Changes in A1C Levels Are Significantly Associated With Changes in Levels of the Cardiovascular Risk Biomarker hs-CRP: Results from the SteP study

Details of the STeP study design have been reported previously (18–20). In brief, this was a 12-month, cluster-randomized clinical trial designed to assess the use of structured SMBG, as part of a comprehensive, collaborative intervention, on glycemic control compared with enhanced usual care in 483 patients with non–insulin-treated T2DM.

The study was conducted in 34 primary care practices across the eastern U.S. Patients were randomized to the structured SMBG group (STG) or to an active control group (ACG) for comparison. STG patients received enhanced usual care and used the Accu-Chek 360 View blood glucose analysis system (Roche Diagnostics, Indianapolis, IN) to record and plot a 7-point SMBG profile (preprandial/postprandial at each meal and at bedtime) on 3 consecutive days before a study visit. STG patients received training in the use of the analysis system and how to interpret their results and use their findings to make changes to their diet and physical activity. STG physicians also received training on interpreting the SMBG data from the device. Enhanced usual care comprised quarterly clinic visits focusing on diabetes management with office point-of-care glycated hemoglobin measurement. ACG patients received enhanced usual care. ACG patients and physicians received no additional training. Free blood glucose meters and test strips (Accu-Chek Aviva blood glucose meter system, Roche Diagnostics) were provided to patients in both study arms. Patient visits occurred at baseline and at months 1, 3, 6, 9, and 12.

The study protocol was approved by the Copernicus Group (Central Institutional Review Board) and is in compliance with the Declaration of Helsinki (21). Written informed consent was obtained from all patients.

As reported previously (18), A1C analysis was conducted by a central laboratory (Covance, Indianapolis, IN) using the Variant II and Variant II Turbo hemoglobin testing systems (Bio-Rad Laboratories, Hercules, CA). Measurements of fasting glucose and postprandial excursions were based on STG patient-reported data from the quarterly 7-point glycemic profiles; accuracy of these data were confirmed using downloaded blood glucose meter data. The 7-point glycemic profiles were also used to assess glycemic variability, which was reported as the MAGE.

Measurements of LDL-cholesterol, HDL-cholesterol, triglycerides, and hs-CRP were taken at baseline and at months 3, 6, 9, and 12. The hs-CRP analysis was also conducted at a central laboratory (Covance). Cutoff points for CV risk were defined as low (mean hs-CRP levels <1.0 mg/L) elevated (mean hs-CRP levels 1.0 to 3 mg/L), and high (mean hs-CRP levels >3.0 mg/L) (22).

Details of the STeP study statistical analysis methodologies have been reported previously (18). Briefly, a cluster-randomization strategy was chosen, whereby all patients within a given practice were assigned to the same study arm. The analysis of change in A1C and other dependent variables was performed using linear mixed models (LMM) analysis with SAS PROC MIXED (23,24). Control variables in all analyses included baseline-dependent variables of patient age, sex, and race (white/nonwhite) as fixed effects; and practice site and subject as random effects. Missing data were estimated using maximum likelihood methods (25). Additional analyses of patient attrition at each step in the protocol also were undertaken.

The primary analysis methods used for this study were similar to those previously reported in the STeP study. LMM analysis was performed for the natural logarithm of hs-CRP at postbaseline visits, with group (ACG or STG), baseline log_e (hs-CRP), age, sex, and race (white/nonwhite) as fixed effects, and patient and site as random effects. The geometric mean estimates at postbaseline visits were derived from the LMM. Geometric mean estimates at baseline were calculated from raw data. The values reported for change from baseline in hs-CRP concentration (mg/L) are absolute differences from baseline in geometric means (95% CI, delta method) at postbaseline visits. Relationships between change in glycemic control, glycemic variability, and change in hs-CRP (log-scale) were examined with a general linear model, with patient demographics, diabetes duration, and BMI as controls. Tests of mediated models via lipids were assessed following the recommendations of Baron and Kenny (26).

As reported previously (18), 13 primary care practices were randomized to the ACG and 21 to the STG. Patient demographic and disease-related characteristics at baseline between the two study groups differed only by age and ethnicity; these differences were controlled in all subsequent analyses (Table 1). Attrition was higher in the STG (n = 81, 28.6%) than in the ACG (n = 43, 18.1%) group. Characteristics of dropout patients did not differ between the two groups.

Geometric mean baseline hs-CRP levels were recorded for 481 patients (Table 1). At baseline, more than 30% of patients were classified as at an elevated CV risk; whereas, almost 60% were found to be at high CV risk, according to hs-CRP level. Patients at high CV risk tended to be younger, more likely to be female, less educated, have higher BMI, shorter diabetes duration, higher diastolic blood pressure, and higher cholesterol levels.

The number of study patients taking statin, β-blocker, and/or ACE-inhibitor medications at baseline was relatively equal among the groups (ACG, n = 157; STG, n = 182) and remained so throughout the study period. Geometric mean (SD) baseline hs-CRP values (mg/L) for these patients were 3.19 (2.76) for ACG patients and 3.67 (2.93) for STG patients. Mean (SD) baseline lipids for the full intent-to-treat cohort were: LDL-cholesterol, 107.1 (42.1) mg/dL; HDL-cholesterol, 44.4 (11.9) mg/dL; and triglycerides, 238.6 (183.6) mg/dL, with no significant between-group differences.

In both study arms, there was a consistent decrease in geometric mean hs-CRP levels over the study duration, which was significantly associated with reductions in A1C observed throughout the study (P < 0.0001; Fig. 1). Reductions in mean hs-CRP values were significantly greater in the STG group at months 3 (P = 0.005), 6 (P = 0.0003), and 12 (P = 0.04) than in the ACG group (Table 2). However, LMM analysis of hs-CRP over time showed no significant difference in change in geometric mean hs-CRP between adherent and nonadherent STG patients: −0.86 vs. −1.01 mg/L, respectively (P = 0.84). STG patients at high CV risk (adherent and nonadherent) showed significantly greater reductions in hs-CRP levels than ACG patients at high CV risk, with the greatest between-group difference seen at month 6: −3.64 (95% CI −4.21 to −3.06) vs. −2.18 (−2.93 to −1.43) mg/dL, respectively (P = 0.002).

Among patients treated with statin, β-blocker, and/or ACE-inhibitors, the STG showed a significantly greater reduction in hs-CRP levels from baseline compared with the ACG: −1.08 (95% CI, −1.39 to −0.77) vs. 0.59 (−0.99 to −0.19) mg/L, respectively (P = 0.049). Although significantly more STG than ACG subjects received a thiazolidinedione (TZD) during the study, 22.7 vs. 12.3%, respectively (P = 0.004), which can influence hs-CRP levels, there was no significant association between use of these medications or change in these medications and change in hs-CRP (P = NS). This finding remained the same even after excluding patients who were receiving TZD therapy at baseline.

Our analysis showed that the reductions in A1C were significantly related to reductions in hs-CRP over time in the sample as a whole and in each of the two study groups: the standardized coefficient (β) was 0.25 for the entire cohort (P < 0.0001), 0.31 for STG (P < 0.0001), and 0.16 for ACG (P = 0.002). These associations were consistent across the study sample and did not vary by patient demographic or diabetes characteristics.

In analyses that included A1C plus MAGE or MAGE alone, changes in MAGE were unrelated to changes in hs-CRP over time (P = 0.82 or P = 0.37). Similar results were recorded when changes in postprandial glucose excursions were included (P = 0.20 or P = 0.99). Hence, changes in glycemic variability, in combination with changes in A1C or alone, did not account for changes in hs-CRP.

We assessed the potential role of changes in lipids as a mediator in the relationship between changes in A1C and changes in hs-CRP over time: there was no reduction in the relationship between changes in A1C and changes in hs-CRP when changes in LDL-cholesterol, HDL-cholesterol, and triglycerides were evaluated in the model individually or as a group, following Baron and Kenny (26). Thus, lipids did not mediate the relationship between changes in A1C and changes in hs-CRP over time.

Our study is the first to report a relationship between the significant glycemic outcomes of a structured SMBG intervention and changes in hs-CRP, a clinical marker of CV risk, with a demonstrated association between inflammation and atherosclerosis (27). A significant positive association was found in both study groups between the degree of reduction of A1C and the degree of reduction in hs-CRP; these improvements were especially significant among STG patients who were at the highest CV risk (hs-CRP >3 mg/L), and their A1C levels dropped the most. Although we were unable to identify the underlying mechanism(s) to explain the relationship between reductions in A1C and hs-CRP, our analyses did rule out several commonly hypothesized mechanisms. For example, postprandial excursions and overall glycemic variability have been linked with oxidative stress (28) and other markers of vascular disease (e.g., carotid intima-media thickening) (29); however, our analyses found no relationship between reduced postprandial glucose excursions or glycemic variability (as measured by MAGE and the magnitude of postprandial glucose excursions) and changes in hs-CRP, nor did HDL-cholesterol, LDL-cholesterol, or triglycerides (individually or combined) mediate the relation between changes in A1C and changes in hs-CRP. Zero-order correlations among these variables also were not significant. Moreover, because the reductions in hs-CRP seen in both study groups were independent of treatment with statins, β-blockers, ACE-inhibitors, and/or TZDs, use of lipid-lowering and/or antihypertensive medications did not appear to be a factor in our findings. However, when one considers the effects of oxidative stress, which is commonly considered to be the link between hyperglycemia and diabetes complications and is believed to be one of the earliest pathophysiologic changes in the inflammatory process that triggers endothelial dysfunction (14), the effects of “metabolic memory” may partially explain the relationship between changes in A1C and changes in hs-CRP.

The concept of metabolic memory hypothesizes that diabetic vascular stresses persist after glycemia has been reduced and that early aggressive treatment aiming to “normalize” metabolic control, as seen in the STeP trial (20), in combination with the agents that reduce cellular reactive species and glycation, may minimize long-term diabetes complications (30). Given that both STeP study groups experienced significantly greater A1C reductions early in the study compared with ACG patients at month 3 and even greater reductions among STG subjects (adherent and nonadherent) at month 6, this early improvement in glycemia possibly conferred a long-term protective effect against oxidative stress that resulted in lower hs-CRP levels even though A1C levels in nonadherent STG patients deteriorated to the same level as in ACG patients at 12 months.

Regardless of the mechanism(s) involved, we showed that reductions in A1C are significantly linked with reductions in CV risk (as assessed by hs-CRP levels) in non–insulin-treated T2DM. Although there are several possible approaches to reducing A1C levels, there is a growing body of evidence demonstrating the effectiveness and practicality of structured SMBG-based interventions in lowering A1C and markers of metabolic risk in this population (31,32). Long-term follow-up, including assessment of compliance, would determine whether the results found in our analyses translate into clinical benefits such as long-term improvement of cardiovascular outcome.

Funding for the STeP study was provided by Roche Diagnostics, Indianapolis, Indiana. Funding for the current STeP study analysis was provided by Roche Diagnostics, Deutschland GmbH, Mannheim, Germany.

I.A.-Z, A.M., and M.A.S. are employed by Roche Diagnostics. O.S., Z.J., J.L.B., C.G.P., L.F., and W.H.P. have worked as consultants for Roche Diagnostics. No other potential conflicts of interest relevant to this article were reported.

O.S., I.A.-Z., A.M., C.G.P., M.A.S., L.F., and W.H.P. contributed to analysis and interpretation of data, drafting the article, revising it critically for important intellectual content, and final approval of the version to be published. Z.J. and J.L.B. performed the statistical analysis, contributed to analysis and interpretation of data, drafted the article, revised it critically for important intellectual content, and gave final approval of the version to be published. O.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.

Parts of this study were presented in poster form at the 71st Scientific Sessions of the American Diabetes Association, San Diego, California, 24–28 June 2011.