Response to Comment on Sharif et al. HDL Cholesterol as a Residual Risk Factor for Vascular Events and All-Cause Mortality in Patients With Type 2 Diabetes. Diabetes Care 2016;39:1424–1430

We appreciate the thoughtful comments and suggestion by Aryan et al. (1) on our recent article on HDL cholesterol (HDL-c) as a residual risk factor for vascular events and all-cause mortality in patients with type 2 diabetes (T2D) (2). The findings of our study in the Second Manifestations of ARTerial disease (SMART) cohort showed that higher HDL-c is related to an unexpected higher risk of cardiovascular events (hazard ratio [HR] 1.10, 95% CI 1.07–1.21) and all-cause mortality (HR 1.14, 95% CI 1.07–1.21) in patients with very low levels of LDL cholesterol (LDL-c) (<2.0 mmol/L). In light of the findings in an ongoing cohort of 2,607 Iranian patients with T2D, Aryan et al. (1) suggest that this unexpected finding can be ascribed to chronic systemic inflammatory status, as they found that higher HDL-c was related to a lower risk of cardiovascular events in T2D patients with very low LDL-c levels (<2.0 mmol/L) after adjusting for hs-CRP (HR 0.771, 95% CI 0.664–0.895).

However, the addition of baseline hs-CRP as a confounder to the multivariable model is debatable. Inflammation is linked to the development of insulin resistance via various cytokines and molecular pathways (3). In turn, insulin resistance leads to dyslipidemia as seen in T2D, characterized by the triad of high triglycerides, low HDL-c, and small dense LDL-c (4,5). As inflammation therefore precedes insulin resistance, it can be argued that hs-CRP is part of the causal pathway in the relation between HDL-c and vascular events and is thus by definition not a confounding factor.

Nonetheless, we performed additional analyses in the strata of LDL-c by adding baseline hs-CRP to the Cox regression model, after imputing 299 missing hs-CRP values. The effect estimates did not change in magnitude or direction (stratum LDL-c <2.0 mmol/L: cardiovascular events HR 1.10, 95% CI 1.02–1.18; all-cause mortality HR 1.14, 95% CI 1.08–1.21). As lipid-lowering therapy intensity might also influence hs-CRP levels (6), we also studied the addition of baseline hs-CRP to the Cox regression model for analysis in the strata of lipid-lowering therapy intensity. Again, adding hs-CRP did not change these results.

We speculate that the contrasting results in SMART cohort and the cohort of Aryan et al. (1) might be attributable to differences in study population and/or medication use. The SMART cohort (7) is a unique cohort consisting of 1,829 high-risk T2D patients of whom 61% receive lipid-lowering treatment (29% usual dose; 32% intensive lipid-lowering therapy) (2). These patients have a median duration of diabetes of 4 years (interquartile range [IQR] 1–9) and the majority (66%) has clinical manifest vascular disease (2). The cohort studied by Aryan et al. (1) seems to consist of a healthier study population compared with the SMART cohort (e.g., age 55 ± 10 years vs. 60 ± 10 years; males 44% vs. 70%). Data on lipid-lowering therapy or vascular disease was not available (8). In the SMART cohort, 82% of the patients in the very low LDL-c stratum received statin treatment, which decreases both hs-CRP and LDL-c levels (6). Therefore, adding hs-CRP to the multivariable model in this stratum probably did not change the results as the inflammation is already substantially reduced by statin treatment (hs-CRP mg/L: LDL-c <2 mmol/L, median 1.9 [IQR 0.8–4.5]; LDL-c 2.0–2.5 mmol/L, median 2.2 [IQR 1.0–5.4]; LDL-c >2.5 mmol/L, median 3.2 [IQR 1.0–7.7]).