 2662 (25.99) | 2244 (26.65) | 418 (22.93) | |
Previous cardiovascular event, n (%) | 3609 (35.21) | 2640 (31.34) | 969 (53.07) | ༜0.001 |
Previous congestive heart failure, n (%) | 494 (4.82) | 327 (3.88) | 167 (9.15) | ༜0.001 |
Previous hyperlipidemia, n (%) | 7165 (69.90) | 5862 (69.58) | 1303 (71.36) | 0.136 |
Previous hypertension, n (%) | 7726 (75.37) | 6301 (74.79) | 1425 (78.04) | 0.003 |
Cigarette-smoking status, n (%) | | | | ༜0.001 |
Current | 1429 (13.94) | 1146 (13.60) | 283 (15.50) | |
Former | 4540 (44.29) | 3664 (43.49) | 876 (47.97) | |
Never | 4282 (41.77) | 3615 (42.91) | 667 (36.53) | |
Weight (kg) | 93.51 ± 18.41 | 93.28 ± 18.400 | 94.58 ± 18.40 | 0.006 |
Body mass index (Kg/cm2) | 32.22 ± 5.40 | 32.21 ± 5.41 | 32.28 ± 5.37 | 0.625 |
Blood pressure (mmHg) | | | | |
Systolic | 136.36 ± 17.11 | 136.00 ± 16.88 | 138.02 ± 18.04 | ༜0.001 |
Diastolic | 74.88 ± 10.66 | 75.14 ± 10.48 | 73.70 ± 11.37 | ༜0.001 |
Medications, n (%) | | | | |
Insulin | 3260 (31.80) | 2559 (30.37) | 701 (38.39) | ༜0.001 |
Metformin | 6554 (63.94) | 5467 (64.90) | 1087 (59.53) | ༜0.001 |
Any sulfonylurea | 5474 (53.40) | 4530 (53.77) | 944 (51.70) | 0.109 |
Any thiazolidinedione | 2258 (22.03) | 1912 (22.70) | 346 (18.95) | ༜0.001 |
ACEI/ARB | 7102 (69.28) | 5835 (69.26) | 1267 (69.39) | 0.933 |
Aspirin | 5579 (54.68) | 4538 (54.12) | 1041 (57.26) | 0.016 |
Statin | 6500 (63.66) | 5314 (63.33) | 1186 (65.16) | 0.147 |
Cholesterol absorption inhibitors | 207 (2.03) | 169 (2.02) | 38 (2.09) | 0.854 |
Niacin and nicotinic acid | 183 (1.79) | 142 (1.69) | 41 (2.26) | 0.118 |
Duration of diabetes (year) | 10.80 ± 7.60 | 10.50 ± 7.42 | 12.18 ± 8.21 | ༜0.001 |
Glycated hemoglobin (%) | 8.30 ± 1.06 | 8.28 ± 1.05 | 8.41 ± 1.09 | ༜0.001 |
Fasting plasma glucose (mg/dL) | 175.19 ± 56.17 | 174.04 ± 55.31 | 180.51 ± 59.72 | ༜0.001 |
Serum creatinine (mg/dL) | 0.91 ± 0.23 | 0.90 ± 0.23 | 0.97 ± 0.25 | ༜0.001 |
eGFR (mL/min/1.73m2) | | | | ༜0.001 |
30–49 mL/min/1.73m2 | 271 (2.64) | 192 (2.28) | 79 (4.33) | |
>50 mL/min/1.73m2 | 9980 (97.36) | 8233 (97.72) | 1747 (95.67) | |
Plasma triglycerides (mmol/L) | 2.13 ± 1.68 | 2.11 ± 1.65 | 2.26 ± 1.81 | 0.001 |
Total plasma cholesterol (mmol/L) | 4.71 ± 1.13 | 4.71 ± 1.12 | 4.76 ± 1.19 | 0.059 |
Plasma LDL-C (mmol/L) | 2.70 ± 0.90 | 2.69 ± 0.89 | 2.74 ± 0.94 | 0.024 |
Plasma HDL-C (mmol/L) | 1.08 ± 0.31 | 1.09 ± 0.31 | 1.03 ± 0.31 | ༜0.001 |
Atherogenic index of plasma (AIP) | 0.54 ± 0.75 | 0.51 ± 0.75 | 0.64 ± 0.74 | ༜0.001 |
ACEI: Angiotensin-converting-enzyme inhibitor; ARB: |
Between patients who developed MACE and patients who did not, there was no significant difference in their living condition (live alone), history of hyperlipidemia, body mass index and prescription record (i.e. sulfonylurea, ACEI/ARB, statin, cholesterol absorption inhibitors, or niacin and nicotinic acid). There was no marked difference in plasma cholesterol level between the MACE group and the non-MACE group, suggesting that it would not be a promising indicator to predict MACE. In comparison to the non-MACE group, traditional risk factors for CVD including old age, male, hypertension and smoking were more prevalent in diabetic patients with MACEs. Patients with MACEs also had significantly larger body weight, higher blood pressure, longer duration of diabetes and a higher incidence of cardiovascular event and congestive heart failure. In addition, they showed significantly higher concentration of fasting plasma glucose, HbA1c, plasma triglycerides and LDL-C than non-MACE individuals. On the other hand, plasma HDL-C was lower in patients with MACEs than those who did not develop MACEs. Subsequently, AIP, which is the marker for abnormal lipid and glucose metabolism and calculated as the ratio between LDL-C and HCL-C on a logarithmic scale, was significantly higher in diabetic patients with MACEs than those without MACEs.
The relationship between AIP and prognosis in patients with T2DM
To explore whether AIP was associated with the poor outcomes of diabetic patient, we obtained an optimal threshold of AIP that would best separate MACE and non-MACE individuals via ROC curve analysis. Our result showed that AIP had an area under curve (AUC) of 0.5512, suggesting that there was an association between AIP and the risk of MACEs (Fig. 1). Furthermore, the optimal cut-off point for AIP was 0.34 obtained from the curve. The study population was thereby assigned to two groups based on AIP: high AIP (greater than or equal to 0.34) and low AIP (less than 0.34).
Next, AIP was assessed as a continuous covariate via univariate cox proportional hazards regression. During follow-up, 1233 of the patients with high AIP developed MACEs while only 593 patients with low AIP had the same outcome (Table 2). Similar results were found in the analysis of secondary outcomes. 1263 patients with high AIP resulted in death from any cause, while such poor outcomes were observed in 695 patients with low AIP (Table 2). AIP was an independent prognostic marker and associated with primary outcomes (HR: 1.383, 95% CI: 1.254–1.525, P < 0.001) and secondary outcomes (all-cause death, HR: 1.205, 95% CI: 1.099–1.322, P < 0.001) in T2DM patients with MACEs (Table 2). More specifically, high AIP presented the highest risk in cardiovascular deaths (HR: 1.500, 95%CI: 1.270–1.765, P < 0.001) and nonfatal myocardial infarction (HR: 1.499, 95%CI: 1.304–1.722, P < 0.001) (Table 2), suggesting that it could be used as a strong predictor of the two outcomes. Kaplan-Meier curves were used to visualize the probability of primary outcomes, the probability of specific cardiovascular events including cardiovascular deaths, nonfatal myocardial infarction, nonfatal strokes (Fig. 2A-D), as well as secondary outcomes, total strokes, and congestive heart failure (Fig. 2E-G). In comparison to patients with low AIP, the probability of poor patient outcomes was significantly higher in the high AIP group (P < 0.05), further illustrating that AIP could be used as a good prognostic marker among patients with T2DM.
Table 2
Univariate cox regression analysis of primary and secondary outcome
Outcomes | Total (n = 10251) | Low AIP (n = 4039) | High AIP (n = 6212) | Univariate |
HR | 95% CI | P |
Primary outcome | 1826 (17.81) | 593 (14.68) | 1233 (19.85) | 1.383 | 1.254–1.525 | ༜0.001 |
Cardiovascular cause death | 669 (6.53) | 205 (5.08) | 464 (7.47) | 1.500 | 1.270–1.765 | ༜0.001 |
Nonfatal myocardial infarction | 936 (9.13) | 287 (7.11) | 649 (10.45) | 1.499 | 1.304–1.722 | ༜0.001 |
Nonfatal stroke | 488 (4.76) | 171 (4.23) | 317 (5.10) | 1.219 | 1.012–1.468 | 0.037 |
Secondary outcomes (Any cause death) | 1958 (19.10) | 695 (17.21) | 1263 (20.33) | 1.205 | 1.099–1.322 | ༜0.001 |
Total stroke | 516 (5.03) | 178 (4.41) | 338 (5.44) | 1.248 | 1.041–1.496 | 0.017 |
Congestive heart failure | 696 (6.79) | 227 (5.62) | 469 (7.55) | 1.372 | 1.171–1.608 | ༜0.001 |
Data are expressed as OR ± 95% CIs (reported in parentheses) as assessed by univariate cox regression analysis. |
Then, the hazard ratio of AIP for different patient outcomes was adjusted for confounding risk factors. Three multivariate regression models were established, each with a different number of confounders taken into consideration. Model 1 was adjusted for age, gender, history of cardiovascular events, smoking, BMI, and duration of diabetes, AIP showed a hazard ratio of 1.333 for MACEs (95% CI: 1.205–1.474, P༜0.001 ) and a lower hazard ratio of 1.184 for all-cause mortality (95% CI: 1.077–1.303, P༜0.001) (Table 3). Model 2 was adjusted for additional variables on top of Model 1, including history congestive heart failure, eGFR, HbA1c, plasma triglycerides, total plasma cholesterol, and plasma HDL-C. Model 3 was based on Model 2, with additional confounders in regard to prescription records, including the use of insulin, biguanide, sulfonylurea, thiazolidinediones, statin, other lipid-lowering medications, niacin, and fibrate. The association between AIP and MACEs remained to be true with a hazard ratio of 1.171 under model 2 and 1.194 under model 3 (Model 2: 95% CI: 1.030–1.333, P = 0.016; Model 3: 95% CI: 1.049–1.360, P = 0.007) (Table 3). Association was also observed between AIP and cardiovascular deaths or nonfatal myocardial infarction (Table 3). However, after adjustment for confounders, AIP showed hazard ratios less than 1 for nonfatal stroke, total stroke, congestive heart failure, and all-cause mortality, representing weak or no association (Table 3).
Table 3
Multivariate cox regression analysis of primary and secondary outcome.
Outcome | Model 1 | Model 2 | Model 3 |
HR | 95% CI | P | HR | 95% CI | P | HR | 95% CI | P |
Primary outcome (MACEs) | 1.333 | 1.205–1.474 | ༜0.001 | 1.171 | 1.030–1.333 | 0.016 | 1.194 | 1.049–1.360 | 0.007 |
Cardiovascular cause death | 1.422 | 1.201–1.683 | ༜0.001 | 1.237 | 0.995–1.538 | 0.056 | 1.264 | 1.015–1.573 | 0.036 |
Nonfatal myocardial infarction | 1.447 | 1.255–1.669 | ༜0.001 | 1.252 | 1.045–1.499 | 0.015 | 1.284 | 1.071–1.539 | 0.007 |
Nonfatal stroke | 1.190 | 0.984–1.441 | 0.073 | 1.078 | 0.841–1.381 | 0.590 | 1.090 | 0.849–1.399 | 0.680 |
Secondary outcomes (all-cause mortality) | 1.184 | 1.077–1.303 | ༜0.001 | 1.037 | 0.917–1.173 | 0.559 | 1.065 | 0.942–1.206 | 0.315 |
Total stroke | 1.232 | 1.023–1.484 | 0.028 | 1.132 | 0.888–1.444 | 0.316 | 1.143 | 0.895–1.459 | 0.284 |
Congestive heart failure | 1.264 | 1.074–1.487 | 0.005 | 1.035 | 0.840–1.276 | 0.746 | 1.017 | 0.823–1.255 | 0.879 |
Data are expressed as OR ± 95% CIs (reported in parentheses) as assessed by multivariate cox regression analysis; HR: hazard ratio; CI: confidence interval. Covariates included in multivariate cox regression models were model 1: age, gender, previous cardiovascular event, smoking, BMI, and duration of diabetes. Model 2: age, gender, previous cardiovascular event, smoking, BMI, duration of diabetes, previous congestive heart failure, eGFR, HbA1c, plasma triglycerides, total plasma cholesterol, and plasma HDL-C. Model 3: age, gender, previous cardiovascular event, smoking, BMI, duration of diabetes, previous congestive heart failure, eGFR, HbA1c, plasma triglycerides, total plasma cholesterol, plasma HDL-C, insulin, biguanide, sulfonylurea, thiazolidinediones, statin, other lipid-lowering medications, niacin, and fibrate. |
The association between AIP and MACEs in the different subgroups of the study population.
Next, to explore the association between AIP and MACEs in more detail, we categorized the study population based on patient demographics and medical records including gender, age, race/ethnicity, history of cardiovascular diseases (CVD), treatment given and trial enrolled, HbA1c level and incidence of depression. Subsequently, we performed stratified analyses to test for interactions and stratified confounders in the association between AIP and MACEs in the different subgroups (Table 4, Supplemental Fig. 1). Our results showed that gender might play roles in the association between AIP and MACEs, leading to a stronger prediction of MACEs by AIP among women. However, we did not detect any interaction among different demographic factors and clinical records in male patients. On the other hand, gender seemed to also interact with the association between AIP and nonfatal myocardial infarction. Taken together, the stratified analyses suggested that AIP might be a stronger prognostic marker among elderly women. Furthermore, we evaluated the association between AIP and nonfatal myocardial infarction and similar results were found across different population subgroups (Table 4, Supplemental Fig. 2).
Table 4
Hazard ratios for the primary outcome and death from any Cause in prespecified Subgroups.
Outcome | Low AIP | High AIP | HRa | 95%CI | P | P for interactionb |
Events/n | % | Events/n | % |
Primary outcome |
Gender | | | | | | | | 0.024 |
Male | 389/2207 | 17.63 | 874/4092 | 21.36 | 1.230 | 1.091–1.386 | 0.001 | |
Female | 204/1832 | 11.14 | 359/2120 | 16.93 | 1.566 | 1.318–1.859 | ༜0.001 | |
Age | | | | | | | | 0.912 |
༜65 | 298/2416 | 12.33 | 690/4073 | 16.94 | 1.426 | 1.245–1.634 | ༜0.001 | |
≥65 | 295/1623 | 18.18 | 543/2139 | 25.39 | 1.410 | 1.224–1.625 | ༜0.001 | |
Race/ethnicity | | | | | | | | 0.557 |
White | 336/1987 | 16.91 | 929/4406 | 21.08 | 1.286 | 1.135–1.457 | ༜0.001 | |
Non-White | 257/2052 | 12.52 | 304/1806 | 16.83 | 1.368 | 1.159–1.615 | ༜0.001 | |
CVD history | | | | | | | | 0.254 |
Yes | 300/1260 | 23.81 | 669/2349 | 28.48 | 1.241 | 1.083–1.422 | 0.002 | |
No | 293/2779 | 10.54 | 564/3863 | 14.60 | 1.391 | 1.208–1.602 | ༜0.001 | |
Glycemia arm | | | | | | | | 0.716 |
Standard | 301/2021 | 14.89 | 629/3102 | 20.28 | 1.408 | 1.227–1.615 | ༜0.001 | |
Intensive | 292/2018 | 14.47 | 604/3110 | 19.42 | 1.358 | 1.181–1.562 | ༜0.001 | |
Trail | | | | | | | | 0.131 |
BP | 311/2295 | 13.55 | 468/2438 | 19.20 | 1.473 | 1.277-1.700 | ༜0.001 | |
Lipid | 282/1744 | 16.17 | 765/3774 | 20.27 | 1.263 | 1.102–1.448 | 0.001 | |
HbA1c | | | | | | | | 0.072 |
༜8.0 | 243/1971 | 12.33 | 548/2898 | 18.91 | 1.553 | 1.335–1.806 | ༜0.001 | |
≥8.0 | 350/2068 | 16.92 | 685/3314 | 2067 | 1.260 | 1.108–1.433 | ༜0.001 | |
Depression | | | | | | | | 0.257 |
Yes | 122/802 | 15.21 | 363/1619 | 22.42 | 1.519 | 1.238–1.865 | ༜0.001 | |
No | 470/3235 | 14.53 | 870/4593 | 18.94 | 1.325 | 1.184–1.482 | ༜0.001 | |
Nonfatal myocardial infarction |
Gender | | | | | | | | 0.064 |
Male | 190/2207 | 8.61 | 457/4092 | 11.17 | 1.318 | 1.112–1.561 | 0.001 | |
Female | 97/1832 | 5.29 | 192/2120 | 9.06 | 1.747 | 1.369–2.230 | ༜0.001 | |
Age | | | | | | | | 0.761 |
༜65 | 146/2416 | 6.04 | 373/4073 | 9.16 | 1.566 | 1.293–1.896 | ༜0.001 | |
≥65 | 141/1623 | 8.69 | 276/2139 | 12.90 | 1.498 | 1.223–1.835 | ༜0.001 | |
Race/ethnicity | | | | | | | | 0.648 |
White | 170/1987 | 8.56 | 501/4406 | 11.37 | 1.362 | 1.144–1.621 | ༜0.001 | |
Non-White | 117/2052 | 5.7 | 148/1806 | 8.19 | 1.462 | 1.147–1.864 | 0.002 | |
CVD history | | | | | | | | 0.158 |
Yes | 152/1260 | 12.06 | 356/2349 | 15.16 | 1.287 | 1.064–1.555 | 0.009 | |
No | 135/2779 | 4.86 | 293/3863 | 7.58 | 1.572 | 1.282–1.928 | ༜0.001 | |
Glycemia arm | | | | | | | | 0.891 |
Standard | 151/2021 | 7.47 | 341/3102 | 10.99 | 1.513 | 1.249–1.832 | ༜0.001 | |
Intensive | 136/2018 | 6.74 | 308/3110 | 9.90 | 1.485 | 1.214–1.817 | ༜0.001 | |
Trail | | | | | | | | 0.025 |
BP | 143/2295 | 6.23 | 257/2438 | 10.54 | 1.747 | 1.424–2.144 | ༜0.001 | |
Lipid | 144/1744 | 8.26 | 392/3774 | 10.39 | 1.266 | 1.046–1.532 | 0.016 | |
HbA1c | | | | | | | | 0.600 |
༜8.0 | 112/1779 | 6.30 | 255/2602 | 9.80 | 1.568 | 1.255–1.958 | ༜0.001 | |
≥8.0 | 175/2260 | 7.74 | 394/3610 | 10.91 | 1.450 | 1.214–1.733 | ༜0.001 | |
Depression | | | | | | | | 0.203 |
Yes | 61/802 | 7.61 | 208/1619 | 12.85 | 1.729 | 1.300–2.300 | ༜0.001 | |
No | 226/3235 | 6.99 | 441/4593 | 9.60 | 1.394 | 1.188–1.637 | ༜0.001 | |
a Low AIP as reference, the Hazard Ratio of High AIP for primary outcome or nonfatal myocardial infarction in each subgroup in gender, age, race/ethnicity, CVD history, Glycemia arm, Trail, HbA1c and Depression. b Interaction between categorical factor AIP and gender, age, race/ethnicity, CVD history, Glycemia arm, Trail, HbA1c and Depression, respectively. |